Target identification method and system based on dynamic clustering of interest points

Abstract

The invention discloses a target recognition method and system based on interest point dynamic clustering, and the method comprises the steps: obtaining a screen width physical value and a screen width pixel value of to-be-recognized target display equipment in response to a target recognition request, calculating the screen resolution of the target display equipment, and carrying out the recognition of the target display equipment according to the screen resolution. Calculating dynamic clustering parameters; obtaining a to-be-clustered point-of-interest data set, and scaling point-of-interest data in the point-of-interest data set based on the dynamic clustering parameters to obtain clustering space coordinates; obtaining an interest point array constructed by the interest point data corresponding to the clustering space coordinates, and constructing a hash table according to the clustering space coordinates and the interest point array corresponding to the clustering space coordinates; and performing multi-dimensional sorting on each interest point array, generating a clustering center of the interest point data set according to a multi-dimensional sorting result, and performing target identification based on each generated clustering center. According to the method provided by the invention, the accuracy of target identification is improved.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and in particular, to an object recognition method and system based on dynamic clustering of points of interest. Background Art

[0002] With the rapid development of geographic information systems (GIS) and location-based services (LBS), the application scenarios of points of interest (POIs) data are becoming increasingly extensive, such as map navigation, virtual reality (VR), augmented reality (AR), etc. However, in the face of a vast amount of POI data, traditional display methods may lead to information overload, and users cannot quickly obtain useful information, thus affecting the immersive experience and interaction effect.

[0003] In the prior art, clustering algorithms are often used to reduce the amount of visualized information. However, existing clustering algorithms rely on a fixed grid size or fixed clustering parameters and cannot adjust the clustering density according to changes in the device screen resolution and display area, resulting in significant differences in the clustering results of the same physical area on different devices and a decrease in the accuracy of object recognition.

[0004] Therefore, it has become an urgent technical problem for those skilled in the art to solve how to improve the existing clustering method to enhance the accuracy of object recognition. Summary of the Invention

[0005] The present application provides an object recognition method and system based on dynamic clustering of points of interest to solve the technical problem of how to improve the existing clustering method and enhance the accuracy of object recognition.

[0006] To solve the above technical problem, an embodiment of the present application provides an object recognition method based on dynamic clustering of points of interest, including: In response to an object recognition request, obtain a dataset of points of interest to be clustered, and obtain the physical value and pixel value of the screen width of the target display device to be recognized; Based on the physical value of the screen width and the pixel value of the screen width, calculate the screen resolution of the target display device, and calculate dynamic clustering parameters according to the screen resolution; Scale each point of interest data in the points of interest dataset based on the dynamic clustering parameters to obtain the clustering space coordinates corresponding to the points of interest data, where each point of interest data includes coordinate information, level information, and unique ID information; Obtain an array of points of interest constructed from each of the points of interest data corresponding to the clustering space coordinates, and construct a hash table based on the clustering space coordinates and the array of points of interest corresponding to the clustering space coordinates; Perform multi-dimensional sorting on each of the said arrays of points of interest, generate the clustering centers of the said dataset of points of interest according to the results of the multi-dimensional sorting, and perform target recognition based on the generated respective clustering centers, wherein the multi-dimensional sorting is designed to perform sorting successively based on the said rank information, the said coordinate information, and the said unique ID information of each of the said point-of-interest data.

[0007] As one of the preferred solutions, the obtaining of the dataset of points of interest to be clustered includes: Obtain an initial dataset of points of interest from multiple data sources, where the data sources include but are not limited to map databases, user-defined lists of points of interest, and data interfaces of third-party geographic information service providers; Perform data cleaning and format unification processing on the said initial dataset of points of interest to obtain the dataset of points of interest to be clustered.

[0008] As one of the preferred solutions, the calculating of the dynamic clustering parameter according to the said screen resolution includes: Wherein, represents the dynamic clustering parameter, represents the power operation, represents rounding up, represents the screen resolution, represents the physical value of the screen width, represents the pixel value of the screen width.

[0009] As one of the preferred solutions, the calculation frequency of the said dynamic clustering parameter is adjusted based on the usage scenario of the said target display device, specifically including: When the screen resolution or the display area of the said target display device changes, recalculate the said dynamic clustering parameter.

