Method and apparatus for determining a device

Abstract

The application discloses a kind of determination method and device of equipment, it is related to computer field, comprising: obtaining the multiple sets of equipment features that target equipment is decomposed into independent category, by constructing the independent vector space of each group of features, based on preset similarity condition respectively filter out the candidate equipment feature set that each group of features is highly consistent, the complete equipment information of target candidate equipment is generated by intelligent splicing the candidate equipment corresponding to different feature groups, reach the purpose of cross equipment reuse optimal local feature, to realize the significant improvement of matching precision, the effective reduction of artificial design cost, solve the technical problems that equipment matching efficiency is lower in related art, accuracy is difficult to guarantee.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computers, and particularly relates to a device determination method and device. BACKGROUND

[0002] In the technical field of data center device management, matching and deployment of cabinets are key links of device management, which directly affect management efficiency, heat dissipation performance and space utilization. The existing technology usually adopts a single-cabinet global matching scheme, which compares the features of the cabinet as a single entity. However, this scheme has significant limitations when dealing with complex cabinet composite features. Specifically, the features of modern cabinets are multi-dimensional and heterogeneous, and the features are difficult to effectively distinguish in global matching. When a new cabinet has local similarity with a historical cabinet in some features, global matching is likely to cause feature confusion and cannot accurately identify the optimal matching scheme. This reduces the accuracy and reliability of matching, increases the need for manual intervention, and affects the efficiency of data center operation and maintenance. SUMMARY

[0003] The present application provides a device determination method and device to at least solve the problem of low device matching efficiency and difficulty in ensuring accuracy in related technologies.

[0004] The present application provides a device determination method, comprising: obtaining a plurality of sets of device features of a target device, wherein the plurality of sets of device features include a first set of device features and a second set of device features, and the categories of the first set of device features and the second set of device features are different; determining a plurality of sets of candidate device features that satisfy a preset similarity condition based on the plurality of sets of device features, wherein the plurality of sets of device features and the plurality of sets of candidate device features correspond one-to-one, the plurality of sets of candidate device features include a first set of candidate device features that satisfy the preset similarity condition with the first set of device features and a second set of candidate device features that satisfy the preset similarity condition with the second set of device features; determining candidate devices based on the plurality of sets of candidate device features, and combining the candidate devices to generate device information of a target candidate device, wherein the candidate devices include a first candidate device and a second candidate device, the first candidate device corresponds to the first set of candidate device features, and the second candidate device corresponds to the second set of candidate device features.

[0005] The application further provides a device determination apparatus, comprising: an acquisition module, configured to acquire a plurality of sets of device features of a target device, wherein the plurality of sets of device features comprise a first set of device features and a second set of device features, and the first set of device features and the second set of device features are of different categories; a determination module, configured to determine a plurality of sets of candidate device features that satisfy a preset similarity condition based on the plurality of sets of device features, wherein the plurality of sets of device features and the plurality of sets of candidate device features correspond to each other, the plurality of sets of candidate device features comprise a first set of candidate device features that satisfy the preset similarity condition with the first set of device features and a second set of candidate device features that satisfy the preset similarity condition with the second set of device features; and a combination module, configured to determine candidate devices based on the plurality of sets of candidate device features, and combine the candidate devices to generate device information of a target candidate device, wherein the candidate devices comprise a first candidate device and a second candidate device, the first candidate device corresponds to the first set of candidate device features, and the second candidate device corresponds to the second set of candidate device features.

[0006] In an example embodiment, the apparatus is configured to acquire the plurality of sets of device features of the target device by: acquiring device parameters of the target device; performing a feature extraction operation on the device parameters to determine a set of device features of the target device; and dividing the features in the set of device features according to spatial dimensions to obtain the plurality of sets of device features.

[0007] In an example embodiment, the apparatus is configured to divide the features in the set of device features according to spatial dimensions to obtain the plurality of sets of device features by: dividing features in the set of device features that belong to a front view of the target device into the first set of device features; dividing features in the set of device features that belong to a rear view of the target device into the second set of device features; and dividing features in the set of device features that belong to a top view of the target device into a third set of device features, wherein the third set of device features is different from the first set of device features and the second set of device features in category, the plurality of sets of candidate device features comprise a third set of candidate device features that satisfy the preset similarity condition with the third set of device features, and the candidate devices comprise a third candidate device, which corresponds to the third set of candidate device features.

[0008] In an example embodiment, the apparatus is configured to divide the features belonging to the front view of the target device in the set of device features into the first group of device features by: obtaining first device parameters belonging to the front view of the target device, wherein the first device parameters comprise at least one of: contour parameters corresponding to the front view, slot information corresponding to the front view, indicator light layout data corresponding to the front view; and integrating features in the set of device features corresponding to the first device parameters into the first group of device features.

[0009] In an example embodiment, the apparatus is configured to divide the features belonging to the back view of the target device in the set of device features into the second group of device features by: obtaining second device parameters belonging to the back view of the target device, wherein the second device parameters comprise at least one of: heat dissipation hole information corresponding to the back view, interface information corresponding to the back view, heat dissipation fan data corresponding to the back view, external expansion slot information corresponding to the back view; and integrating features in the set of device features corresponding to the second device parameters into the second group of device features.

[0010] In an example embodiment, the apparatus is configured to divide the features belonging to the top view of the target device in the set of device features into the third group of device features by: obtaining third device parameters belonging to the top view of the target device, wherein the third device parameters comprise at least one of: size parameters corresponding to the top view, load bearing data corresponding to the top view; and integrating features in the set of device features corresponding to the third device parameters into the third group of device features.

[0011] In an example embodiment, the apparatus is configured to determine a plurality of groups of candidate device features satisfying a preset similarity condition based on the plurality of groups of device features respectively by: obtaining sample candidate device parameters of a predetermined sample candidate device; performing feature extraction on the sample candidate device parameters to determine a set of sample candidate device features of the sample candidate device; dividing features in the set of sample candidate device features according to spatial dimensions to obtain sample candidate device features corresponding to the sample candidate device; determining similarities between a target group of device features and different target groups of sample candidate features, wherein the plurality of groups of device features comprise the target group of device features, the sample candidate device features comprise the target group of sample candidate features, the different target groups of sample candidate features correspond to different sample candidate devices respectively, and the target group of sample candidate features and the target group of device features belong to a same category; and determining the target group of sample candidate features satisfying the preset similarity condition as a group of candidate device features corresponding to the target group of device features in the plurality of groups of candidate device features.

[0012] In an example embodiment, the apparatus is configured to determine the similarity between each target device feature in the target group device features and each target candidate feature in the target group sample candidate features by determining a sub-similarity between each target device feature in the target group device features and each target candidate feature in the target group sample candidate features, wherein the target device feature and the target candidate feature belong to the same feature dimension, and combining the sub-similarities to determine the similarity between the target group device features and the target group sample candidate features.

[0013] In an example embodiment, the apparatus is configured to determine the sub-similarity between each target device feature in the target group device features and each target candidate feature in the target group sample candidate features by generating a target feature vector based on the target device feature and generating a candidate feature vector based on the target candidate feature, and determining a distance between the target feature vector and a plurality of the candidate feature vectors in a target vector space, and determining the sub-similarity based on the distance.

[0014] In an example embodiment, the apparatus is configured to perform a feature extraction operation on the sample candidate device parameters to determine a sample candidate device feature set of the sample candidate device by performing a feature extraction operation on the sample candidate device parameters to determine a sample candidate device feature set of the sample candidate device, calculating a mean value and a variance for sample candidate device features belonging to different sample candidate device feature sets and belonging to the same feature dimension, and normalizing each feature in each sample candidate device feature set based on the mean value and the variance to obtain an updated sample candidate device feature set.

[0015] In an example embodiment, the apparatus is further configured to add a preset weight coefficient to each feature corresponding to a feature dimension in the target group sample candidate features, dynamically adjust the preset weight coefficient to a target weight coefficient based on a preset environmental parameter threshold, and optimize each feature corresponding to a feature dimension in the target group sample candidate features according to the target weight coefficient.

[0016] In an example embodiment, the apparatus is configured to determine a candidate device based on the plurality of groups of candidate device features by determining a front view corresponding to the first candidate device, a rear view corresponding to the second candidate device, and a top view corresponding to the third candidate device, and combining the front view, the rear view, and the top view to obtain a combined view according to the size information of the target device, wherein the combined view is labeled with the first candidate device, the second candidate device, and the third candidate device, and the combined view is used to represent the device information of the target candidate device.

[0017] In an example embodiment, the device is further configured to: after the front view, the rear view and the top view are scaled and spliced according to the size information of the target device to obtain a combined view, generate front view contribution information, rear view contribution information and top view contribution information based on the similarity contribution degrees of different feature dimensions in the plurality of sets of candidate device features; and display the front view contribution information, the rear view contribution information and the top view contribution information on the combined view.

[0018] The present application also provides an electronic device, comprising: a memory configured to store a computer program; and a processor configured to execute the computer program to implement the steps of the determination method of any of the above devices.

[0019] The present application also provides a computer-readable storage medium, which stores a computer program, wherein the computer program is executed by a processor to implement the steps of the determination method of any of the above devices.

[0020] The present application also provides a computer program product, comprising a computer program, wherein the computer program is executed by a processor to implement the steps of the determination method of any of the above devices.

[0021] Through the embodiments of the present application, a plurality of sets of device features (such as a first set of device features and a second set of device features) are obtained by decomposing a target device, an independent vector space of each feature set is constructed, and a set of candidate device features (including a first set of candidate device features and a second set of candidate device features) highly matched with each feature set is selected based on a preset similarity condition (such as cosine similarity calculation), thereby solving the confusion problem caused by feature coupling in global matching. Finally, through combination optimization, candidate devices corresponding to different feature sets (such as a first candidate device and a second candidate device) are intelligently spliced to generate complete device information of a target candidate device, achieving the purpose of cross-device reuse of optimal local features, thereby significantly improving matching accuracy and effectively reducing artificial design costs, and solving the technical problems of low device matching efficiency and difficulty in ensuring accuracy in related technologies. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present application, the drawings required in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0023] Figure 1 is a schematic diagram of an application environment of an optional device determination method according to an embodiment of the present application;

[0024] Figure 2 is a flow diagram of a method for determining an optional device according to an embodiment of the application;

[0025] Figure 3 is a schematic diagram of a method for determining an optional device according to an embodiment of the application;

[0026] Figure 4 is a schematic diagram of a method for determining an optional device according to an embodiment of the application;

[0027] Figure 5 is a schematic diagram of a method for determining an optional device according to an embodiment of the application;

[0028] Figure 6 is a schematic diagram of a method for determining an optional device according to an embodiment of the application;

[0029] Figure 7 is a schematic diagram of a method for determining an optional device according to an embodiment of the application;

[0030] Figure 8 is a schematic diagram of a device for determining an optional device according to an embodiment of the application. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, but not all embodiments of the present application. Based on the embodiments in the present application, any other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0032] It should be noted that, in the description of the present application, the terms "comprise", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. The terms "first", "second" and the like in the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence.

[0033] In order for those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0034] In combination with the specific application environment architecture or specific hardware architecture on which the execution of the device determination method depends, the specific application environment architecture or specific hardware architecture is described here.

[0035] According to one aspect of the embodiments of this application, a method for determining a device is provided. Optionally, in this embodiment, the above-described method for determining a device can be applied to, for example... Figure 1 The hardware environment shown consists of server 101 and terminal device 103. For example... Figure 1 As shown, server 101 is connected to terminal 103 via a network and can be used to provide services to terminal devices or applications installed on terminal devices. The applications can be video applications, instant messaging applications, browser applications, educational applications, game applications, etc. Database 105 can be set up on the server or independently of the server to provide data storage services for server 101, such as a game data storage server. The network mentioned above can include, but is not limited to, wired networks and wireless networks. The wired network includes local area networks, metropolitan area networks, and wide area networks. The wireless network includes Bluetooth, WIFI, and other networks that enable wireless communication. Terminal device 103 can be a terminal configured with an application, and can include, but is not limited to, at least one of the following: mobile phones (such as Android phones, iOS phones, etc.), laptops, tablets, handheld computers, MID (Mobile Internet Devices), PADs, desktop computers, smart TVs, smart voice interaction devices, smart home appliances, vehicle terminals, aircraft, virtual reality (VR) terminals, augmented reality (AR) terminals, mixed reality (MR) terminals, and other computer devices. The server mentioned above can be a single server, a server cluster composed of multiple servers, or a cloud server.

[0036] Combination Figure 1 As shown, the method for determining the above-mentioned device can be executed by an electronic device, which can be a terminal device or a server. The method for determining the above-mentioned device can be implemented by the terminal device or the server respectively, or by the terminal device and the server together.

[0037] Embodiments of this application provide a method for determining a device, such as... Figure 2 As shown, the method for determining the equipment is described in detail, taking into account the execution flow of the method for determining the equipment.

[0038] S202, Obtain multiple sets of device features of the target device, wherein the multiple sets of device features include a first set of device features and a second set of device features, and the categories of the first set of device features and the second set of device features are different;

[0039] S204, respectively determine a plurality of groups of candidate device features satisfying the preset similarity condition based on the plurality of groups of device features, wherein the plurality of groups of device features correspond to the plurality of groups of candidate device features one by one, and the plurality of groups of candidate device features include a first group of candidate device features satisfying the preset similarity condition with the first group of device features and a second group of candidate device features satisfying the preset similarity condition with the second group of device features.

[0040] S206, determine a candidate device based on the plurality of groups of candidate device features, and combine the candidate device to generate device information of a target candidate device, wherein the candidate device includes a first candidate device and a second candidate device, the first candidate device corresponds to the first group of candidate device features, and the second candidate device corresponds to the second group of candidate device features.

[0041] Optionally, in the embodiments of the present application, the target device can include but is not limited to a cabinet to be intelligently matched to determine a suitable feature combination or a corresponding candidate device, and the core requirement is to find a historical device feature combination that can meet the functional, performance and environmental adaptation requirements of the target device through multi-dimensional feature comparison, so as to achieve the goal of efficiently reusing historical design resources and reducing design or deployment costs.

[0042] Optionally, in the embodiments of the present application, the plurality of groups of device features can include but are not limited to a set of information extracted from different dimensions or categories of the target device and used to describe device-specific attributes, and the sets of features are divided according to a preset classification rule. Each category of features focuses on the attributes of a certain specific aspect of the device, is independent of each other and complementary to each other, and together constitutes a comprehensive and accurate feature description of the target device, thereby providing a quantitative basis for subsequent similarity comparison with historical device features.

[0043] Optionally, in the embodiments of the present application, the first group of device features can include but is not limited to a certain category of device features in the plurality of groups of device features, which is different from other groups. This category of features focuses on the attributes of a certain specific dimension of the target device, has a clear description range and quantitative indicators, and is the core basis for subsequent similarity judgment with historical device features of the same category. The category division is usually based on the functional attributes or spatial perspective of the device, so as to ensure that the key characteristics of a certain aspect of the device can be accurately reflected.

[0044] Optionally, in the embodiments of the present application, the second group of device features can include but is not limited to another category of device features in the plurality of groups of device features, which is different from the first group of device features. This category of features focuses on the attributes of another dimension of the target device, and has obvious differences with the first group of device features in terms of description range and functional direction. The two are complementary to each other to fully present the key characteristics of the target device, and the purpose of setting is to capture device attributes from different angles to ensure that subsequent matching can cover multiple aspects of device requirements.