[0010] As one of the preferred solutions, the scaling of each point-of-interest data in the said dataset of points of interest based on the said dynamic clustering parameter to obtain the clustering space coordinates corresponding to the said point-of-interest data includes: Extract the coordinate information of each of the said point-of-interest data, where the coordinate information includes the abscissa and the ordinate, expressed as ; Perform scaling processing on the abscissa and the ordinate respectively based on the said dynamic clustering parameter to obtain the clustering space coordinates, expressed as: Wherein, represents the abscissa of the clustering space coordinates, The ordinate representing the clustering space coordinates, represents the dynamic clustering parameter.

[0011] As one of the preferred solutions, after obtaining the clustering space coordinates corresponding to the point of interest data, it further includes: Performing a legality check on the clustering space coordinates, specifically including that if the clustering space coordinates exceed the preset threshold range, the dynamic clustering parameter is corrected.

[0012] As one of the preferred solutions, the constructing a hash table according to the clustering space coordinates and the array of points of interest corresponding to the clustering space coordinates includes: Constructing a hash table with the clustering space coordinates as the key and the array of points of interest corresponding to the clustering space coordinates as the value; Wherein, during the construction process of the hash table, the load factor of the hash table is monitored in real time. If the load factor exceeds the preset load threshold, an expansion operation is performed on the hash table; the load factor is the ratio of the number of elements already stored in the hash table to the capacity of the hash table.

[0013] As one of the preferred solutions, the performing multi-dimensional sorting on each array of points of interest and generating the clustering center of the point of interest data set according to the result of the multi-dimensional sorting includes: Sorting the level information of each point of interest data in the array of points of interest to obtain a first sorting result; wherein the level information reflects the importance or priority of each point of interest data; Obtaining each first point of interest data with the same sorting in the first sorting result, calculating the distance information between each first point of interest data and the clustering center, and updating the first sorting result based on the distance information to obtain a second sorting result; wherein the clustering center is expressed as , represents the clustering space coordinates, represents the dynamic clustering parameter; Obtaining each second point of interest data with the same sorting in the second sorting result, sorting according to the unique ID information of each second point of interest data to obtain a third sequence of points of interest; Obtaining the clustering center of the point of interest data set according to the third sequence of points of interest.

[0014] Another embodiment of the present application provides an object recognition system based on dynamic clustering of points of interest, including: A first acquisition module, configured to obtain a data set of points of interest to be clustered in response to an object recognition request, and obtain the physical value and pixel value of the screen width of the target display device to be recognized; A calculation module, configured to calculate the screen resolution of the target display device based on the physical value of the screen width and the pixel value of the screen width, and calculate dynamic clustering parameters according to the screen resolution; A second acquisition module, configured to scale each piece of point-of-interest data in the point-of-interest dataset based on the dynamic clustering parameters to obtain the clustering space coordinates corresponding to the point-of-interest data, where each piece of point-of-interest data includes coordinate information, level information, and unique ID information; A construction module, configured to obtain a point-of-interest array constructed by each piece of point-of-interest data corresponding to the clustering space coordinates, and construct a hash table according to the clustering space coordinates and the point-of-interest array corresponding to the clustering space coordinates; A sorting module, configured to perform multi-dimensional sorting on each point-of-interest array, perform multi-dimensional sorting on each point-of-interest array, generate the clustering centers of the point-of-interest dataset according to the results of the multi-dimensional sorting, and perform target recognition based on the generated clustering centers, where the multi-dimensional sorting is designed to perform sorting in sequence based on the level information, coordinate information, and unique ID information of each piece of point-of-interest data.

[0015] As one of the preferred solutions, the sorting module is further configured to: Sort the level information of each piece of point-of-interest data in the point-of-interest array to obtain a first sorting result; where the level information reflects the importance or priority of each piece of point-of-interest data; Obtain each first piece of point-of-interest data with the same sorting in the first sorting result, calculate the distance information between each first piece of point-of-interest data and the clustering center, and update the first sorting result based on the distance information to obtain a second sorting result; where the clustering center is expressed as , represents the clustering space coordinates, represents the dynamic clustering parameters; Obtain each second piece of point-of-interest data with the same sorting in the second sorting result, and sort according to the unique ID information of each second piece of point-of-interest data to obtain a third point-of-interest sequence; Obtain the clustering centers of the point-of-interest dataset according to the third point-of-interest sequence.

[0016] Compared with the prior art, the beneficial effects of the embodiments of the present application are at least one of the following: (1) By calculating the physical size and pixel width of the device screen in real time, the present application dynamically adjusts the clustering parameters to achieve the self-adaptation of the clustering density. Therefore, there is no need to develop separate adaptation logics for devices such as VR, AR, and mobile phones, and one set of algorithms covers the entire scenario, reducing a large amount of adaptation costs.

[0017] (2) The POIs obtained by clustering in this application are sorted in the following order: priority, Manhattan distance, and unique ID. In an emergency scenario (such as a fire), high-priority targets (such as fire hydrants and hospitals) can be identified first, thus shortening the response time.

[0018] (3) In this application, the POI data is scaled based on dynamic clustering parameters to obtain the clustering space coordinates, and a hash table is constructed. This spatial hash storage method can achieve fast aggregation and retrieval of POIs with a time complexity close to a constant, greatly improving the recognition efficiency and meeting application scenarios with strict requirements for fast target recognition, such as real-time navigation and emergency response. Brief Description of the Drawings

[0019] Figure 1 It is a schematic flowchart of a target recognition method based on dynamic clustering of POIs in one embodiment of this application; Figure 2 It is a schematic diagram of a target recognition system based on dynamic clustering of POIs in one embodiment of this application. Detailed Embodiment

[0020] Next, the technical solutions in the embodiments of this application will be clearly and completely described in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. The purpose of providing these embodiments is to make the disclosure of this application more thorough and comprehensive. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of this application.

[0021] In the description of this application, terms such as "first", "second", "third", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first", "second", "third", etc. may explicitly or implicitly include one or more of such features. In the description of this application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0022] In the description of the present application, it should be noted that, unless otherwise clearly defined and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, and can be the communication inside two components. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for the purpose of illustration, rather than indicating or implying that the device or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present application. The term "and / or" used herein includes any and all combinations of one or more of the related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific circumstances.

[0023] In the description of the present application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present application have the same meanings as those commonly understood by those skilled in the technical field to which the present application belongs. The terms used in the specification of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific circumstances.

[0024] An embodiment of the present application provides a target recognition method based on dynamic clustering of points of interest. Specifically, please refer to Figure 1 [[ID=e8]], Figure 1 which shows a schematic flow chart of the target recognition method based on dynamic clustering of points of interest in one of the embodiments of the present application, and it includes steps S1-S5: S1: In response to a target recognition request, obtain a data set of points of interest to be clustered, and obtain the physical value and pixel value of the screen width of the target display device to be recognized; When a target recognition request is received, an event listening mechanism is first triggered. This mechanism can be implemented through callback functions or message queues at the software level. For example, in a target recognition scenario based on a mobile application, when the user clicks a specific button on the map interface to initiate a target recognition operation, the event processor inside the application will capture this request and pass it to the module responsible for obtaining device screen information. In a more complex distributed system, the request may be sent from the client to the server side through the network. After the request processing service on the server side receives the request, it starts to execute the process of obtaining device screen information.

[0025] Preferably, in an embodiment of the present application, obtaining a data set of points of interest to be clustered includes: Obtain an initial point of interest dataset from multiple data sources, where the data sources include but are not limited to map databases, user-defined lists of points of interest, and data interfaces of third-party geographic information service providers; Perform data cleaning and format unification processing on the initial point of interest dataset to obtain a point of interest dataset to be clustered.

[0026] In order to comprehensively and accurately obtain point of interest data, it is necessary to collect information from multiple different data sources, and these data sources have their own characteristics and advantages.

[0027] The map database is one of the important sources of point of interest data. It usually contains a large amount of geographical information, such as commercial sites, public facilities, transportation stations, etc. This data has been professionally collected and sorted, and has high accuracy and integrity.

[0028] Users may define some specific points of interest according to their own needs and preferences during use. These points of interest have special significance for users, and may be places they often visit, locations they are concerned about, etc. The method of this application allows users to upload or input their own defined list of points of interest and incorporate it into the initial point of interest dataset to meet personalized needs.