[0045] Optionally, in the embodiments of the present application, the preset similarity condition can include, but is not limited to, a standard or rule for judging whether the target device certain group of device features and the historical device features have similarity, which is formulated in combination with the quantification attribute of the device features and the actual application scene requirement, and covers the similarity threshold requirement at the feature vector level and can also include the hard matching range of the key attribute, so as to ensure that the selected candidate features can be matched with the target features in a specific dimension, and provide reliable basis for subsequent determination of the candidate device.

[0046] Optionally, in the embodiments of the present application, the multiple groups of candidate device features can include, but are not limited to, multiple groups of device features based on the target device, a feature set corresponding to each group of device features and obtained by screening from the historical device feature library through the preset similarity condition, each group of candidate device features is derived from the historical device, and satisfies the similarity requirement with the group of device features of the target device in the corresponding category, and the feature set is the key to realize the cross-device combination matching, and provides the support at the feature level for subsequent selection of the advantage part from different historical devices for combination.

[0047] Optionally, in the embodiments of the present application, the first group of candidate device features can include, but is not limited to, a feature set corresponding to each group of device features and satisfying the preset similarity condition in the multiple groups of candidate device features, each feature in the set is derived from the corresponding category feature of the historical device (such as the historical feature with the same category as the first group of device features), and has high similarity with the first group of device features of the target device in the core attribute, and the function is to provide direct basis for determining the candidate device that can be matched with the target device in the corresponding dimension of the first group of device features.

[0048] Optionally, in the embodiments of the present application, the second group of candidate device features can include, but is not limited to, a feature set corresponding to each group of device features and satisfying the preset similarity condition in the multiple groups of candidate device features, each feature in the set is derived from the corresponding category feature of the historical device (with the same category as the second group of device features), and has high similarity with the second group of device features of the target device in the core attribute, and the first group of candidate device features and the second group of candidate device features provide support for the matching of the target device from different dimensions respectively, so as to ensure that the subsequent combined candidate device can cover the multiple aspect requirements of the target device.

[0049] Optionally, in the embodiments of the present application, the candidate device can include, but is not limited to, the historical cabinet selected from the historical device library through the preset similarity condition, and each candidate device corresponds to a group of features in the multiple groups of candidate device features, which is the key carrier from the feature similarity judgment to the device level selection, and the value lies in providing the device features verified by practice, so as to facilitate subsequent combination to generate the scheme meeting the target requirement.

[0050] Optionally, in the embodiments of the present application, the first candidate device can include, but is not limited to, a historical cabinet corresponding to the first group of candidate device features in the candidate device, and the core basis for being selected is that a certain type of feature (the same as the first group of device feature categories) of the cabinet can meet the preset similarity condition of the target device in this group of features, and has an advantage in the key attribute or actual operation performance of this type of feature, and is the core device source in the first dimension when generating the target candidate device.

[0051] Optionally, in the embodiments of the present application, the second candidate device can include, but is not limited to, a historical cabinet corresponding to the second group of candidate device features in the candidate device, which is different from the first candidate device in the corresponding feature category, and the core advantage is reflected in the dimension corresponding to the second group of device features, which can meet the functional requirements of the target device in this dimension, and is the key device source in the second dimension when generating the target candidate device, and is complementary to the first candidate device to cover the multi-faceted requirements of the target device.

[0052] Optionally, in the embodiments of the present application, the device information of the target candidate device can include, but is not limited to, the information set formed by combining the candidate devices determined based on multiple groups of candidate device features, which comprehensively describes the key attributes, feature sources and matching basis of the target candidate cabinet, and the information not only covers the feature parameters of each dimension of the target candidate cabinet, but also labels the source cabinet, possible adjustment parameters (such as scaling ratio) and similarity contribution degree of each part of the feature, providing intuitive and detailed reference for user evaluation and decision-making.

[0053] Exemplarily, the above obtaining the multiple groups of device features of the target device can be understood as, in the cabinet intelligent matching scene, first, systematic feature extraction work is carried out on the target cabinet (i.e. the target device) that needs to be matched. This process is not to extract single-dimensional features, but to divide the features into multiple different categories according to the attribute characteristics and matching requirements of the cabinet, form multiple groups of device features, and explicitly contain at least two categories of obviously different features (the first group and the second group of device features) to ensure comprehensive coverage of the key attributes of the cabinet. In actual operation, the extraction rules will be formulated with reference to the spatial view and functional attributes of the cabinet. For example, the front view features of the cabinet are taken as the first group of device features. When extracting, the length and height parameters of the front view are obtained through image recognition technology, the contour coefficient is calculated by calculating the ratio of the perimeter to the area, the number of hard disk slots is counted and the relative position of each slot is recorded through coordinate positioning, the number and two-dimensional coordinates of light emitting diode lamps (Light Emitting Diode Lamp, LED for short) are identified, and a front view feature set containing these parameters is formed. At the same time, the rear view features of the cabinet are taken as the second group of device features. When extracting, the number of heat dissipation holes is counted and the proportion of the total area of the heat dissipation holes to the area of the rear view is calculated, the interface type is coded (such as Universal Serial Bus 3.0 (USB3.0) is coded as 001) and the two-dimensional coordinates of the interface are recorded (with the lower left corner of the rear view as the origin (0, 0)), the number and position of the cooling fans are identified, the number and coordinates of the expansion card slots are counted, and a rear view feature set is formed. In this way, multiple groups of device features with different categories and accurately reflecting different dimensional attributes of the cabinet are obtained, laying a foundation for subsequent matching.

[0054] Exemplarily, the above-mentioned determination of multiple sets of candidate device features satisfying the preset similarity condition based on multiple sets of device features can be understood as in the intelligent matching process of the cabinet, for the multiple sets of device features of the target cabinet obtained, each set of features needs to be compared and screened with the corresponding category of features in the historical cabinet feature library, through the preset similarity judgment standard, the historical features similar to the target features are selected to form a one-to-one corresponding candidate feature set, and the candidate feature set corresponding to the first set and the second set of device features is explicitly included, to ensure the pertinence and accuracy of the matching. In operation, the pre-constructed historical cabinet feature database will be called first, which stores the feature vectors and key parameters (such as feature mean, standard deviation) of historical cabinets classified by categories such as front view and rear view. When processing the first set of device features (such as front view features), the target front view feature vector and the historical front view feature vector will be standardized (converted to standard normal distribution with mean 0 and standard deviation 1) first, and then the cosine similarity formula is used to calculate the similarity degree of the two, if the similarity value is higher than the preset threshold (such as 0.85), and the key attributes such as the number of hard disk slots and the number of LED lights of the historical cabinet front view deviate from the target cabinet within the allowable range, then the historical feature is included in the first set of candidate device features; When processing the second set of device features (such as rear view features), the weight of the heat dissipation hole and the heat dissipation fan features will be dynamically adjusted (the adjusted weight is calculated according to the formula) combined with the current computer room temperature (such as 36℃), then the standardization and cosine similarity calculation are performed, and the historical features with similarity up to standard and adaptive attributes such as interface type and heat dissipation hole proportion are selected to form the second set of candidate device features, and finally the one-to-one correspondence between multiple sets of device features and multiple sets of candidate device features is realized.

[0055] Exemplarily, the above-mentioned device information of the target candidate device generated by combining the candidate devices determined based on the multiple sets of candidate device features can be understood as follows in the intelligent matching scenario of the cabinet: first, the historical cabinets to which each set of candidate device features belongs (i.e., the candidate devices) are determined according to the correspondence between each set of candidate device features and the historical cabinets; then, the optimal devices in each dimension are selected from the candidate devices, the advantage features of these devices are integrated, and the device information that can comprehensively reflect the attributes, sources and matching basis of the target candidate cabinet is formed, and the first candidate device and the second candidate device corresponding to the first set of candidate device features and the second set of candidate device features are explicitly included to ensure the completeness and pertinence of the combination scheme. In actual operation, the candidate devices are determined in combination with the feature similarity and historical operation data. For example, from the historical cabinets A, B and C corresponding to the first set of candidate device features, the cabinet A with the highest similarity (such as the similarity of 0.92 of cabinet A) and no fault in the past year is selected as the first candidate device; from the historical cabinets D and E corresponding to the second set of candidate device features, the cabinet D with better heat dissipation performance is selected as the second candidate device in combination with the temperature of the computer room; if there is a top view candidate feature, a corresponding third candidate device (such as cabinet F) is also determined. When the device information is generated by combination, the front view feature of cabinet A, the rear view feature of cabinet D and the top view feature of cabinet F are scaled at a ratio of 1:1 according to the size of the target cabinet (such as the rear view of cabinet D is scaled by 1.05 times), to ensure that the views can be perfectly spliced, and the source cabinet (such as "front view source: cabinet A") and the scaling ratio of each feature are labeled, the similarity contribution analysis (such as the contribution of the front view of cabinet A is 40%) is generated, and these contents are integrated to form the device information of the target candidate device, providing clear and comprehensive reference for the user.

[0056] By the embodiments of the present application, the target device is decomposed into multiple sets of device features of independent categories (such as the first set of device features and the second set of device features), the independent vector spaces of each feature set are constructed, the separation of heterogeneous device features is realized, the candidate device feature sets (including the first set of candidate device features and the second set of candidate device features) highly matched with each set of features are respectively selected based on the preset similarity condition (such as the cosine similarity calculation) through the grouping and parallel matching strategy, the confusion problem caused by feature coupling in global matching is solved, and finally the candidate devices corresponding to different feature sets (such as the first candidate device and the second candidate device) are intelligently spliced through the combination optimization algorithm to generate the complete device information of the target candidate device, achieving the purpose of reusing the optimal local features across devices, thereby significantly improving the matching accuracy and effectively reducing the artificial design cost, and solving the technical problems of low device matching efficiency and difficulty in ensuring accuracy in the related art.

[0057] In an example embodiment, the plurality of groups of device features of the target device are acquired, including: acquiring device parameters of the target device; performing a feature extraction operation on the device parameters to determine a set of device features of the target device; and dividing the features in the set of device features according to a spatial dimension to obtain the plurality of groups of device features.

[0058] Optionally, in the embodiment of the present application, the device parameters can include, but are not limited to, raw data for describing physical structure, functional attributes and performance indicators of the target cabinet, which are the basis for subsequent extraction of device features, cover quantifiable information of the cabinet under different spatial views, include size data reflecting the cabinet form, component configuration data embodying the cabinet function, and performance data affecting the cabinet use adaptability, and need to be comprehensively and accurately collected to ensure the accuracy of subsequent feature extraction.

[0059] Optionally, in the embodiment of the present application, the feature extraction operation can include, but is not limited to, a process of converting the raw parameters into feature information capable of quantitatively describing specific attributes of the cabinet by using preset calculation, statistics, encoding and other processing rules for the collected device parameters of the target cabinet, which needs to adopt an adaptive processing method in combination with the attribute characteristics of the cabinet parameters in each dimension to ensure that the extracted features can accurately reflect the structure, function and performance characteristics of the cabinet, and lay a foundation for subsequent division of features according to the spatial dimension.

[0060] Optionally, in the embodiment of the present application, the set of device features can include, but is not limited to, the sum of feature information comprehensively reflecting the multi-dimensional attributes of the cabinet after performing the feature extraction operation on the device parameters of the target cabinet, which covers the key features of the cabinet in terms of structural form, functional configuration, performance bearing and the like, and each feature is presented in a quantifiable or encodable form, which is a direct basis for subsequent division of the plurality of groups of device features according to the spatial dimension, and ensures that the divided features can focus on the attributes of the cabinet under different spatial views.

[0061] Optionally, in the embodiment of the present application, the spatial dimension can include, but is not limited to, a spatial view division standard based on the physical structure of the cabinet, which takes the common spatial view of the cabinet in the actual observation and use process as the basis, classifies the overall features of the cabinet according to different observation views, and each spatial dimension corresponds to a specific view direction of the cabinet, so that the divided features can focus on reflecting the attributes of the cabinet under the view, and provide a classification basis for subsequent independent matching of features for different views.

[0062] Exemplarily, the above-mentioned obtaining the device parameters of the target device can be understood as in the cabinet intelligent matching scene, to realize the multi-group device feature acquisition of the target cabinet (target device), firstly, the original data capable of describing the physical structure, function configuration and performance index of the cabinet need to be systematically collected, and these data need to cover the key attributes under different spatial views of the cabinet, to provide comprehensive and accurate basic information for subsequent feature extraction. In actual operation, the data can be collected by combining professional measuring tools (such as a tape measure, a coordinate measuring instrument) and device parameter queries (such as a cabinet product manual, a factory detection report), for example, for the front view of the target cabinet, the length and height dimensions are measured, the number of hard disk slots is counted and the spacing of each slot is recorded, and the installation position of the LED lamp is determined; for the rear view, the number of heat dissipation holes is counted, the aperture of a single heat dissipation hole is measured to calculate the total area, the type (such as USB, Ethernet) and installation position of the interface are recorded, the number and speed parameters of the cooling fan are queried, and the number of expansion card slots is counted; for the top view, the length and width dimensions are measured, the maximum load capacity value of the cabinet is queried, and through these ways, the device parameters of the target cabinet are completely obtained, and the subsequent feature extraction operation is prepared.

[0063] Exemplarily, the above-mentioned obtaining the device parameters of the target device; performing feature extraction operation on the device parameters to determine the device feature set of the target device can be understood as in the cabinet intelligent matching process, after completing the device parameter collection of the target cabinet, the original parameters need to be processed and transformed according to the preset processing rules, to be transformed into feature information capable of quantitatively describing the attributes of the cabinet in each dimension, and then integrated to form a feature set reflecting the attributes of the cabinet. Different extraction methods will be adopted according to the parameter type during operation, for example, for the length and height parameters of the front view, the contour coefficient is obtained by calculating the ratio of the perimeter to the area (perimeter = (length + height) x 2, area = length x height, coefficient = perimeter / area); for the hard disk slot parameter, the total number of slots is counted and the relative coordinates of each slot are calculated combined with the measured spacing; for the LED lamp position parameter, it is directly arranged as a two-dimensional coordinate array; for the heat dissipation hole parameter of the rear view, the total area of all heat dissipation holes is calculated and the ratio of the heat dissipation hole area to the rear view area (length x height) is obtained to obtain the heat dissipation hole ratio; for the interface type parameter, the interface coordinates are recorded according to the preset coding rule (such as USB3.0 coded as 001); for the length and width parameters of the top view, they are directly extracted, and the load capacity parameter is directly recorded. The extracted features (such as contour coefficient, heat dissipation hole ratio, length x width dimension, etc.) are integrated, and the device feature set of the target cabinet is formed.