[0029] Third-party geographic information service providers have professional geographic data collection and processing capabilities and can provide diverse point of interest data. For example, some companies that specialize in providing geographic information services collect and sort point of interest data through satellite remote sensing, field surveys, etc., and provide it to other systems for use through data interfaces. By calling these interfaces, different types and sources of point of interest data can be obtained to enrich the initial dataset.

[0030] The initial point of interest dataset obtained from multiple data sources may have problems such as uneven data quality and inconsistent formats. Therefore, data cleaning and format unification processing are required. The purpose of data cleaning is to remove noise, duplicate data, and incorrect data in the data and improve the data quality. Specific operations include checking the integrity of the data and removing points of interest that lack key information (such as coordinate information, name, etc.); checking the accuracy of the data and correcting incorrect coordinate, name, etc. information; removing duplicate points of interest to avoid data redundancy. For example, by comparing the coordinates and names of points of interest, points of interest with the same coordinates and similar names are regarded as duplicate data and removed.

[0031] The POI data provided by different data sources may have different formats and structures. To facilitate subsequent processing and analysis, it is necessary to uniformly process the data formats. This includes unifying the field names, data types, and encoding methods of the POI data, etc. For example, uniformly converting the coordinate information in different data sources into the coordinate system format of the pseudo-Mercator projection, and uniformly encoding the names of POIs using UTF-8, etc.

[0032] S2: Calculate the screen resolution of the target display device based on the physical value of the screen width and the pixel value of the screen width, and calculate the dynamic clustering parameter according to the screen resolution; Preferably, in an embodiment of the present application, calculating the dynamic clustering parameter according to the screen resolution includes: wherein, represents the dynamic clustering parameter, represents the power operation, represents rounding up, represents the screen resolution, represents the physical value of the screen width, represents the pixel value of the screen width.

[0033] The principle of the screen resolution calculation formula is to associate the physical size of the screen with the number of pixels to obtain the physical size corresponding to a single pixel, that is, the screen resolution. This value reflects the fineness of the screen and is crucial for subsequent POI clustering and target recognition. For example, on a high-resolution device, the physical area represented by each pixel is smaller, and more refined clustering is required to accurately present POIs; while on a low-resolution device, the physical area represented by pixels is larger, and the clustering can be relatively coarser.

[0034] When performing the calculation, it is necessary to ensure the accuracy and consistency of the obtained physical value of the screen width and the pixel value of the screen width. If the obtained values are abnormal (such as the physical value of the screen width is negative or the pixel value of the screen width is 0), error handling is required. Error information can be recorded through logs, and the user can be prompted that there is an abnormality in obtaining the device screen information, asking the user to check the device settings or perform the acquisition operation again. At the same time, to improve the calculation accuracy, a high-precision data type (such as a floating-point number) can be used for calculating the screen resolution to avoid precision loss caused by data type conversion.

[0035] After obtaining the screen resolution, the dynamic clustering parameter is further calculated. The calculation formula of the dynamic clustering parameter obtains a dynamic clustering parameter that can reflect the characteristics of the device screen through the comprehensive operation of the screen resolution and the pixel value of the screen width. Through such a calculation method, the clustering granularity can be dynamically adjusted according to the screen characteristics of different devices. For example, for a device with a high resolution, the calculated value of the dynamic clustering parameter will be relatively small, which means that the clustering will be more refined; while for a device with a low resolution, the value of the dynamic clustering parameter will be larger and the clustering will be coarser.

[0036] Preferably, in an embodiment of the present application, the calculation frequency of the dynamic clustering parameter is adjusted based on the usage scenario of the target display device, specifically including: When the screen resolution or display area of the target display device changes, recalculate the dynamic clustering parameter.

[0037] In order to enable the dynamic clustering parameter to better adapt to the changes of the device screen, it is necessary to adjust its calculation frequency according to the usage scenario of the target display device. When the screen resolution or display area of the target display device changes, immediately recalculate the dynamic clustering parameter. This can be achieved by listening to the screen change event of the device. For example, in a mobile device, a broadcast receiver for screen resolution change can be registered; in a desktop computer system, the screen change notification mechanism provided by the operating system can be used to obtain the screen change event. After detecting the screen change event, the system will re-obtain the physical value of the screen width and the pixel value of the screen width, and recalculate the screen resolution and the dynamic clustering parameter according to the above steps to ensure that the clustering parameter can reflect the latest state of the screen in time.