[0064] Exemplarily, the above-mentioned performing a feature extraction operation on the device parameters, and determining a device feature set of the target device can be understood as follows: in the intelligent matching process of the cabinet, after obtaining the device feature set of the target cabinet, the features in the feature set need to be classified according to the spatial view (front view, rear view, top view) of the cabinet as the classification standard, so that each category of features focuses on reflecting the attributes of the cabinet under a certain spatial view, and finally forms multiple groups of independent device features. When performing the classification, the spatial view to which each feature belongs is determined, for example, the features (outline coefficient, number and coordinates of hard disk slot, number and coordinates of LED lamp) related to the front structure and function of the cabinet are classified into the device feature group corresponding to the front view; the features (number and proportion of heat dissipation holes, type code and coordinates of interface, number and coordinates of heat dissipation fans, number and coordinates of expansion card slots) related to the rear function of the cabinet are classified into the device feature group corresponding to the rear view; the features (length, width and bearing capacity) related to the planar size and bearing capacity of the cabinet are classified into the device feature group corresponding to the top view. Through such classification, the original complete device feature set is split into three groups of independent device features, each group of features can accurately reflect the core attributes of the cabinet under the corresponding spatial view, and provides a clear classification basis for subsequent independent matching of different view features. Figure Three

[0065] Through the embodiments of the present application, the spatial dimension division strategy is adopted to perform feature extraction on the device parameters of the target device, the device feature set is defined, and multiple groups of device features are divided according to the spatial view (such as front / rear view), thereby realizing logical decoupling of the physical structure features; through the construction of a standardized vector space, the original parameters are converted into quantifiable and comparable feature data, thereby achieving the purpose of feature data standardization processing, thereby providing clear structure and consistent dimension basis data support for subsequent grouping matching, and solving the problem of low matching efficiency caused by feature mixing in the traditional scheme.

[0066] In one exemplary embodiment, the features in the device feature set are divided according to the spatial dimension, and multiple groups of device features are obtained, including:

[0067] The features in the device feature set belonging to the front view of the target device are divided into a first group of device features;

[0068] The features in the device feature set belonging to the rear view of the target device are divided into a second group of device features;

[0069] ​The features in the top view belonging to the target device in the device feature set are divided into a third group of device features, wherein the third group of device features is different from the categories of the first group of device features and the second group of device features, the multiple groups of candidate device features include a third group of candidate device features that satisfy a preset similarity condition with the third group of device features, and the candidate device includes a third candidate device corresponding to the third group of candidate device features.

[0070] Optionally, in the embodiments of the present application, the third group of device features can include but is not limited to a feature set that is different from the feature categories of the front view (first group) and the rear view (second group) and is divided from the device feature set of the target cabinet (target device). This group of features focuses on the core attributes in the top view of the cabinet, mainly reflects the characteristics of the cabinet in the aspects of space occupation and physical carrying capacity in the plane, which are directly related to the space adaptability of the cabinet in the data center room and the support capacity for the weight of the equipment. It is a key dimension that must be considered before the cabinet is deployed and the equipment is installed. The category division is based on the natural attributes of the cabinet space view, which ensures that the front and rear view features are clearly distinguished in the description direction and application scenario.

[0071] Optionally, in the embodiments of the present application, the third group of candidate device features can include but is not limited to a feature set that satisfies a preset similarity condition with the third group of device features of the target cabinet and is filtered from the historical cabinet feature library. Each feature in this set is derived from the top view features of the historical cabinet and has high similarity with the third group of device features of the target cabinet in the core attributes such as space size and carrying capacity. The filtering process needs to combine the preset similarity judgment standard (such as feature vector similarity threshold, key attribute deviation range), to ensure that the filtered features can match the needs of the target cabinet in the dimensions of space adaptation and carrying capacity, and is the direct basis for determining the third candidate device. At the same time, together with the first and second groups of candidate device features, it forms multiple groups of candidate device features, providing complete feature support for cross-cabinet combination matching.

[0072] Optionally, in the embodiments of the present application, the third candidate device can include but is not limited to a historical cabinet corresponding to the third group of candidate device features in the candidate device. The core basis for being selected is that the top view features (of the same category as the third group of device features) of the cabinet can satisfy the preset similarity condition with the third group of device features of the target cabinet, and have advantages in space size adaptability and carrying capacity reliability. It is a key equipment source in the dimensions of space occupation and carrying capacity when generating the target candidate device. It is complementary to the first candidate device (corresponding to the front view features) and the second candidate device (corresponding to the rear view features) in the functional dimension, and together ensures that the target candidate device can cover all aspects of requirements such as cabinet structure form, functional configuration, space adaptation, and carrying performance.

[0073] Exemplarily, the above-mentioned division of the features belonging to the front view of the target device in the device feature set into the first group of device features can be understood as follows: in the intelligent matching scenario of the cabinet, when dividing the device feature set of the target cabinet (target device) according to the spatial dimension, first, the feature range corresponding to the front view perspective is determined, all features describing the structure, shape and front function configuration of the cabinet are filtered out from the feature set, these features are classified and integrated to form an independent first group of device features. The division process needs to be based on the relevance of the features and the front view perspective to ensure that the features classified can accurately reflect the core attributes of the front of the cabinet. For example, the target cabinet device feature set contains features such as contour coefficient (calculated by the length and height of the front view, reflecting the compactness of the front), number and coordinates of hard disk slots (located on the front of the cabinet, related to the layout of the device installation), number and position of LED lights (distributed on the front, used for state monitoring), etc. These features are all directly related to the front view perspective of the cabinet. Through manual annotation or automatic classification algorithm (such as keyword matching based on feature description), these features are separated from the set and uniformly classified into the first group of device features, so that this group of features focuses on describing the key attributes of the front of the cabinet, laying the foundation for subsequent matching with the historical cabinet front view features.

[0074] Exemplarily, the above-mentioned division of the features belonging to the front view of the target device in the device feature set into the first group of device features; the features belonging to the rear view of the target device in the device feature set are divided into the second group of device features can be understood as follows: after completing the division of the front view features, continue to filter the features related to the rear view perspective of the cabinet from the device feature set according to the spatial dimension. These features mainly describe the functional configuration and structural details of the back of the cabinet, and there is a significant difference in the description of the perspective and attribute type from the front view features. Through classification and integration, an independent second group of device features is formed to ensure that this group of features can focus on reflecting the core functions (such as heat dissipation, device connection) of the back of the cabinet. For example, the target cabinet device feature set also contains features such as the number and proportion of heat dissipation holes (located on the back of the cabinet, affecting the heat dissipation efficiency), interface type and coordinates (back interface for connecting devices with the outside), number and position of cooling fans (back fans for auxiliary cooling), number and coordinates of external card slots (back card slots for expanding devices), etc. These features are all directly related to the rear view perspective of the cabinet. By judging the physical location corresponding to the features (such as the keywords “heat dissipation hole”, “back interface”, etc.), they are filtered out from the feature set and classified into the second group of device features, so that this group of features focuses on describing the functional attributes of the back of the cabinet, and forms a clear perspective and category distinction with the first group of front view features.

[0075] Exemplarily, the above-mentioned division of the features belonging to the rear view of the target device in the device feature set into a second group of device features, and the division of the features belonging to the top view of the target device in the device feature set into a third group of device features can be understood as follows: after the division of the features of the front view and the rear view is completed, the features related to the top view of the cabinet are screened from the remaining device feature set, which focus on the planar spatial size and physical carrying capacity of the cabinet, neither involving the front structure of the front view nor the back function of the rear view, and the categories are different from the first two groups of features, and are integrated to form the third group of device features; meanwhile, in the subsequent matching process, the historical cabinet top view features that meet the preset similarity condition need to be screened for the third group of device features to form the third group of candidate device features, and the corresponding third candidate device (i.e., the historical cabinet with the candidate features of this group) is determined. For example, the remaining features in the target cabinet device feature set include the top view length and width dimensions (reflecting the cabinet floor area), and the carrying capacity value (reflecting the device support capacity), which are completely different from the categories of the contour coefficient of the front view and the heat dissipation hole proportion of the rear view, and are classified into the third group of device features; in the subsequent matching, the similarity of these features and the historical cabinet top view features is calculated, and the features with small size deviation and suitable carrying capacity are screened to form the third group of candidate device features, and then the optimal cabinet F is selected from the historical cabinets corresponding to these candidate features (such as cabinets F and G) as the third candidate device, to ensure that the target candidate device meets the requirements in terms of spatial adaptation and carrying performance.

[0076] It should be noted that when the equipment feature set is divided according to the spatial dimension, in addition to the basic mode of dividing the front view features into the first group, the rear view into the second group, and the top view into the third group, the feature division under each view can also be refined and expanded according to the special type of the data center cabinet or the actual application requirements, which is not limited in the present application. For example, for some special cabinets with side view function requirements (such as cabinets with maintenance doors or expansion interfaces on the side), a fourth group of equipment features can be added to the original three groups of features, which can include the size of the maintenance door, the type and position of the side interface, and other features. The corresponding multiple groups of candidate equipment features also add a fourth group of candidate equipment features (screening similar features in the historical cabinet side view features), and the candidate equipment adds a fourth candidate equipment (corresponding to the fourth group of candidate equipment features) to meet the matching requirements of the special cabinet for the side function; for example, for the top view features, in addition to the basic length, width and bearing capacity, if the target cabinet is a multi-layer structure, the interlayer spacing and the bearing distribution of each layer can be added to the third group of equipment features. At this time, when screening the third group of candidate equipment features, not only the length, width and total bearing capacity need to be compared, but also the interlayer spacing and the bearing distribution of each layer need to be checked for similarity with the target requirements. The adaptability of the corresponding third candidate equipment is also considered to ensure that the division mode can flexibly adapt to different types of cabinet matching scenarios; in addition, for some scenes with high precision requirements for view features, the front view features can be further divided into front view left area features and front view right area features, which are respectively classified into subcategories of the first group of equipment features. When matching, candidate features can be independently screened for different area features to further improve the accuracy of matching. These diversified division cases can be adjusted according to actual needs to better serve the core goal of cabinet intelligent matching.

[0077] Through the embodiments of the present application, the three-view spatial division mechanism (front view, rear view, top view) is adopted, and the equipment feature set is classified into three independent view groups (first / second / third group of equipment features), which realizes the accurate classification of multi-dimensional physical features. By defining independent feature categories (such as area and bearing capacity) for the newly added view groups such as top view, the feature coverage dimension is expanded. In the matching stage, by screening the third group of candidate equipment features and the corresponding third candidate equipment, the purpose of multi-view collaborative optimization is achieved, thereby realizing the comprehensive coverage of the complex requirements of cabinet structure, heat dissipation, and bearing capacity, and significantly improving the matching integrity in complex scenarios.

[0078] In one example embodiment, the features of the front view of the equipment feature set belonging to the target equipment are divided into the first group of equipment features, including:

[0079] obtaining a first device parameter of the front view of the target device, wherein the first device parameter comprises at least one of the following: a contour parameter corresponding to the front view, slot information corresponding to the front view, and indicator light layout data corresponding to the front view;

[0080] integrating a feature corresponding to the first device parameter in the device feature set as a first group of device features.

[0081] Optionally, in the embodiments of the present application, the first device parameter can include but is not limited to a set of raw data for describing the core attributes of the front view of the target cabinet (target device), which directly relates to the structural form, functional slot configuration and indicator light layout of the cabinet under the front view perspective, and is the basis for extracting the front view features and forming the first group of device features. Among them, the contour parameter focuses on the size features of the front view of the cabinet, determines the structural compactness and spatial form of the front of the cabinet; the slot information focuses on the slot configuration for installing devices on the front view, which is related to the accommodation capacity and installation layout of the cabinet for hardware devices; the indicator light layout data reflects the distribution of the indicator lights on the front view, which affects the visual monitoring effect of the running state of the cabinet. The collection of these parameters needs to ensure accuracy to ensure the reliability of subsequent feature extraction and integration, which directly serves the similarity matching of the front view features of the cabinet and the historical cabinet features.

[0082] Optionally, in the embodiments of the present application, the contour parameter corresponding to the front view can include but is not limited to the quantitative data collected from the front view perspective of the target cabinet for describing the size and structural form of the front of the cabinet, which is the core input for calculating the contour coefficient of the front view (reflecting the compactness and shape features of the front of the cabinet), and directly determines the description accuracy of the overall structural form of the front view of the cabinet. Its collection needs to be based on standard measurement methods to ensure the accuracy and consistency of the data, so as to be compared with the historical cabinet front view contour features for effective comparison and judgment of the similarity of the structural form. The core role of the contour parameter is to convert the shape features of the front view of the cabinet into calculable and comparable indicators through quantitative size data, to provide a basic dimension for feature extraction, and also an important basis for evaluating whether the cabinet can adapt to the installation space of the peripheral equipment on the front side, whether it can meet the aesthetic appearance and space utilization rate of the front layout of the machine room.

[0083] Optionally, in the embodiments of the present application, the slot information corresponding to the front view can include, but is not limited to, relevant data collected from the target cabinet front view perspective, which is used to describe the slots on the front of the cabinet for installing various types of hardware devices. These data record in detail the number, arrangement, position distribution and size specification of the slots, which are directly related to the accommodation capacity and installation compatibility of the cabinet for hardware devices, are an important dimension for evaluating the functional adaptability of the cabinet, and are also the core basis for subsequent extraction of front view slot features and similarity matching with historical cabinet slot features. The completeness and accuracy of the slot information determine whether the subsequent matching can find a historical cabinet feature that can meet the hardware installation requirements of the target cabinet, avoiding the problem of hardware devices that cannot be normally installed or unreasonable layout due to slot mismatch.

[0084] Optionally, in the embodiments of the present application, the indicator light layout data corresponding to the front view can include, but is not limited to, relevant data collected from the target cabinet front view perspective, which is used to describe the number, type, installation position and function correspondence of the indicator lights on the front of the cabinet. These data are directly related to the visual monitoring function of the cabinet running state. The indicator lights feed back in real time the key information such as the power state, data transmission state and equipment fault state of the cabinet through different colors, flashing frequencies or on-off states. Therefore, the completeness of the indicator light layout data determines the accuracy of the feature description of the cabinet monitoring function, is an important basis for subsequent similarity matching of the front view indicator light features of the historical cabinet, and ensures the compatibility of the candidate features after matching to meet the monitoring requirements of the cabinet.

[0085] Through the embodiments of the present application, the front view feature directional extraction mechanism is adopted, the contour parameters, slot information and LED layout and other key parameters are obtained, the related data in the device feature set are integrated into the first group of device features, the special quantitative description of the device appearance and function layout is realized, the structured vector generation (such as coordinate array) is achieved, the purpose of feature computability is achieved, and high-precision data basis is provided for the front view similarity matching.

[0086] In one example embodiment, the features of the rear view belonging to the target device in the device feature set are divided into a second group of device features, including:

[0087] Obtaining the second device parameters of the rear view belonging to the target device, wherein the second device parameters include at least one of the following: rear view corresponding vent information, rear view corresponding interface information, rear view corresponding cooling fan data, and rear view corresponding expansion slot information;

[0088] Integrating the features corresponding to the second device parameters in the device feature set into the second group of device features.

[0089] Optionally, in the embodiments of the present application, the second device parameter can include but is not limited to a set of original data collected from the rear view of the target cabinet (target device) for describing the core function attributes of the back of the cabinet, which focuses on the key information related to the heat dissipation, device connection and expansion of the back of the cabinet, and is the basis for extracting the rear view features and integrating them into the second group of device features. Among them, the heat dissipation hole information and the heat dissipation fan data are directly related to the heat dissipation performance of the cabinet, which determines whether the cabinet can effectively dissipate the heat generated by the device running; the interface information is related to the connection compatibility of the cabinet and the external device or network; and the external card slot information affects the hardware expansion capability of the cabinet. The collection of these parameters needs to ensure comprehensiveness and accuracy to ensure that the subsequent feature extraction can accurately reflect the functional characteristics of the rear view of the cabinet, provide a reliable basis for similarity matching with the historical rear view features of the cabinet, and then meet the actual needs of the data center for cabinet heat dissipation, connection and expansion.