[0038] S3: Scale each piece of point-of-interest data in the point-of-interest dataset based on the dynamic clustering parameter to obtain the clustering space coordinates corresponding to the point-of-interest data, where each piece of point-of-interest data includes coordinate information, level information, and unique ID information; Preferably, in an embodiment of the present application, scaling each piece of point-of-interest data in the point-of-interest dataset based on the dynamic clustering parameter to obtain the clustering space coordinates corresponding to the point-of-interest data includes: Extract the coordinate information of each piece of point-of-interest data, where the coordinate information includes the abscissa and the ordinate, expressed as ; Perform scaling processing on the abscissa and the ordinate respectively based on the dynamic clustering parameter to obtain the clustering space coordinates, expressed as: where, represents the abscissa of the clustering space coordinates, represents the ordinate of the clustering space coordinates, Represents dynamic clustering parameters.

[0039] After obtaining the dataset of points of interest to be clustered, it is necessary to scale each point of interest data based on the previously calculated dynamic clustering parameters to obtain the corresponding clustering space coordinates.

[0040] Each point of interest data includes coordinate information, level information, and unique ID information. The coordinate information is usually represented in the form of a coordinate system of pseudo-Mercator projection, which is used to determine the position of the point of interest in the geographical space; the level information represents the importance or priority of the point of interest. For example, hospitals, transportation hubs, etc. may have a higher level; the unique ID information is used to uniquely identify each point of interest, facilitating subsequent queries and management.

[0041] The purpose of scaling is to adjust the position of the point of interest in the clustering space according to the dynamic clustering parameters, making the clustering more in line with the display characteristics of the device screen. The specific scaling operation can be achieved by multiplying the coordinate information of the point of interest by the dynamic clustering parameters. In this way, the point of interest is mapped from the original geographical space to the clustering space, making the clustering effect of the point of interest more reasonable under different screen resolutions and display areas.

[0042] The obtained clustering space coordinates are used for subsequent clustering operations. The system can divide the points of interest into different clusters according to these coordinates. During the clustering process, the clustering space coordinates can more accurately reflect the relative position relationship between the points of interest, thereby improving the accuracy and efficiency of clustering. For example, in a density-based clustering algorithm, the clustering space coordinates can be used to calculate the distance between points of interest to determine whether they belong to the same cluster.

[0043] In practical applications, due to factors such as the diversity of device screen characteristics, errors in the data calculation process, and the particularity of the distribution of points of interest, some clustering space coordinates may exceed the reasonable range. If these abnormal clustering space coordinates are used for subsequent clustering and target recognition operations, problems such as chaotic clustering results and inaccurate target recognition may occur, seriously affecting the performance of the entire system and the user experience. Therefore, it is very necessary to check the legality of the clustering space coordinates.

[0044] Preferably, in an embodiment of the present application, after obtaining the clustering space coordinates corresponding to the point of interest data, it further includes: Checking the legality of the clustering space coordinates, specifically including if the clustering space coordinates exceed the preset threshold range, then correcting the dynamic clustering parameters.

[0045] When the clustering space coordinates corresponding to the point of interest data are obtained, the system will perform a legality check according to the preset threshold range. First, check whether the abscissa of the clustering space is within the preset width-direction threshold range. If it exceeds this range, it indicates that the clustering space coordinates of this point of interest may be abnormal in the width direction. Similarly, perform a height-direction check on the ordinate of the clustering space. If it exceeds the range, it also indicates that the clustering space coordinates of this point of interest are abnormal in the height direction.