[0090] Optionally, in the embodiments of the present application, the heat dissipation hole information corresponding to the rear view can include but is not limited to quantitative data collected from the rear view of the target cabinet for describing the key attributes of the heat dissipation hole, which is the core basis for evaluating the heat dissipation capacity of the rear view of the cabinet and directly affects the adaptability of the cabinet to the internal device heat dissipation demand. As an important structure for cabinet heat dissipation, the number, size and distribution of the heat dissipation holes determine the circulation efficiency of air in the cabinet, and then affect the stability of device running. The collection of this information needs to be based on standardized measurement methods to ensure that the data can be accurately converted into calculable heat dissipation features (such as heat dissipation hole ratio) so as to be compared with the historical heat dissipation hole features of the rear view of the cabinet in the subsequent process, to screen out candidate features with similar heat dissipation performance, and to meet the needs of the data center to adjust the heat dissipation matching weight according to the room temperature.

[0091] Optionally, in the embodiments of the present application, the interface information corresponding to the rear view can include but is not limited to related data collected from the rear view of the target cabinet for describing the interface type, number, position and function, which directly determines the connection compatibility and convenience of the cabinet and external devices (such as servers, switches, power modules) and network, and is the core embodiment of the functional adaptability of the cabinet. Different types of interfaces correspond to different connection demands, and the position distribution of the interface affects the rationality of wiring and the convenience of maintenance. The collection of this information needs to record the key attributes of the interface in detail to ensure that it can be converted into a feature that can be coded and located, so as to be compared with the historical interface features of the rear view of the cabinet in the subsequent process, to screen out candidate features that can meet the connection demand of the target cabinet, and to avoid the situation that the device cannot be normally connected or the maintenance is difficult due to the mismatch of the interface.

[0092] Optionally, in the embodiments of the present application, the rear view corresponding to the heat dissipation fan data can include but is not limited to relevant data collected from the target cabinet rear view perspective, which is used to describe the heat dissipation fan configuration and installation condition. These data are important supplements for assisting in evaluating the cabinet heat dissipation performance, and together with the heat dissipation hole information, determine the overall heat dissipation efficiency of the cabinet. The heat dissipation fan accelerates the discharge of hot air inside the cabinet through active blowing or suction, and its running state is crucial for equipment heat dissipation, especially when the room temperature exceeds the threshold. The collection of this data needs to cover the core configuration information of the fan to ensure that it can be converted into a feature reflecting the function of the fan, so that in subsequent dynamic weight adjustment, the matching weight of the fan feature can be flexibly adjusted according to the room temperature, and the historical cabinet rear view features with similar heat dissipation fan configurations are screened out.

[0093] Optionally, in the embodiments of the present application, the rear view corresponding to the external plug-in card slot information can include but is not limited to relevant data collected from the target cabinet rear view perspective, which is used to describe the external plug-in card slot configuration condition. These data are directly related to the hardware expansion capability of the cabinet, and determine whether the cabinet can add or upgrade functional modules (such as expansion graphics card, sound card, data acquisition card, etc.) according to business needs. The number, size, position distribution, etc. of the external plug-in card slot affect the installation compatibility and maintenance convenience of the expansion module. The collection of this information needs to record the key attributes of the slot in detail to ensure that it can be converted into a feature that can be counted and located, so that in subsequent comparison with the historical cabinet rear view external plug-in card slot features, candidate features with similar expansion capabilities are screened out to meet the needs of flexible expansion of the data center cabinet according to business growth.

[0094] Through the embodiments of the present application, the rear view composite feature integration mechanism is adopted, the parameters such as heat dissipation hole ratio, interface type and position, fan and external plug-in card slot are extracted, the second group of device features are formed, and the independent modeling of heat dissipation and interface function is realized; through multi-dimensional feature vector fusion (such as type coding + coordinate positioning), the purpose of systematic expression of complex functional features is achieved, thereby solving the problem of difficult unified matching of heterogeneous features in the rear view.

[0095] In one example embodiment, the features of the top view belonging to the target device in the device feature set are divided into a third group of device features, including:

[0096] Obtaining third device parameters belonging to the top view of the target device, wherein the third device parameters include at least one of the following: size parameters corresponding to the top view, and bearing data corresponding to the top view;

[0097] Integrating the features corresponding to the third device parameters in the device feature set into the third group of device features.

[0098] Optionally, in the embodiments of the present application, the third device parameter can include, but is not limited to, a set of raw data collected from the perspective of the top view of the target cabinet (target device) for describing the spatial attributes and physical carrying capacity of the cabinet plane, which is the core basis for extracting the top view features and integrating into the third set of device features, and is directly related to the spatial adaptability of the cabinet in the data center room and the safety of supporting the weight of the device. Among them, the size parameter corresponding to the top view determines the floor area of the cabinet in the room, which needs to be matched with the reserved installation space in the room; the bearing data corresponding to the top view determines the upper limit of the weight of the devices that the cabinet can accommodate, which needs to be adapted to the total weight of the devices to be installed. The collection of these parameters needs to follow the standardized measurement specification to ensure the accuracy of the data, so as to be converted into quantifiable features for subsequent similarity matching with historical cabinet top view features, and to support the subsequent pre-screening of candidate cabinets (based on area and bearing requirements to clean up invalid candidate data).

[0099] Optionally, in the embodiments of the present application, the size parameter corresponding to the top view can include, but is not limited to, quantified data collected from the perspective of the top view of the target cabinet for describing the geometric shape of the cabinet plane, which directly determines the spatial occupancy range of the cabinet in the data center room and is a key indicator for evaluating whether the cabinet can adapt to the reserved installation space in the room. As a perspective reflecting the top plane shape of the cabinet, the size parameter of the top view mainly focuses on the length and width in the horizontal direction, and the product of the two parameters is the floor area of the cabinet, which needs to be matched with the size of the installation area planned in the room to avoid the cabinet being unable to be deployed or wasting space due to size mismatch. The collection of this parameter needs to use high-precision measurement tools (such as laser range finder) to ensure that the data error is controlled within the allowable range, so as to be converted into the length and width features of the top view for subsequent similarity comparison with the size features of the historical cabinet top view, and to provide accurate basis for the comparison, and in the pre-screening stage of the candidate cabinet, historical cabinets with size mismatch can be quickly excluded based on this parameter, improving the matching efficiency.

[0100] Optionally, in the embodiments of the present application, the bearing data corresponding to the top view can include, but is not limited to, quantitative data associated with the target cabinet top view perspective for describing the overall support capacity of the cabinet, which directly determines the upper limit of the total weight of the equipment that the cabinet can safely bear, and is a key indicator to ensure the structural stability and safe operation of the equipment after installation. The bearing capacity of the cabinet is mainly determined by factors such as the frame material (such as the thickness of the steel material), the structural design (such as the beam spacing), etc., and is obtained through professional bearing detection experiments, usually expressed in weight units (such as kilograms). In the data center scenario, the total weight of the equipment to be installed (such as servers, storage arrays) needs to be strictly controlled within the bearing capacity range of the cabinet to avoid cabinet deformation, tilting, and even equipment damage or safety accidents due to insufficient bearing. The collection of this data needs to be based on authoritative detection reports or factory parameters to ensure data authenticity and reliability, so as to be subsequently converted into the bearing capacity characteristics of the top view, providing a basis for similarity comparison with historical cabinet top view bearing characteristics, and at the same time, in the pre-screening stage of the candidate cabinet, the historical cabinets with insufficient bearing capacity can be quickly excluded based on this data to ensure that the matching results meet the safe operation requirements of the equipment.

[0101] Through the embodiments of the present application, by adopting the top view physical parameter aggregation mechanism, the size and bearing data are extracted and integrated into the third group of equipment characteristics, realizing the special analysis of equipment space occupation and mechanical properties; by simplifying the feature dimension (length / width / bearing value), the purpose of efficient matching of core parameters is achieved, thereby providing key decision basis for the spatial adaptability of cabinet deployment.

[0102] In one example embodiment, based on the plurality of groups of equipment characteristics, a plurality of groups of candidate equipment characteristics satisfying a preset similarity condition are determined, comprising:

[0103] Obtaining sample candidate equipment parameters of a predetermined sample candidate equipment;

[0104] Performing a feature extraction operation on the sample candidate equipment parameters to determine a sample candidate equipment characteristic set of the sample candidate equipment;

[0105] Dividing the features in the sample candidate equipment characteristic set according to the spatial dimension to obtain sample candidate equipment characteristics corresponding to the sample candidate equipment;

[0106] Determining the similarity between the target group of equipment characteristics and different target groups of sample candidate characteristics, wherein the plurality of groups of equipment characteristics include the target group of equipment characteristics, the sample candidate equipment characteristics include the target group of sample candidate characteristics, different target groups of sample candidate characteristics correspond to different sample candidate equipment, and the target group of sample candidate characteristics and the target group of equipment characteristics belong to the same category;

[0107] The target group sample candidate feature satisfying the preset similarity condition is determined as a group of candidate device features corresponding to the target group device feature in the multiple groups of candidate device features.

[0108] Optionally, in the embodiment of the present application, the sample candidate device described above can include, but is not limited to, historical cabinets in a data center device management scene, which are pre-stored in a historical cabinet feature library, used for feature comparison with a target cabinet (target device). These cabinets are mature devices that have been actually deployed or designed and verified, and their parameters and features have been collected and stored through a standardized process, which can provide a reference basis for feature matching of the target cabinet. The core value of the sample candidate device lies in providing reusable historical feature resources, and through similarity comparison with the target cabinet features, candidate features that meet the target requirements are screened out, and then an optimal matching scheme is generated. It includes, but is not limited to, cabinet A with complete front view contour parameters, rear view cooling hole configuration and top view load data, which has been used for high-density computing device deployment, cabinet B with a multi-fan cooling system and rich external card slot, and cabinet C suitable for small computer rooms, compact size and moderate load, etc. The parameters and features of these sample candidate devices have been stored according to the specification and can be used as a basis for comparison with the target cabinet features.

[0109] Optionally, in the embodiment of the present application, the sample candidate device parameters described above can include, but are not limited to, a set of original data collected from each perspective of the sample candidate device (historical cabinet) for describing its physical structure, functional configuration and performance indicators. These data are consistent with the collection dimension of the target device parameters, cover the core original information corresponding to the front view, rear view and top view of the sample candidate device, and are the basis for extracting the sample candidate device features and forming the sample candidate device feature set. The collection of sample candidate device parameters needs to follow the same standardized process as the target device parameters to ensure consistency in data format and accuracy, so as to facilitate effective comparison with the target device features in the future and avoid similarity judgment deviation caused by differences in data collection standards. These parameters include contour parameters (length, height), slot information, indicator light layout data of the front view of the sample candidate device, cooling hole information, interface information, cooling fan data, external card slot information of the rear view, and size parameters, load data of the top view, etc. Each type of parameter needs to be recorded in detail to provide comprehensive support for feature extraction.

[0110] Optionally, in the embodiments of the present application, the sample candidate device feature set can include, but is not limited to, the sum of feature information comprehensively reflecting the multi-dimensional attributes of the sample candidate device after performing a feature extraction operation on the sample candidate device parameters. The set is consistent with the device feature set of the target device in terms of feature category and structure, covers all key features corresponding to the front view, rear view and top view of the sample candidate device, and each feature is presented in a quantifiable or encodable form, which is a direct basis for subsequent division of sample candidate device features according to spatial dimensions. The construction of the sample candidate device feature set needs to follow the same extraction rules as the target device feature set, such as calculating the contour coefficient and the proportion of heat dissipation holes, counting the number of slots and interface type codes, and obtaining the position features of indicator lights and fans through coordinate arrangement, to ensure comparability with the target device features. The core role of the set is to convert the original parameters of the sample candidate device into comparable features, providing a unified feature carrier for subsequent similarity calculation with the target device features.

[0111] Optionally, in the embodiments of the present application, the sample candidate device features can include, but are not limited to, feature subsets belonging to different view categories obtained by dividing the sample candidate device feature set according to spatial dimensions. These feature subsets correspond one-to-one to the multiple groups of device features (first group, second group, third group of device features) of the target device. Each group of sample candidate device features focuses on the core attributes of a certain spatial view of the sample candidate device, such as the sample candidate device features corresponding to the front view, the sample candidate device features corresponding to the rear view, and the sample candidate device features corresponding to the top view. The categories of each group of features are completely consistent with the categories of the corresponding group of features of the target device, ensuring that they can be accurately compared for similarity with the corresponding group of features of the target device. The division process of the sample candidate device features is the same as that of the multiple groups of device features of the target device, and they are classified according to the spatial dimensions of the front view, rear view and top view to avoid confusion of feature categories due to differences in division rules, which affects the matching accuracy. These features are the core units of the sample candidate device participating in similarity comparison, and through similarity calculation with the corresponding group of features of the target device, candidate features that meet the preset conditions are selected.

[0112] Optionally, in the embodiments of the present application, the target group device features can include, but are not limited to, a certain group of specific category features selected from the multiple groups of device features of the target device, which are currently required to be compared with the sample candidate device features, and the group of features focuses on the core attributes of a certain spatial view of the target device, for example, if the front view features are currently required to be compared, the target group device features are the first group of device features (front view features); if the rear view features are required to be compared, the target group device features are the second group of device features (rear view features); if the top view features are required to be compared, the target group device features are the third group of device features (top view features). The selection of the target group device features needs to follow the order of the matching process, and is usually selected in turn according to the view category, so as to ensure that all categories of the multiple groups of device features can be covered.

[0113] Optionally, in the embodiments of the present application, the target group sample candidate features can include, but are not limited to, a certain group of features in the sample candidate device features which have the same category as the target group device features, that is, if the target group device features are the front view features, the target group sample candidate features are the front view sample candidate features of the sample candidate device; if the target group device features are the rear view or top view features, the rear view or top view sample candidate features of the sample candidate device are selected correspondingly. Different target group sample candidate features are respectively derived from different sample candidate devices.

[0114] Exemplarily, the sample candidate device parameters of the pre-determined sample candidate device can be understood as follows: in the intelligent matching scene of the cabinet, in order to filter the multiple groups of candidate device features based on the multiple groups of device features of the target device, the original parameters of the sample candidate device (historical cabinet) pre-stored in the historical cabinet feature library need to be retrieved first, and these parameters are the basis for the extraction of the sample candidate device features, and need to be consistent with the collection dimensions of the target device parameters, so as to ensure the effectiveness of the subsequent feature comparison. In actual operation, the historical cabinet feature library stores the parameters of the sample candidate device in a standardized format, for example, through a database query instruction, the front view length and height, the rear view number of heat dissipation holes, the interface type, the top view length, width and bearing data of the sample candidate device A are retrieved; the front view slot information and indicator light layout, the rear view heat dissipation fan position and external card slot, the top view size and bearing data of the sample candidate device B are retrieved, and the like. These parameters need to be complete and accurate, so as to avoid incomplete feature extraction due to missing parameters, and to affect the reliability of the similarity calculation result, so as to prepare data for subsequent feature extraction operation.

[0115] Exemplarily, the above-mentioned obtaining the sample candidate device parameters of the predetermined sample candidate device; performing a feature extraction operation on the sample candidate device parameters to determine the sample candidate device feature set of the sample candidate device can be understood as after obtaining the sample candidate device parameters, the original parameters need to be processed and transformed according to the same rules and methods as the target device feature extraction, and the original parameters are transformed into quantifiable and comparable feature information, and then integrated to form the feature set of the sample candidate device, so as to ensure that the set is completely consistent with the device feature set of the target device in structure and category, and has the basis for direct comparison. For example, when performing feature extraction on the parameters of the sample candidate device C, the contour coefficient is calculated according to the length and height of the front view, the number of hard disk slot positions is counted and the coordinates are recorded, the coordinates are obtained by arranging the LED indicator light position; the proportion is calculated according to the number and size of the rear view cooling hole, the coordinates are recorded by coding the interface type, the number and position of the cooling fan are extracted, and the number and position of the external card slot are extracted; the length, width and bearing data of the top view are directly extracted, and these extracted features (contour coefficient, cooling hole proportion, length x width size, etc.) are integrated according to the category to form the sample candidate device feature set of the sample candidate device C. The feature category and arrangement order of the set are consistent with the target device feature set, and preparation is made for subsequent division of features according to spatial dimensions.

[0116] Exemplarily, the above-mentioned performing a feature extraction operation on the sample candidate device parameters to determine the sample candidate device feature set of the sample candidate device; dividing the features in the sample candidate device feature set according to spatial dimensions to obtain the sample candidate device features corresponding to the sample candidate device can be understood as after obtaining the sample candidate device feature set, the features in the set need to be classified according to the view category to which they belong according to the same spatial dimension standard (front view, rear view, top view) as the target device feature division, and divided into multiple independent sample candidate device features. Each group of features corresponds to the attributes of a spatial view of the sample candidate device, and one-to-one corresponds to the multiple device features of the target device. For example, when dividing the sample candidate device feature set of the sample candidate device D, the features related to the front view (contour coefficient, hard disk slot feature, LED indicator light feature) are classified into the front view sample candidate feature; the features related to the rear view (cooling hole proportion, interface feature, cooling fan feature, external card slot feature) are classified into the rear view sample candidate feature; the features related to the top view (length, width size feature, bearing feature) are classified into the top view sample candidate feature. Through such division, three independent sample candidate device features are obtained, which correspond to the first group, the second group and the third group of device features of the target device respectively, and provide clear feature units for subsequent grouping for similarity calculation.

[0117] Exemplarily, the above dividing the features in the sample candidate device feature set according to the spatial dimension to obtain the sample candidate device features corresponding to the sample candidate device, and determining the similarity between the target group device features and different target group sample candidate features can be understood as follows: after the sample candidate device features are divided, a group of device features of the target device is selected as the target group device features, and the target group sample candidate features of the same category as the target group device features are selected from all sample candidate devices, and the similarity between the target group device features and each target group sample candidate feature is calculated by using a preset similarity calculation method (such as cosine similarity), and each target group sample candidate feature is derived from a different sample candidate device, so as to ensure that enough historical features are covered for comparison. For example, the third group of device features (top view features, including length, width, and load-bearing capacity) of the target device are selected as the target group device features, and the top view sample candidate features of sample candidate devices A, B, C, and D (all of which are of the same category as the target group device features) are selected as different target group sample candidate features, the feature vectors of the target group device features and the top view feature vectors of each sample candidate device are normalized and then substituted into the cosine similarity formula to calculate the similarity values between the target group device features and the top view sample candidate features of sample candidate devices A, B, C, and D, thereby providing a quantitative basis for subsequent screening of qualified features.

[0118] Exemplarily, the above determining the similarity between the target group device features and different target group sample candidate features can be understood as follows: after the similarity between the target group device features and each target group sample candidate feature is calculated, each similarity value is compared with a preset similarity condition (such as a similarity threshold value and a key attribute deviation range) to screen all target group sample candidate features that meet the condition, and these qualified features constitute a group of candidate device features corresponding to the target group device features, and the group of candidate device features belongs to part of the multiple groups of candidate device features, and subsequent candidate feature screening corresponding to other groups of device features needs to be completed according to the same process. For example, the preset similarity condition is that the similarity value is not less than a certain threshold value, the length and width deviation of the top view is not more than a certain proportion of the target value, and the load-bearing capacity is not less than the target value, if the calculated similarity values between the target group device features (top view) and the top view sample candidate features of sample candidate devices B and D meet the threshold value requirement, and the length, width deviation, and load-bearing capacity all meet the condition, while the similarity with sample candidate devices A and C does not meet the requirement, then the top view sample candidate features of sample candidate devices B and D are determined as a group of candidate device features corresponding to the target group device features (the third group of device features), and this group of features is the third group of candidate device features in the multiple groups of candidate device features, and the above process is repeated for the first group and the second group of device features of the target device to obtain the complete multiple groups of candidate device features.

[0119] It should be noted that, in the process of determining a plurality of groups of candidate device features based on a plurality of groups of device features, in addition to obtaining sample candidate device parameters, extracting feature sets, dividing features, and calculating similarity according to the above-mentioned process, each link can be flexibly adjusted according to the actual needs of the data center, and the present application does not make specific limitations. For example, when obtaining sample candidate device parameters, if the sample candidate device is an old cabinet of a long time ago, some original parameters are missing, a parameter completion mechanism can be introduced to reasonably complete the missing parameters based on the parameter distribution law of the same type of cabinet (such as the load data range of the same size cabinet, the distribution characteristics of the heat dissipation holes of the similar structure cabinet), to ensure the integrity of the parameters, and to avoid incomplete feature extraction due to missing parameters, for example, the load data of the overhead view of a certain old sample candidate device is missing, the load data distribution of the same size and material cabinet can be referred to to complete a load value in a reasonable interval, and then subsequent feature extraction is performed. For another example, when calculating the similarity, in addition to using cosine similarity, other similarity calculation methods can also be selected according to the feature type, if the parameters in the target group of device features are mainly discrete (such as interface type coding, slot number), the Jaccard similarity can be used to calculate the feature overlap, if they are mainly continuous (such as size, load), the Euclidean distance can be used to calculate the feature difference, for example, when calculating the similarity between the target group of device features (rear view interface feature, including multiple interface type coding) and the target group of sample candidate features, the Jaccard similarity is used to calculate the overlap ratio of the interface types, if the overlap ratio is higher than a preset threshold, it is determined that the similarity condition is met. In addition, in the setting of the preset similarity condition, in addition to the fixed threshold, a dynamic threshold mechanism can also be introduced to adjust the threshold according to the number of sample candidate devices and the feature distribution, if the number of sample candidate devices is large, the threshold can be appropriately increased to select better candidate features, if the number of sample candidate devices is small, the threshold can be appropriately reduced to ensure that there are enough candidate features to participate in subsequent combination, for example, when the number of features of the same category in the sample candidate device and the target group of device features exceeds a certain number, the similarity threshold is increased, otherwise it is reduced, to ensure that the matching result is accurate and covers the needs, these diversified adjustment methods can be flexibly applied according to the actual scene, to improve the flexibility and adaptability of determining a plurality of groups of candidate device features.

[0120] Through the embodiments of the present application, the sample feature hierarchical matching mechanism is adopted, the sample feature groups are generated by performing spatial dimension division on the sample candidate device parameters, the candidate feature groups that meet the threshold condition are selected by calculating the similarity (such as cosine similarity) between the target group of device features and the sample feature groups; by establishing the category alignment rule of “target group-sample group”, the purpose of accurate comparison of heterogeneous features is achieved, thereby solving the feature mispositioning problem in historical data reuse, and significantly improving the reliability of the candidate scheme.

[0121] In an example embodiment, determining the similarity between the target group device feature and the different target group sample candidate features respectively comprises:

[0122] determining a sub-similarity between each target device feature in the target group device feature and each target candidate feature in the target group sample candidate feature, wherein the target device feature and the target candidate feature belong to the same feature dimension;

[0123] merging the sub-similarities to determine the similarity between the target group device feature and the target group sample candidate feature.

[0124] Optionally, in the embodiments of the present application, the sub-similarity can include but is not limited to, in the cabinet intelligent matching scenario, the similarity degree quantization value between the specific features belonging to the same feature dimension in the target group device feature (such as the features of a certain view of the target cabinet) and the target group sample candidate feature (such as the features of the same view of the historical cabinet). These feature dimensions are the basic units that constitute the target group device feature and the target group sample candidate feature, and each dimension corresponds to a specific attribute of the cabinet, such as the contour coefficient dimension in the front view, the hard disk slot number dimension, the heat dissipation hole proportion dimension in the rear view, the interface type dimension, etc. The calculation of the sub-similarity needs to be carried out independently for each independent feature dimension, and the calculation method suitable for the feature type of this dimension (such as numerical difference calculation for continuous features and matching degree calculation for discrete features) is adopted, and the result directly reflects the similarity of the two features in a single attribute, which is the basis for subsequent merging to obtain the overall similarity. The accuracy of the sub-similarity is crucial to the overall similarity result, and it is necessary to ensure that the sub-similarity calculation logic of each feature dimension is consistent with the characteristics of this dimension attribute, and avoid the deviation of similarity judgment due to improper calculation method.

[0125] Optionally, in the embodiments of the present application, the feature dimensions can include, but are not limited to, in the cabinet intelligent matching scenario, independent classification units for describing a certain specific attribute of the cabinet obtained after the target group equipment features and the target group sample candidate features are disassembled, each feature dimension focuses on the specific characteristics of a certain aspect of the cabinet, and is a basic component of the equipment feature vector. The division of the feature dimensions needs to be based on the attribute characteristics and matching requirements of the cabinet, to ensure that each dimension has a clear description object and a quantitative method. For example, the front view features can be divided into contour coefficient, hard disk slot number, LED lamp number, etc. dimensions, the rear view features can be divided into heat dissipation hole proportion, interface type number, heat dissipation fan number, etc. dimensions, and the top view features can be divided into length, width, load capacity, etc. dimensions. Different feature dimensions correspond to different functions or structural attributes of the cabinet, and the importance thereof can change due to actual scenarios (such as the temperature of the machine room and the type of equipment). After the sub-similarity is calculated, the sub-similarity needs to be integrated in combination with the weight of the feature dimension, to obtain an overall similarity that is more suitable for actual requirements. The explicit division of the feature dimensions disassembles the complex equipment features into units that can be independently calculated and evaluated, reduces the complexity of the similarity calculation, and also provides the possibility of dynamically adjusting the weight (such as increasing the weight of the heat dissipation related dimension according to the temperature of the machine room).

[0126] Optionally, in the embodiments of the present application, the target equipment features can include, but are not limited to, specific quantitative values or description information belonging to a certain specific feature dimension in the target group equipment features (such as the front view, rear view or top view features of the target cabinet), which are basic units of the target group equipment features. Each target equipment feature corresponds to a specific attribute of a certain dimension of the cabinet, such as “contour coefficient = 0.8” and “hard disk slot number = 12” in the front view features, “heat dissipation hole proportion = 30%” and “interface type number = 3 types” in the rear view features, and “length = 120 cm” and “load capacity = 500 kg” in the top view features. The target equipment features need to belong to the same feature dimension as the target candidate features (features of the same dimension of the sample candidate equipment), to calculate the sub-similarity. For example, the target equipment features of the “hard disk slot number” dimension can only calculate the sub-similarity with the target candidate features of the “hard disk slot number” dimension in other features. The acquisition of these features needs to be through the extraction and standardization processing of the target equipment parameters, to ensure that the format and precision thereof are consistent with those of the target candidate features, so as to facilitate effective similarity comparison. The accuracy of the target equipment features directly affects the calculation result of the sub-similarity, and further affects the reliability of the overall similarity, therefore, the standardization process needs to be strictly followed in the extraction process, to avoid data errors.

[0127] Optionally, in the embodiments of the present application, the target candidate features can include, but are not limited to, specific quantitative values or description information belonging to a certain feature dimension in the target group sample candidate features (such as the features of a certain view of the historical cabinet), which are basic units constituting the target group sample candidate features and belong to the same feature dimension as the target device features, and are used to calculate the sub-similarity with the target device features. The target candidate features are derived from the extraction and standardization processing of the sample candidate device parameters, and the format and precision thereof are consistent with the target device features, thereby ensuring the effectiveness of the sub-similarity calculation. For example, the target candidate feature of the front view of the sample candidate device “hard disk slot number = 11” can calculate the sub-similarity with the target device feature of the front view of the target device “hard disk slot number = 12”; and the target candidate feature of the rear view of the sample candidate device “heat dissipation hole ratio = 28%” can calculate the sub-similarity with the target device feature of the rear view of the target device “heat dissipation hole ratio = 30%”. Different target candidate features correspond to different sample candidate devices, and through the sub-similarity calculation with the target device features, the sample features similar to the target device in a certain dimension can be screened out, thereby providing a basis for subsequent merging of the overall similarity and determination of the candidate device features. The integrity and accuracy of the target candidate features depend on the standardized process of the sample candidate device parameter acquisition and feature extraction, and it is necessary to ensure that the information of each feature dimension is recorded completely.

[0128] According to the embodiments of the present application, the hierarchical similarity fusion mechanism is adopted, the feature group is disassembled into sub-dimensions (such as the hard disk slot coordinates) and the sub-similarity is calculated, and then the group similarity is merged, thereby achieving the purpose of precise matching of local features through independent comparison of multi-dimensional features and aggregation of results.

[0129] In one example embodiment, the sub-similarity between each target device feature in the target group device features and each target candidate feature in the target group sample candidate features is determined, including:

[0130] A target feature vector is generated based on the target device features, and a candidate feature vector is generated based on the target candidate features;

[0131] The distances of the target feature vector and the plurality of candidate feature vectors in the target vector space are determined, and the sub-similarity is determined based on the distances.

[0132] Optionally, in the embodiments of the present application, the target feature vector can be understood as an ordered numerical set formed by integrating the target device features (i.e. the specific quantified information of a certain specific feature dimension in the target group device features) according to a preset rule, which is used for mathematical calculation. Its core function is to convert the unstructured features of the cabinet into a structured vector form, so as to compare the similarity in the vector space. In the cabinet intelligent matching scenario, the target device features usually revolve around the front view, rear view and top view of the cabinet. Therefore, the target feature vector needs to cover the quantified values of the key feature dimensions in the corresponding view, and the order of the elements in the vector is consistent with the definition order of the feature dimensions, ensuring the consistency and comparability of the subsequent calculation with the candidate feature vector.

[0133] Optionally, in the embodiments of the present application, the candidate feature vector can be understood as an ordered numerical set formed by integrating the target candidate features (i.e. the specific quantified information of a certain specific feature dimension in the target group sample candidate features, which comes from the sample candidate device) according to the same rule as the target feature vector. It is completely consistent with the target feature vector in terms of the number of elements and corresponding feature dimensions, which is the prerequisite for distance calculation between the two vectors in the vector space. In the cabinet scenario, the sample candidate device is a historical cabinet, and the target candidate features are standardized features extracted from the historical cabinet parameters. Therefore, the candidate feature vector needs to strictly follow the construction rule of the target feature vector to ensure that each element can form an accurate comparison with the corresponding element of the target feature vector. For example, the target feature vector corresponds to the "outline coefficient-hard disk slot number-LED light number" of the front view, and the candidate feature vector also needs to integrate the corresponding features of the front view of the historical cabinet in this order.