[0046] S4: Obtain the point-of-interest array constructed from the data of each point of interest corresponding to the clustering space coordinates, and construct a hash table based on the clustering space coordinates and the point-of-interest array corresponding to the clustering space coordinates; Preferably, in an embodiment of the present application, constructing a hash table based on the clustering space coordinates and the point-of-interest array corresponding to the clustering space coordinates includes: Construct a hash table with the clustering space coordinates as the key and the point-of-interest array corresponding to the clustering space coordinates as the value; Among them, during the construction process of the hash table, the load factor of the hash table is monitored in real time. If the load factor exceeds the preset load threshold, an expansion operation is performed on the hash table; the load factor is the ratio of the number of elements already stored in the hash table to the capacity of the hash table.

[0047] In the target recognition method based on dynamic point-of-interest clustering, constructing a hash table is for efficiently storing and retrieving point-of-interest data. First, group the point-of-interest data with the same clustering space coordinates into an array. For example, after dynamic parameter scaling, multiple points of interest may be mapped to the same clustering coordinate (cx, cy), and the coordinate, level, ID, etc. information of these points of interest will be organized into an array to form the candidate data set of this clustering unit. When creating a hash table, a preset initial capacity (such as 16 buckets) is set, and each bucket is used to store key-value pairs. The key is the clustering space coordinate (cx, cy), and the value is the corresponding point-of-interest array. Initially, the hash table is empty and waiting for data to be filled.

[0048] Traverse all clustering coordinates and their corresponding point-of-interest arrays, and map each key-value pair to a bucket through a hash function. For example, use the hash algorithm of (cx * 31 + cy) % capacity to convert the coordinates into bucket indices to ensure uniform distribution of different coordinates.

[0049] Load factor = number of elements already stored / total capacity of the hash table. For example, when a hash table with a capacity of 100 stores 75 elements, the load factor is 0.75. This indicator reflects the storage density of the hash table. After each new key-value pair is inserted, the system automatically calculates the current load factor. When the load factor exceeds the preset threshold (usually 0.75), an expansion operation is triggered to prevent performance degradation caused by increased hash collisions. During expansion, the capacity of the hash table is doubled (such as from 16 to 32), and a new array of buckets is created. The hash function of the new capacity will recalculate the hash values of all existing keys to ensure uniform distribution of data.

[0050] In one embodiment of the present application, the steps of data expansion include: traversing each bucket of the original hash table; for each key-value pair, recalculating the hash value using the new capacity to determine the new bucket position; inserting the key-value pair into the new bucket and handling possible hash conflicts (through linked lists or linear probing); updating the hash table capacity and element count to complete the migration.

[0051] S5: Perform multi-dimensional sorting on each point-of-interest array, generate the cluster centers of the point-of-interest dataset according to the results of the multi-dimensional sorting, and perform target recognition based on the generated cluster centers, where the multi-dimensional sorting is designed to sort sequentially based on the rank information, coordinate information, and unique ID information of each point-of-interest data.

[0052] Preferably, in one embodiment of the present application, performing multi-dimensional sorting on each point-of-interest array and generating the cluster centers of the point-of-interest dataset according to the results of the multi-dimensional sorting includes: Sorting the rank information of each point-of-interest data in the point-of-interest array to obtain the first sorting result; where the rank information reflects the importance or priority of each point-of-interest data; Obtaining each first point-of-interest data with the same sorting in the first sorting result, calculating the distance information between each first point-of-interest data and the cluster center, and updating the first sorting result based on the distance information to obtain the second sorting result; where the cluster center is expressed as , represents the cluster space coordinates, represents the dynamic clustering parameter; Obtaining each second point-of-interest data with the same sorting in the second sorting result, sorting according to the unique ID information of each second point-of-interest data to obtain the third point-of-interest sequence; Obtaining the cluster centers of the point-of-interest dataset according to the third point-of-interest sequence.

[0053] The rank information in the point-of-interest data reflects the importance or priority of the point of interest. For example, in the point-of-interest data of a city map, the ranks of some important public facilities may be higher, while the ranks of ordinary stores may be lower.

[0054] To sort the array of points of interest according to the level information, the present application uses the quicksort algorithm. Quicksort is an efficient sorting algorithm with an average time complexity of O(nlogn). Its basic idea is to select a pivot element and divide the array into two parts such that all elements in the left part are less than or equal to the pivot element, and all elements in the right part are greater than the pivot element, and then recursively sort the left and right parts. After quicksort, the present application obtains the first sorting result arranged in descending (or ascending, depending on specific requirements) order according to the level information. In this way, in subsequent processing, points of interest with higher levels will be considered first.