[0134] Optionally, in the embodiments of the present application, the target vector space can be understood as a multi-dimensional mathematical space constructed with the elements (i.e. the quantified values of each feature dimension) of the target feature vector and the candidate feature vector as the coordinate axes. Each feature dimension corresponds to a coordinate axis in the space, and each point in the vector space uniquely corresponds to a feature vector (target feature vector or candidate feature vector). Its core function is to provide a geometric environment for vector distance calculation, and the distance between two points in the space directly reflects the similarity of two feature vectors. In the cabinet scenario, the dimension of the target vector space is determined by the number of elements of the feature vector. For example, the front view feature vector contains 3 elements (outline coefficient, hard disk slot number, LED light number), so the target vector space is a three-dimensional space. The rear view feature vector contains 4 elements (heat dissipation hole proportion, interface type coding number, heat dissipation fan number, external card slot number), so the target vector space is a four-dimensional space, and the meaning of each coordinate axis is associated with the physical meaning of the corresponding feature dimension.

[0135] Optionally, in the embodiments of the present application, the above-mentioned distance can be understood as the geometric distance between the point corresponding to the target feature vector and the point corresponding to the candidate feature vector in the target vector space, and the size of the distance is negatively correlated with the degree of similarity between the two feature vectors, that is, the smaller the distance, the more similar the cabinet features corresponding to the two feature vectors; the larger the distance, the lower the degree of similarity. In the cabinet intelligent matching scene, the calculation method of the distance needs to be selected according to the type of the feature vector, and the commonly used methods include the cosine distance (suitable for measuring the similarity of vector direction, ignoring the difference in vector length, suitable for the scene where the different dimensions of the cabinet feature have similar weights) and the Euclidean distance (suitable for measuring the comprehensive size of the numerical difference of the vector elements, suitable for the scene where the feature numerical accuracy is required to be high), and the feature vector needs to be standardized before calculation to avoid the calculation deviation caused by the difference in the numerical range of the feature dimension.

[0136] By the embodiments of the present application, the similarity is quantitatively evaluated by vector space mapping (target feature vector and candidate feature vector) and space distance calculation, so as to guarantee the calculation efficiency while avoiding the defect of global feature weight imbalance.

[0137] In one exemplary embodiment, a feature extraction operation is performed on sample candidate equipment parameters to determine a sample candidate equipment feature set of the sample candidate equipment, including:

[0138] A feature extraction operation is performed on sample candidate equipment parameters to determine a sample candidate equipment feature set of the sample candidate equipment;

[0139] The mean and variance are calculated for sample candidate equipment features belonging to different sample candidate equipment feature sets and belonging to the same feature dimension;

[0140] The features in each sample candidate equipment feature set are standardized based on the mean and variance to obtain an updated sample candidate equipment feature set.

[0141] Optionally, in the embodiments of the present application, the above-mentioned mean can be understood as the arithmetic mean of the numerical values of all sample candidate equipment features belonging to the same feature dimension in the plurality of sample candidate equipment feature sets, which is used to reflect the overall distribution center level of the feature dimension in the historical sample candidate equipment, and is one of the core parameters for subsequent standardization processing of the sample candidate equipment features. In the cabinet scene, the same feature dimension refers to the same attribute of different historical cabinets under the same view, such as the "outline coefficient" dimension of all sample candidate equipment front views, the "heat dissipation hole proportion" dimension of rear views, and the "load bearing capacity" dimension of top views, and when calculating the mean, the numerical values of all effective sample candidate equipment features under the dimension need to be collected to ensure that the sample size is large enough to reflect the true distribution.

[0142] Optionally, in the embodiments of the present application, the above-mentioned variance can be understood as the arithmetic mean of the square of the deviation between the numerical value of all sample candidate device features belonging to the same feature dimension in the plurality of sample candidate device feature sets and the mean value of the dimension, which is used to reflect the numerical value dispersion degree of the feature dimension in the historical sample candidate device, i.e., the fluctuation range of the feature numerical value around the mean value, which is a key parameter for realizing feature standardization processing in cooperation with the mean value. In the cabinet scene, the calculation of the variance needs to be based on the mean value result of the same feature dimension, and the numerical stability of the feature dimension can be judged through the variance, for example, a smaller variance indicates that the feature dimension is distributed in a concentrated manner in the historical cabinet, and the numerical difference is small; a larger variance indicates that the distribution is dispersed, and the numerical difference is large, which provides a basis for eliminating the numerical range difference in subsequent standardization processing.

[0143] Optionally, in the embodiments of the present application, the above-mentioned standardization processing can be understood as a process of converting the numerical values of all sample candidate device features under the same feature dimension into standard normal distribution numerical values with a mean value of 0 and a standard deviation of 1 based on the mean value and the variance of the same feature dimension, the core purpose of which is to eliminate the weight deviation caused by the numerical range difference of different feature dimensions, so that all feature dimensions have equal influence in the similarity calculation, and at the same time, the numerical scales of different sample candidate device features are unified, and the effectiveness of cross-cabinet and cross-dimension feature comparison is improved. In the cabinet scene, the standardization processing needs to be performed separately for each feature dimension, and the mean value and the variance of the dimension need to be used, and the processing formula is usually “standardized numerical value=(original numerical value-mean value) / standard deviation” (standard deviation is the square root of variance), and the processed data is more suitable for subsequent vector space distance calculation and similarity evaluation.

[0144] Optionally, in the embodiments of the present application, the above-mentioned updated sample candidate device feature set can be understood as a feature set containing standardized numerical values of each feature dimension obtained after the original feature set of the sample candidate device is subjected to standardization processing, which is consistent with the original feature set in terms of the number of feature dimensions and the corresponding view category, but all feature numerical values have been converted into standard normal distribution form, which is a core data set for subsequent similarity comparison with the target device features. In the cabinet scene, the updated feature set needs to be stored by view category (front view, rear view, top view), and when a preset number (such as 10) of sample candidate devices are added, the mean value and the variance of the related feature dimensions need to be recalculated, and the feature set of all sample candidate devices needs to be re-standardized based on the new mean value and the variance, so as to realize dynamic updating of the feature set and ensure data timeliness.

[0145] Through the embodiments of the present application, a dynamic standardization updating mechanism is adopted, periodic mean and variance calculation is performed on the sample features (such as updating every 10 cabinets are added), and normal distribution standardization processing (mean 0, standard deviation 1) is performed on the feature values; by establishing a feature distribution parameter table (such as Table 1) and calling it in real time, the purpose of uniformity of cross-cabinet data comparability is achieved, thereby solving the problem of invalidation of historical data caused by equipment iteration and ensuring the stability of long-term matching.

[0146] In an example embodiment, the above method further comprises:

[0147] adding a preset weight coefficient to each feature corresponding to each feature dimension in the target group sample candidate features;

[0148] dynamically adjusting the preset weight coefficient to a target weight coefficient based on a preset environmental parameter threshold;

[0149] optimizing the features corresponding to each feature dimension in the target group sample candidate features according to the target weight coefficient.

[0150] Optionally, in the embodiments of the present application, the above-mentioned preset weight coefficient can be understood as a numerical value that is pre-set for each feature dimension in the target group sample candidate features and is used to measure the importance of the dimension in similarity calculation, which usually ranges from 0 to 1, and the sum of the preset weight coefficients of all feature dimensions under the same view is 1. The core role is to allocate basic importance to each feature dimension when real-time environmental factors are not considered, ensuring that the similarity calculation can give priority to the influence of key features. In the cabinet scene, the setting of the preset weight coefficient needs to be based on the regular importance of each feature dimension of the cabinet. For example, in the front view, the "hard disk slot number" directly affects the equipment installation, and the weight coefficient is higher than that of the "LED light number". In the rear view, the "interface type coding number" and the "number of cooling fans" are related to the connection and cooling of the equipment, and the weight coefficient is higher than that of the "external card slot number". The preset weight coefficient needs to be verified through a large amount of historical matching data to ensure its rationality.

[0151] Optionally, in the embodiments of the present application, the above-mentioned preset environmental parameter threshold can be understood as a pre-set environmental parameter critical value for triggering dynamic adjustment of the weight coefficient of the feature dimension. The threshold is determined based on the actual environmental needs of the data center cabinet operation. When the real-time environmental parameter exceeds or is lower than the threshold, it means that the degree of influence of the corresponding environmental factor on the cabinet function has changed, and the weight coefficient of the related feature dimension needs to be adjusted to adapt to the actual needs. In the cabinet scene, the environmental parameter mainly refers to the key external conditions that affect the operation of the cabinet, such as the temperature and humidity of the computer room. Among them, the temperature of the computer room is the core environmental parameter, and its preset threshold needs to be set in combination with the heat dissipation capacity of the cabinet and the equipment operating temperature requirement, to ensure that the weight adjustment triggered by the threshold can effectively adapt to the change of heat dissipation demand.

[0152] Optionally, in the embodiments of the present application, the target weight coefficient can be understood as a value for measuring the importance of the feature dimension in the current environment after dynamically adjusting the preset weight coefficient based on the preset environmental parameter threshold. The core difference between the target weight coefficient and the preset weight coefficient is that the influence of the real-time environment factor is integrated. The target weight coefficient of the same feature dimension will change with the change of the environmental parameter, and the sum of the target weight coefficients of all feature dimensions in the same view still remains 1, ensuring the rationality of the weight distribution. In the cabinet scene, the adjustment of the target weight coefficient is mainly aimed at the feature dimensions that are greatly affected by the environmental parameter. For example, when the temperature of the computer room exceeds the preset threshold, the weight coefficients of the rear view "heat dissipation hole ratio" and "heat dissipation fan quantity" are increased, and the weight coefficients of other feature dimensions that are less affected by the environment (such as the number of external plug-in card slots) are reduced, so as to highlight the importance of the heat dissipation feature.

[0153] Through the embodiments of the present application, an environment adaptive weighting mechanism is adopted. The initial weight coefficient (such as the front view hard disk weight 0.6) is preset for the feature dimension, and the weight is dynamically adjusted according to the environmental parameter (such as the temperature of the computer room) (such as increasing the heat dissipation weight when the temperature is greater than 35°C). The sum is maintained constant through weight normalization processing, which achieves the purpose of dynamic optimization of feature importance, so as to realize the real-time cooperation of the matching strategy and the actual demand, and significantly improve the scene adaptability of the scheme.

[0154] In one example embodiment, a candidate device is determined based on a plurality of groups of candidate device features, and device information of a target candidate device is generated by combining the candidate devices, including:

[0155] A front view corresponding to a first candidate device, a rear view corresponding to a second candidate device, and a top view corresponding to a third candidate device are determined based on the plurality of groups of candidate device features;

[0156] The front view, the rear view, and the top view are scaled and spliced according to the size information of the target device to obtain a combined view, wherein the combined view labels the first candidate device, the second candidate device, and the third candidate device, and the combined view is used to represent the device information of the target candidate device.

[0157] Optionally, in the embodiments of the present application, the size information of the target device can be understood as the geometric size data for describing the spatial dimensions of the target cabinet (target device), mainly including the length and height of the front view, the length and height of the rear view, the length and width of the top view, which directly determines the spatial occupation range and appearance form of the target cabinet, and is the core basis for subsequent scaling and splicing of the view of the sample candidate device, ensuring that the combined view after splicing can accurately reflect the actual size of the target cabinet. In the cabinet scene, the size information of the target device needs to be collected by high-precision measurement tools (such as laser range finders), and needs to be consistent with the size measurement standard of the sample candidate device view (such as taking the cabinet outer contour size as the standard, without including protruding parts), to avoid scaling proportion deviation caused by measurement standard difference, affecting the accuracy of the combined view.

[0158] Optionally, in the embodiments of the present application, the combined view can be understood as a visual image that completely reflects the appearance and view characteristics of the target candidate device, which is formed by scaling the front view corresponding to the first candidate device, the rear view corresponding to the second candidate device, and the top view corresponding to the third candidate device according to the size information of the target device, and then splicing according to the spatial structure logic of the cabinet. It not only contains the feature details of each view, but also labels the source of each sample candidate device, which is a direct presentation form of the device information of the target candidate device. In the cabinet scene, the splicing of the combined view needs to follow the physical structure relationship of the cabinet, for example, the front view is located on the front, the rear view is located on the back, and the top view is located on the top. The size of each view after scaling in the corresponding direction needs to be completely consistent with the size of the target device, to ensure that there is no obvious misplacement after splicing, and the labeling information needs to be clear and visible, to facilitate the user to understand the source of each view.

[0159] Through the embodiments of the present application, by using a multi-source view intelligent fusion mechanism, the front view, the rear view and the top view of the optimal candidate device are scaled and spliced into a combined view according to the size of the target device. Through source labeling (such as cabinet A front view + cabinet B rear view) and proportion adaptation, the purpose of physical compatibility of cross-device local characteristics is achieved, so as to generate a target candidate device scheme that can be directly deployed, greatly reducing the cost of manual design.

[0160] In one example embodiment, after scaling and splicing the front view, the rear view and the top view according to the size information of the target device to obtain the combined view, the above method further comprises:

[0161] Generating front view contribution information, rear view contribution information and top view contribution information based on the similarity contribution degrees of different feature dimensions in multiple groups of candidate device characteristics;

[0162] Displaying the front view contribution information, the rear view contribution information and the top view contribution information on the combined view.

[0163] Optionally, in the embodiments of the present application, the similarity contribution degree can be understood as the contribution degree of each feature dimension in the target group sample candidate feature to the overall similarity, and the quantitative value thereof is equal to the product of the target weight coefficient of the feature dimension and the corresponding sub-similarity, for reflecting the role of a certain feature dimension in the similarity matching between the target group device feature and the target group sample candidate feature. The higher the contribution degree is, the greater the influence of the feature dimension on the matching result is. In the cabinet scenario, the calculation of the similarity contribution degree needs to be based on the dynamically adjusted target weight coefficient and the sub-similarity of each feature dimension, and the sum of the similarity contribution degrees of all feature dimensions under the same view is equal to the overall similarity of the view. By analyzing the contribution degree, it can be determined which feature dimensions are the key factors and which are the secondary factors in the matching, thereby providing a basis for subsequent optimization of the matching strategy.

[0164] Optionally, in the embodiments of the present application, the front view contribution degree information can be understood as an information set generated based on the similarity contribution degrees of each feature dimension in the front view, for describing the contribution of the front view and each feature dimension to the matching result. The information set not only contains the overall similarity of the front view (the sum of the contribution degrees of all feature dimensions), but also contains the specific contribution degree value and proportion of each feature dimension (such as the contour coefficient, the number of hard disk slots, and the number of LED lights), and is a quantitative description of the front view matching process and result. In the cabinet scenario, the front view contribution degree information needs to correspond to the front view part in the combined view. Through the information, the role of each feature dimension in the front view in the matching can be intuitively understood. For example, if the contribution degree proportion of the “number of hard disk slots” is the highest, it indicates that this feature dimension is the key factor in the front view matching, thereby providing a basis for user evaluation of the front view matching quality and optimization of the front view feature selection.