[0055] In the first sorting result, there may be multiple points of interest with the same level information. For these points of interest with the same level (i.e., the first point-of-interest data), further processing is required. First, it is necessary to clarify the representation of the clustering center, which is represented as , and then calculate the distance information between each first point-of-interest data and the clustering center. The distance can be calculated using the common Euclidean distance formula. According to the calculated distance information, the part with the same level in the first sorting result is re-sorted so that the points of interest closer to the clustering center are arranged in the front, thereby obtaining the second sorting result. The purpose of doing this is to give priority to the points of interest closer to the clustering center when the levels are the same because they may have a stronger correlation with the clustering center.

[0056] In the second sorting result, there may still be points of interest with the same distance from the clustering center and the same level (i.e., the second point-of-interest data). To further distinguish these points of interest, we use the unique ID information for sorting. The unique ID is the unique identifier of each point of interest. By sorting it, it can be ensured that the final sorting result is uniquely determined. Sort these second point-of-interest data in ascending (or descending) order according to the unique ID to obtain the third point-of-interest sequence.

[0057] After obtaining the third point-of-interest sequence, the clustering center of the point-of-interest dataset can be generated according to this sequence.

[0058] In an embodiment of the present application, select the first few points of interest in the sorted point-of-interest sequence (the specific number can be determined according to actual requirements), and then calculate the average value of the coordinates of these points of interest as the coordinates of the clustering center.

[0059] After generating each clustering center, target recognition can be performed based on these clustering centers. There are many specific methods for target recognition, such as region division, feature matching, etc.

[0060] Through the above process of multi-dimensional sorting and cluster center generation, the points of interest can be clustered and target recognition can be performed more accurately, improving the accuracy and efficiency of recognition.

[0061] Another embodiment of the present application provides a target recognition system based on dynamic clustering of points of interest. Specifically, please refer to Figure 2 , Figure 2 which shows a schematic diagram of a target recognition system based on dynamic clustering of points of interest in one embodiment of the present application, and it includes: A first acquisition module 11, configured to, in response to a target recognition request, acquire a dataset of points of interest to be clustered, and acquire the physical value and pixel value of the screen width of the target display device to be recognized; A calculation module 12, configured to calculate the screen resolution of the target display device based on the physical value and pixel value of the screen width, and calculate dynamic clustering parameters according to the screen resolution; A second acquisition module 13, configured to scale each point-of-interest data in the point-of-interest dataset based on the dynamic clustering parameters to obtain the clustering space coordinates corresponding to the point-of-interest data, where each point-of-interest data includes coordinate information, level information, and unique ID information; A construction module 14, configured to obtain a point-of-interest array constructed by each point-of-interest data corresponding to the clustering space coordinates, and construct a hash according to the clustering space coordinates and the point-of-interest array corresponding to the clustering space coordinates; A sorting module 15, configured to perform multi-dimensional sorting on each point-of-interest array, generate a cluster center of the point-of-interest dataset according to the result of the multi-dimensional sorting, and perform target recognition based on the generated cluster centers, where the multi-dimensional sorting is designed to perform sorting in sequence based on the level information, coordinate information, and unique ID information of each point-of-interest data.

[0062] Preferably, in an embodiment of the present application, the sorting module is further configured to: Sort the level information of each point-of-interest data in the point-of-interest array to obtain a first sorting result; where the level information reflects the importance or priority of each point-of-interest data; Obtain each first point-of-interest data with the same sorting in the first sorting result, calculate the distance information between each first point-of-interest data and the cluster center, and update the first sorting result based on the distance information to obtain a second sorting result; where the cluster center is expressed as , represents the clustering space coordinates, represents the dynamic clustering parameters; Obtain each second point of interest data with the same sorting in the second sorting result, and sort according to the unique ID information of each second point of interest data to obtain a third sequence of points of interest; Obtain the clustering center of the point of interest data set according to the third sequence of points of interest.