[0165] Optionally, in the embodiments of the present application, the rear view contribution degree information can be understood as an information set generated based on the similarity contribution degrees of each feature dimension in the rear view, for describing the contribution of the rear view and each feature dimension to the matching result. The content of the information set is similar to that of the front view contribution degree information, and contains the overall similarity of the rear view, the specific contribution degree value and proportion of each feature dimension (such as the proportion of the heat dissipation hole, the number of interface type encodings, the number of heat dissipation fans, and the number of external card slots), and focuses on the role of each feature dimension in the rear view matching process, especially the contribution of the heat dissipation related features which are greatly affected by the environmental parameters. In the cabinet scenario, the rear view contribution degree information needs to highlight the contribution changes of the heat dissipation related features (the proportion of the heat dissipation hole and the number of heat dissipation fans). For example, when the temperature of the machine room exceeds the preset threshold, the contribution degree proportion of the heat dissipation related features will be significantly improved. Through the information, it can be directly judged whether the rear view matching is suitable for the current environmental requirements.

[0166] Optionally, in the embodiments of the present application, the top view contribution degree information can be understood as a set of information generated based on the similarity contribution degree of each feature dimension of the top view, for describing the contribution of the overall top view and each feature dimension to the matching result, which includes the overall similarity of the top view, the specific contribution degree value and proportion of each feature dimension (such as length, width, load-bearing capacity), and focuses on the role of spatial size and load-bearing capacity features in the top view matching process. These features are directly related to the spatial adaptability of the cabinet and the safety of equipment support. In the cabinet scene, the top view contribution degree information needs to correspond to the top view part in the combined view. Through this information, the importance of each feature dimension of the top view in the matching can be intuitively understood. For example, if the contribution degree proportion of “load-bearing capacity” is the highest, it means that this feature dimension is the key factor in the top view matching, and the load-bearing capacity of the sample candidate equipment needs to be focused on whether it meets the target demand.

[0167] Through the embodiments of the present application, the contribution degree transparent transmission mechanism is adopted, the analysis report is generated by calculating the similarity contribution degree of each view (such as the front view LED layout contribution value), and is dynamically labeled on the combined view; through visual presentation of the key feature influence factor (such as the reuse value of the high-heat-dissipation rear view), the purpose of strengthening design decision assistance is achieved, thereby providing the user with data basis for iterative optimization, and improving the engineering landing efficiency of the scheme.

[0168] The present application will be further explained and described below in combination with specific examples:

[0169] The present application relates to the technical field of data center equipment management, and particularly focuses on an intelligent matching method and system for cabinets. In data center equipment management, the matching and deployment of cabinets is a key link, which is directly related to the management efficiency, heat dissipation performance, and area characteristics of the equipment and many other aspects. Therefore, how to realize accurate matching of cabinet features, especially the composite feature processing of new heterogeneous cabinets, has become a problem to be solved at present.

[0170] The closest prior art is the existing single-cabinet global matching scheme. However, this scheme has two significant limitations. The existing scheme often relies on global features in the matching process, but for new heterogeneous cabinets, their features are more complex and diverse, including hole proportion, contour coefficient, and many other dimensions. These composite features are difficult to accurately process in the global matching scheme, resulting in poor matching effect. In actual scenarios, new cabinets often have local similarities with historical cabinets in some views or features, and the global matching scheme cannot effectively distinguish these local similar features, thereby causing feature confusion and affecting the accuracy and reliability of the matching.

[0171] Therefore, the present application proposes a cabinet intelligent matching method and system based on multi-dimensional feature combination, aiming to solve the problems existing in the prior art and realize accurate matching and efficient management of cabinet features. The present application proposes a cabinet intelligent matching method and system based on multi-dimensional feature combination, especially for the heterogeneous cabinet matching problem in the field of data center equipment management, specifically including:

[0172] Cabinet feature decoupling: The cabinet features are decomposed into front view, rear view, top view, etc. independent dimensions. An independent feature vector space is established for each dimension, such as front view contour coefficient, hard disk slot, LED layout feature; the rear view includes the proportion of cooling holes, interface type, interface position, cooling fan feature, and external card slot feature; the top view includes area feature and load capacity feature. To accurately describe and extract the cabinet features.

[0173] Dynamic combination matching algorithm: Through the dynamic combination matching algorithm, multi-dimensional parallel matching is performed to improve the matching efficiency and accuracy. A weight adaptive mechanism is introduced to dynamically adjust the matching weight according to the view importance (such as the influence of room temperature on the proportion of cooling holes). The optimal combination scheme is output, such as "front view matching A cabinet + rear view matching B cabinet + top view matching C cabinet", to meet the matching needs of complex heterogeneous cabinets.

[0174] Intelligent rendering system: Establish a feature combination relationship graph to intuitively show the correlation between the features of each view and the matching source. Support "jigsaw puzzle" interface rendering, label the matching source of each view with different colors, and facilitate user understanding and operation. Provide similarity contribution visualization analysis to help users evaluate and optimize the matching scheme.

[0175] As shown in Figure 3 , the specific implementation process is as follows:

[0176] S1, cabinet feature decoupling:

[0177] Step one: The physical features of the cabinet are divided according to the spatial dimensions, mainly into front view, rear view and top view. The complex cabinet features are simplified into manageable independent parts.

[0178] Step two: For each view, an independent feature vector space is established. The front view includes contour coefficient, hard disk slot, and LED layout feature; the rear view includes the proportion of cooling holes, interface type, interface position, cooling fan feature, and external card slot feature; the top view includes area feature and load capacity feature.

[0179] Step three: The feature vectors of each view are accurately extracted from the original data to provide basic data for the subsequent matching algorithm. Among them:

[0180] The front view feature vector calculation method is as follows:

[0181] Contour coefficient: Extract the contour of the front view of the cabinet: length, height, calculate the ratio of its perimeter to area, get the contour coefficient. Indicate the compactness and shape characteristics of the cabinet.

[0182] Hard disk feature: Count the number of hard disk slots in the front view, and record the position of each slot (relative coordinates).

[0183] LED layout feature: Identify the position and number of LED lights, use a two-dimensional coordinate array to represent the position of each LED light, such as: LED light position 1_x, LED light position 1_y,..., LED light position n_x, LED light position n_y.

[0184] Finally, the feature vector of the front view is obtained: front view feature vector = [contour coefficient, hard disk slot number, hard disk slot position 1_x, hard disk slot position 1_y, LED light number, LED light position 1_x, LED light position 1_y,..., LED light position n_x, LED light position n_y].

[0185] The back view feature vector calculation method is as follows:

[0186] Radiator hole feature: Count the number of radiator holes, the area of all radiator holes and the proportion of the area of the back view.

[0187] Interface feature: List all interface types on the back view (such as USB, Ethernet, Universal Serial Bus Type-C (Type-C) and other types), use type code to represent, and record the two-dimensional coordinates of each interface (the starting coordinates of the lower left corner of the back view are (0, 0)).

[0188] Radiator fan feature: Identify the number of radiator fans and record their two-dimensional coordinates.

[0189] External card slot feature: Count the number and position of external card slots.

[0190] Finally, the feature vector of the back view is obtained: back view feature vector = [radiator hole number, radiator hole proportion, interface type code 1,..., interface type code m, interface position 1_x, interface position 1_y,..., interface position n_x, interface position n_y, radiator fan number, radiator fan position_x, radiator fan position_y, external card slot number, external card slot position 1_x, external card slot position 1_y,..., external card slot position p_x, external card slot position p_y].

[0191] The top view feature vector calculation method is as follows:

[0192] Area feature: length and width of the computer cabinet top view.

[0193] Load-bearing capacity feature: record the load-bearing capacity value of the cabinet.

[0194] Finally, the feature vector of the top view is obtained: top view feature vector = [length, width, load-bearing capacity value].

[0195] For all historical cabinets, the feature vector is extracted according to the above features, and the feature vector is recorded in the data table (front view feature data table, rear view feature data table, top view feature data table), as shown in Figure 4 .

[0196] Step four: standardize the feature vector, convert the value of each feature dimension to a standard normal distribution with mean 0 and standard deviation 1. The specific steps are as follows taking the front view as an example:

[0197] (1) Calculate the mean and standard deviation of each feature dimension:

[0198] Calculate the mean , (where m is the number of samples, n is the number of feature dimensions. For the jth feature dimension in the feature vector (j = 1, 2,..., n), X ij is the value of the ith sample in the jth feature dimension).

[0199] Calculate the standard deviation .

[0200] (2) Standardize each feature dimension:

[0201] Using the calculated mean and standard deviation, standardize each feature dimension, that is: .

[0202] Where is the standardized value of the ith sample in the jth feature dimension.

[0203] As shown in Figure 5 , the mean μ and standard deviation σ of each key feature need to be pre-calculated and stored in the key feature mean and standard deviation data table, and directly called each time. Update once every 10 new cabinet data.

[0204] (3) Generate the standardized feature vector:

[0205] Each sample and each feature dimension in the feature vector X is standardized as described above, and the standardized feature vector X' is finally obtained. And X' is overwritten in the front view feature data table, and the rest of the rear view and top view are processed according to the above standardization process.

[0206] S2, dynamic combination matching algorithm:

[0207] Step one: the initial weights corresponding to the profile coefficient, hard disk feature, and LED layout feature in the front view are 0.2, 0.6, and 0.2 respectively; the initial weights corresponding to the heat dissipation hole feature, interface feature, heat dissipation fan feature, and external card slot feature in the rear view are 0.1, 0.3, 0.3, and 0.3 respectively; the initial weights corresponding to the area feature and load capacity feature in the top view are 0.5 and 0.5 respectively.

[0208] Step two: introducing a weight self-adaptive mechanism. According to the current environmental parameter room temperature, the matching weights of the heat dissipation hole feature and the heat dissipation fan feature are dynamically adjusted.

[0209] Temperature threshold T 阈值 = 35℃, heat dissipation hole initial weight W 散热孔初始 = 0.1, heat dissipation fan initial weight W 风扇初始 = 0.3, when the temperature exceeds the threshold, the weights of the heat dissipation hole and the fan are increased. The heat dissipation hole weight adjustment formula is:

[0210] The heat dissipation fan weight adjustment formula is:

[0211] Weight normalization: in order to ensure that the sum of the weights of the rear view features remains unchanged, weight normalization is performed:

[0212] The sum of all feature weights is calculated: wherein, is the weight of each feature.

[0213] The weight of each feature is normalized:

[0214] Step three: based on the comparison of the feature vectors of the front view, rear view, and top view of the target cabinet and each candidate cabinet, the similarity values of each view are output by the combination optimization algorithm according to the weights, and the view scheme with the maximum similarity value is taken as the target scheme. The three-layer screening mechanism is shown in Figure 6 , and the specific steps are as follows:

[0215] (1) Feature vector preparation.

[0216] Target cabinet feature vector: extract the feature vectors of the front view, rear view, and top view of the target cabinet, and denote them as V target front , V target back , and V target top

[0217] ​​​​Candidate cabinet feature vector: For each candidate cabinet, the feature vectors of its front view, rear view, and top view are also extracted, respectively denoted as V i front , V i back , V i top , where i represents the number of the candidate cabinet.

[0218] Candidate cabinet feature vector cleaning: According to the requirements of the target cabinet on area and load-bearing capacity, the data of the area feature and load-bearing capacity feature in the top view of the candidate cabinet that do not meet the requirements are cleaned up. It is ensured that the remaining top view feature vector set meets the requirements.

[0219] Vector normalization: In order to improve the accuracy of similarity calculation, the feature vector is normalized so that the value of each dimension is in the same order of magnitude. The normalization formula is as follows:

[0220] , where V is the original feature vector, μ is the mean of the vector, and σ is the standard deviation of the vector.

[0221] Similarity measurement: cosine similarity is used as the similarity measurement method, and its formula is:

[0222] .

[0223] , where A and B are two feature vectors to be compared, ⋅ represents the dot product, and ||A|| and ||B|| represent the modulus of vectors A and B, i.e., the length of the vector.

[0224] View similarity calculation: Based on the calculation method in step (3) above, the similarity of each feature vector is calculated for the target cabinet and the candidate cabinet to be compared from the front view, rear view, and top view, respectively, denoted as S i , S , and S , respectively. Thus, the similarity set S 前 front=[S 前1 , S 前2 , S 前3 ……S 前n ], S 后1 rear=[S 后2 , S 后3 , S 后n ……S 俯1 , and S 俯2 top=[S 俯3 , S 俯n ……S are obtained.

[0225] Based on the obtained similarity set of each view, the view with the highest similarity of the front view, the rear view and the top view is obtained, and the source thereof is: the front view of cabinet A + the rear view of cabinet B + the top view of cabinet C.

[0226] S3, intelligent rendering system:

[0227] Step one: based on the view combination with the highest matching degree calculated in the above step, the front view of cabinet A + the rear view of cabinet B + the top view of cabinet C are scaled in proportion according to the size of each view of the target cabinet. Each view can perfectly match the size of the target cabinet and can be perfectly spliced, and the source and scaling ratio of each view are marked, so that the user can easily see how the current cabinet is combined from different parts of the historical cabinet.

[0228] Step two: provide similarity contribution visualization analysis. This analysis report lists in detail the similarity contribution of each view to the historical cabinet, helping users to evaluate and optimize the matching scheme. For example, if the rear view of a historical cabinet has particularly high heat dissipation efficiency, and the current cabinet has a high requirement for heat dissipation efficiency, then the matching contribution of the rear view of the historical cabinet will be higher, and the user can consider prioritizing the use of this part of the feature in subsequent design, such as Figure 7 The process is shown.

[0229] Through the embodiments of the present application, the matching accuracy and efficiency can be improved. Specifically, by decoupling and independent vector space modeling of three-dimensional view (front / rear / top) features, complex cabinet features are disassembled into quantifiable and standardized submodules, reducing the computational complexity. Through a dynamic weight mechanism (such as temperature adaptive adjustment of heat dissipation weight), it is ensured that the matching result meets the actual scene requirements, improving the practicality of the matching scheme. In addition, the system flexibility can be enhanced to support cross-cabinet combination matching (such as cabinet A front view + cabinet B rear view + cabinet C top view), breaking through the single cabinet limit and maximizing the reuse of historical design schemes. Moreover, the cost of manual design can be reduced, and automatic feature extraction (contour coefficient, interface position, etc.) and intelligent matching algorithm replace manual experience, shortening the cabinet design cycle. At the same time, visual decision support can be met, similarity contribution analysis and combined view rendering are provided, and the matching basis is intuitively displayed to assist users in optimizing the design scheme (such as prioritizing the use of the rear view with high heat dissipation efficiency).

[0230] Through the above description of the embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be realized by means of software and the necessary general hardware platform, of course, it can also be realized by hardware, but in many cases the former is a better embodiment.

[0231] The embodiments of the present application also provide a determination device of an equipment, as shown in Figure 8As shown, the apparatus comprises:

[0232] The acquisition module 802 is configured to acquire a plurality of sets of device features of a target device, wherein the plurality of sets of device features comprises a first set of device features and a second set of device features, and the first set of device features and the second set of device features are different in category.

[0233] The determination module 804 is configured to determine a plurality of sets of candidate device features that satisfy a preset similarity condition based on the plurality of sets of device features, wherein the plurality of sets of device features and the plurality of sets of candidate device features are in one-to-one correspondence, and the plurality of sets of candidate device features comprises a first set of candidate device features that satisfies the preset similarity condition with the first set of device features and a second set of candidate device features that satisfies the preset similarity condition with the second set of device features.

[0234] The combination module 806 is configured to determine a candidate device based on the plurality of sets of candidate device features, and combine the candidate device to generate device information of a target candidate device, wherein the candidate device comprises a first candidate device and a second candidate device, the first candidate device corresponds to the first set of candidate device features, and the second candidate device corresponds to the second set of candidate device features.

[0235] In an example embodiment, the apparatus described above is configured to acquire the plurality of sets of device features of the target device by: acquiring device parameters of the target device; performing a feature extraction operation on the device parameters to determine a set of device features of the target device; and dividing the features in the set of device features according to a spatial dimension to obtain the plurality of sets of device features.

[0236] In an example embodiment, the apparatus described above is configured to divide the features in the set of device features according to the spatial dimension to obtain the plurality of sets of device features by: dividing features in the set of device features that belong to a front view of the target device into a first set of device features; dividing features in the set of device features that belong to a rear view of the target device into a second set of device features; and dividing features in the set of device features that belong to a top view of the target device into a third set of device features, wherein the third set of device features is different from the first set of device features and the second set of device features in category, the plurality of sets of candidate device features comprises a third set of candidate device features that satisfies the preset similarity condition with the third set of device features, and the candidate device comprises a third candidate device, which corresponds to the third set of candidate device features.

[0237] In an example embodiment, the apparatus described above is configured to divide the features in the set of device features that belong to the front view of the target device into the first set of device features by: acquiring first device parameters that belong to the front view of the target device, wherein the first device parameters comprise at least one of the following: contour parameters corresponding to the front view, slot information corresponding to the front view, and indicator light layout data corresponding to the front view; and integrating features in the set of device features that correspond to the first device parameters into the first set of device features.

[0238] In an example embodiment, the apparatus is configured to divide the features in the device feature set that belong to the rear view of the target device into a second group of device features by: obtaining second device parameters that belong to the rear view of the target device, wherein the second device parameters comprise at least one of the following: heat dissipation hole information corresponding to the rear view, interface information corresponding to the rear view, heat dissipation fan data corresponding to the rear view, and expansion slot information corresponding to the rear view; and integrating the features in the device feature set that correspond to the second device parameters into the second group of device features.

[0239] In an example embodiment, the apparatus is configured to divide the features in the device feature set that belong to the top view of the target device into a third group of device features by: obtaining third device parameters that belong to the top view of the target device, wherein the third device parameters comprise at least one of the following: size parameters corresponding to the top view, and load bearing data corresponding to the top view; and integrating the features in the device feature set that correspond to the third device parameters into the third group of device features.

[0240] In an example embodiment, the apparatus is configured to determine a plurality of groups of candidate device features that satisfy a preset similarity condition based on a plurality of groups of device features by: obtaining sample candidate device parameters of a predetermined sample candidate device; performing a feature extraction operation on the sample candidate device parameters to determine a sample candidate device feature set of the sample candidate device; dividing the features in the sample candidate device feature set according to a spatial dimension to obtain sample candidate device features corresponding to the sample candidate device; determining a similarity between a target group of device features and different target groups of sample candidate features, wherein the plurality of groups of device features comprise the target group of device features, the sample candidate device features comprise the target groups of sample candidate features, the different target groups of sample candidate features correspond to different sample candidate devices, and the target groups of sample candidate features and the target group of device features belong to a same category; and determining a target group of sample candidate features that satisfies a preset similarity condition as a group of candidate device features corresponding to the target group of device features in the plurality of groups of candidate device features.

[0241] In an example embodiment, the apparatus is configured to determine a similarity between a target group of device features and different target groups of sample candidate features by: determining a sub-similarity between each target device feature in the target group of device features and each target candidate feature in the target group of sample candidate features, wherein the target device feature and the target candidate feature belong to a same feature dimension; and merging the sub-similarities to determine the similarity between the target group of device features and the target group of sample candidate features.

[0242] In an example embodiment, the apparatus is configured to determine the sub-similarity between each target device feature in the target group device feature and each target candidate feature in the target group sample candidate feature by generating a target feature vector based on the target device feature and generating a candidate feature vector based on the target candidate feature, and determining the distance between the target feature vector and each candidate feature vector in a target vector space, and determining the sub-similarity based on the distance.

[0243] In an example embodiment, the apparatus is configured to determine the sample candidate device feature set of the sample candidate device by performing the feature extraction operation on the sample candidate device parameters, and calculating the mean and variance of the sample candidate device features belonging to different sample candidate device feature sets and belonging to the same feature dimension, and performing the standardization processing on each feature in each sample candidate device feature set based on the mean and variance to obtain the updated sample candidate device feature set.

[0244] In an example embodiment, the apparatus is further configured to add a preset weight coefficient to the feature corresponding to each feature dimension in the target group sample candidate feature, dynamically adjust the preset weight coefficient to a target weight coefficient based on a preset environmental parameter threshold, and optimize the feature corresponding to each feature dimension in the target group sample candidate feature according to the target weight coefficient.

[0245] In an example embodiment, the apparatus is configured to determine the candidate device based on the multiple groups of candidate device features, and combine the candidate devices to generate the device information of the target candidate device by determining the front view corresponding to the first candidate device, the rear view corresponding to the second candidate device, and the top view corresponding to the third candidate device based on the multiple groups of candidate device features, and scaling and splicing the front view, the rear view, and the top view according to the size information of the target device to obtain a combined view, wherein the combined view labels the first candidate device, the second candidate device, and the third candidate device, and the combined view is used to represent the device information of the target candidate device.

[0246] In an example embodiment, the apparatus is further configured to generate the front view contribution information, the rear view contribution information, and the top view contribution information based on the similarity contribution degrees of different feature dimensions in the multiple groups of candidate device features after scaling and splicing the front view, the rear view, and the top view according to the size information of the target device to obtain the combined view, and display the front view contribution information, the rear view contribution information, and the top view contribution information on the combined view.

[0247] It should be noted that the description of the features in the embodiments corresponding to the device determination apparatus described above can refer to the related description of the embodiments corresponding to the device determination method, which will not be repeated here.

[0248] Embodiments of the present application further provide an electronic device comprising a memory and a processor, the memory storing a computer program, and the processor being configured to execute the computer program to perform the steps in any of the above-mentioned device determination method embodiments.

[0249] Embodiments of the present application further provide a computer readable storage medium storing a computer program, wherein the computer program is configured to perform the steps in any of the above-mentioned device determination method embodiments when executed.

[0250] In an example embodiment, the above-mentioned computer readable storage medium can include, but is not limited to, a U disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media capable of storing computer programs.

[0251] Embodiments of the present application further provide a computer program product comprising a computer program, wherein the computer program is executed by a processor to implement the steps in any of the above-mentioned device determination method embodiments.

[0252] Embodiments of the present application further provide another computer program product comprising a non-volatile computer readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the steps in any of the above-mentioned device determination method embodiments.

[0253] The skilled person can further realize that the units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the general description of the examples has been described in terms of functional generalities. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0254] The determination method and device of the equipment provided by the present application are described in detail above. The principles and implementation manners of the present application are described by applying specific examples, and the above description of the examples is only applicable to helping understand the method of the present application and the core idea thereof. It should be pointed out that, for ordinary skilled persons in the technical field, some improvements and modifications can be made to the present application without departing from the principles of the present application, and these improvements and modifications also fall within the protection scope of the claims of the present application.

Claims

1. A method for determining a device, characterized in that, include: Obtain multiple sets of device features of the target device, wherein the multiple sets of device features include a first set of device features and a second set of device features, and the first set of device features and the second set of device features are of different categories; Based on the multiple sets of device features, multiple sets of candidate device features that satisfy preset similarity conditions are determined respectively. The multiple sets of device features correspond one-to-one with the multiple sets of candidate device features. The multiple sets of candidate device features include a first set of candidate device features that satisfy the preset similarity conditions with the first set of device features and a second set of candidate device features that satisfy the preset similarity conditions with the second set of device features. Based on the multiple sets of candidate device features, candidate devices are determined, and the candidate devices are combined to generate device information for a target candidate device. This includes: determining a front view corresponding to a first candidate device, a rear view corresponding to a second candidate device, and a top view corresponding to a third candidate device based on the multiple sets of candidate device features; scaling and stitching the front view, the rear view, and the top view according to the size information of the target device to obtain a combined view, wherein the combined view labels the first candidate device, the second candidate device, and the third candidate device, and the combined view is used to represent the device information of the target candidate device. The candidate devices include the first candidate device and the second candidate device, the first candidate device corresponds to the first set of candidate device features, and the second candidate device corresponds to the second set of candidate device features. The method further includes: acquiring device parameters of the target device; performing feature extraction on the device parameters to determine a set of device features of the target device; classifying features belonging to the front view of the target device in the set of device features into a first group of device features; classifying features belonging to the rear view of the target device in the set of device features into a second group of device features; and classifying features belonging to the top view of the target device in the set of device features into a third group of device features, wherein the third group of device features is different in category from both the first group of device features and the second group of device features, and the multiple sets of candidate device features include a third group of candidate device features that satisfy the preset similarity condition with the third group of device features, wherein the candidate device includes a third candidate device, and the third candidate device corresponds to the third group of candidate device features.

2. The method for determining the equipment according to claim 1, characterized in that, The step of dividing the features belonging to the front view of the target device in the device feature set into the first group of device features includes: Obtain first device parameters belonging to the front view of the target device, wherein the first device parameters include at least one of the following: contour parameters corresponding to the front view, slot information corresponding to the front view, and indicator layout data corresponding to the front view; The features corresponding to the first device parameters in the device feature set are integrated into the first group of device features.

3. The method for determining the equipment according to claim 1, characterized in that, The step of dividing the features belonging to the rear view of the target device in the device feature set into the second group of device features includes: Obtain second device parameters belonging to the rear view of the target device, wherein the second device parameters include at least one of the following: heat dissipation hole information corresponding to the rear view, interface information corresponding to the rear view, cooling fan data corresponding to the rear view, and external card slot information corresponding to the rear view. The features corresponding to the second device parameters in the device feature set are integrated into the second group of device features.

4. The method for determining the equipment according to claim 1, characterized in that, The step of dividing the top-view features of the target device in the device feature set into a third group of device features includes: Obtain a third device parameter belonging to the top view of the target device, wherein the third device parameter includes at least one of the following: the dimension parameter corresponding to the top view, and the load-bearing data corresponding to the top view; The features in the set of device features that correspond to the third device parameter are integrated into the third set of device features.

5. The method for determining the equipment according to claim 1, characterized in that, The step of determining multiple sets of candidate device features that satisfy preset similarity conditions based on the multiple sets of device features includes: Obtain the sample candidate device parameters of the predetermined sample candidate device; Perform feature extraction on the parameters of the candidate devices to determine the candidate device feature set; The features in the sample candidate device feature set are divided according to the spatial dimension to obtain the sample candidate device features corresponding to the sample candidate device; Determine the similarity between the device features of the target group and the candidate features of different target group samples, wherein the multiple device features include the device features of the target group, the candidate device features include the candidate features of the target group samples, different candidate features of the target group samples correspond to different candidate devices, and the candidate features of the target group samples and the device features of the target group belong to the same category; The candidate features of the target group samples whose similarity satisfies the preset similarity condition are determined as a set of candidate device features that correspond to the target group device features among the multiple sets of candidate device features.

6. The method for determining the device according to claim 5, characterized in that, Determining the similarity between the device features of the target group and the candidate features of different target group samples includes: Determine the sub-similarity between each target device feature in the target group device features and each target candidate feature in the target group sample candidate features, wherein the target device features and the target candidate features belong to the same feature dimension; The similarity between the target group device features and the target group sample candidate features is determined by merging the various sub-similarity values.

7. The method for determining the equipment according to claim 6, characterized in that, Determining the sub-similarity between each target device feature in the target group device features and each target candidate feature in the target group sample candidate features includes: A target feature vector is generated based on the target device features, and a candidate feature vector is generated based on the target candidate features; The distances between the target feature vector and the multiple candidate feature vectors in the target vector space are determined, and the sub-similarity is determined based on the distances.

8. The method for determining the equipment according to claim 5, characterized in that, The step of performing feature extraction on the parameters of the candidate devices to determine the feature set of the candidate devices includes: Perform feature extraction on the parameters of the candidate devices to determine the candidate device feature set; For sample candidate device features belonging to different sample candidate device feature sets but belonging to the same feature dimension, calculate the mean and variance; Based on the mean and variance, the features in each of the sample candidate device feature sets are standardized to obtain the updated sample candidate device feature sets.

9. The method for determining the device according to claim 5, characterized in that, The method further includes: Add preset weight coefficients to the features corresponding to each feature dimension in the candidate features of the target group samples; The preset weight coefficient is dynamically adjusted to the target weight coefficient based on a preset environmental parameter threshold. The features corresponding to each feature dimension in the candidate features of the target group samples are optimized according to the target weight coefficient.

10. The method for determining the device according to claim 1, characterized in that, After scaling and stitching the front view, the rear view, and the top view according to the size information of the target device to obtain a combined view, the method further includes: Based on the similarity contribution of different feature dimensions in the multiple sets of candidate device features, front view contribution information, rear view contribution information, and top view contribution information are generated. The front view contribution information, the rear view contribution information, and the top view contribution information are displayed on the combined view.

11. A device for determining equipment, characterized in that, include: The acquisition module is used to acquire multiple sets of device features of the target device, wherein the multiple sets of device features include a first set of device features and a second set of device features, and the first set of device features and the second set of device features are of different categories; The determining module is used to determine multiple sets of candidate device features that satisfy preset similarity conditions based on the multiple sets of device features, wherein the multiple sets of device features correspond one-to-one with the multiple sets of candidate device features, and the multiple sets of candidate device features include a first set of candidate device features that satisfy the preset similarity conditions with the first set of device features and a second set of candidate device features that satisfy the preset similarity conditions with the second set of device features; A combination module is used to determine candidate devices based on the multiple sets of candidate device features, and to combine the candidate devices to generate device information of a target candidate device. This includes: determining a front view corresponding to a first candidate device, a rear view corresponding to a second candidate device, and a top view corresponding to a third candidate device based on the multiple sets of candidate device features; scaling and stitching the front view, the rear view, and the top view according to the size information of the target device to obtain a combined view. The combined view labels the first candidate device, the second candidate device, and the third candidate device. The combined view represents the device information of the target candidate device. The candidate devices include the first candidate device and the second candidate device, where the first candidate device corresponds to the first set of candidate device features, and the second candidate device corresponds to the second set of candidate device features. The device is further configured to: acquire device parameters of the target device; perform feature extraction on the device parameters to determine a set of device features of the target device; classify features belonging to the front view of the target device in the set of device features into a first group of device features; classify features belonging to the rear view of the target device in the set of device features into a second group of device features; classify features belonging to the top view of the target device in the set of device features into a third group of device features, wherein the third group of device features is different in category from both the first group of device features and the second group of device features, the multiple groups of candidate device features include a third group of candidate device features that satisfy the preset similarity condition with the third group of device features, the candidate devices include a third candidate device, and the third candidate device corresponds to the third group of candidate device features.

12. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the method for determining the device as claimed in any one of claims 1 to 10.

Images