[0063] Compared with the prior art, the beneficial effects of the embodiments of the present application are at least one of the following: (1) By calculating the physical size and pixel width of the device screen in real time, the present application dynamically adjusts the clustering parameters to achieve self-adaptation of the clustering density. Therefore, there is no need to develop separate adaptation logics for devices such as VR, AR, and mobile phones, and one set of algorithms covers the entire scenario, reducing a large amount of adaptation costs.

[0064] (2) The present application sorts the clustered POIs in sequence according to the following rules: priority, Manhattan distance, and unique ID. Therefore, in an emergency scenario (such as a fire), high-priority targets (such as fire hydrants and hospitals) will be recognized first, shortening the response time.

[0065] (3) By scaling the point of interest data based on the dynamic clustering parameters, the present application obtains the clustering spatial coordinates and constructs a hash table. This spatial hash storage method can achieve fast aggregation and retrieval of points of interest with a time complexity close to a constant, greatly improving the recognition efficiency and meeting application scenarios with strict requirements for fast target recognition such as real-time navigation and emergency response.

[0066] The above embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several deformations and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A target recognition method based on dynamic clustering of points of interest, characterized in that, Including: In response to a target recognition request, obtain a dataset of points of interest to be clustered, and obtain the physical value and pixel value of the screen width of the target display device to be recognized; Based on the physical value of the screen width and the pixel value of the screen width, calculate the screen resolution of the target display device, and calculate dynamic clustering parameters according to the screen resolution; Scale each point-of-interest data in the point-of-interest dataset based on the dynamic clustering parameters to obtain the clustering space coordinates corresponding to the point-of-interest data, where each point-of-interest data includes coordinate information, level information, and unique ID information; Obtain a point-of-interest array constructed from each of the point-of-interest data corresponding to the clustering space coordinates, and construct a hash table based on the clustering space coordinates and the point-of-interest array corresponding to the clustering space coordinates; Perform multidimensional sorting on each point-of-interest array, generate the clustering centers of the point-of-interest dataset according to the results of the multidimensional sorting, and perform target recognition based on the generated clustering centers, where the multidimensional sorting is designed to perform sorting based on the level information, coordinate information, and unique ID information of each point-of-interest data in sequence.

2. The target recognition method based on dynamic clustering of points of interest according to claim 1, characterized in that, The obtaining of the dataset of points of interest to be clustered includes: Obtain an initial dataset of points of interest from multiple data sources, where the data sources include but are not limited to map databases, user-defined lists of points of interest, and data interfaces of third-party geographic information service providers; Perform data cleaning and format unification processing on the initial dataset of points of interest to obtain a dataset of points of interest to be clustered.

3. The target recognition method based on dynamic clustering of points of interest according to claim 1, characterized in that The calculating of the dynamic clustering parameters according to the screen resolution includes: Among them, represents the dynamic clustering parameter, represents the power operation, represents rounding up, represents the screen resolution, represents the physical value of the screen width, represents the pixel value of the screen width.

4. The target recognition method based on dynamic clustering of points of interest according to claim 3, characterized in that, The calculation frequency of the dynamic clustering parameters is adjusted based on the usage scenario of the target display device, specifically including: When the screen resolution or display area of the target display device changes, recalculate the dynamic clustering parameters.

5. The target recognition method based on dynamic clustering of points of interest according to claim 1, characterized in that, The scaling of each point-of-interest data in the point-of-interest dataset based on the dynamic clustering parameters to obtain the clustering space coordinates corresponding to the point-of-interest data includes: Extract the coordinate information of each of the said point-of-interest data, where the coordinate information includes an abscissa and an ordinate, expressed as ; Perform scaling processing on the abscissa and ordinate respectively based on the dynamic clustering parameters to obtain the clustering space coordinates, expressed as: Among them, represents the abscissa of the clustering space coordinates, represents the ordinate of the clustering space coordinates, represents the dynamic clustering parameter.

6. The method for target recognition based on dynamic clustering of points of interest according to claim 5, characterized in that After obtaining the clustering space coordinates corresponding to the point-of-interest data, it further includes: Perform a legality check on the clustering space coordinates, specifically including if the clustering space coordinates exceed a preset threshold range, then correct the dynamic clustering parameters.

7. The target recognition method based on dynamic clustering of points of interest according to claim 1, characterized in that The constructing of the hash table based on the clustering space coordinates and the point-of-interest array corresponding to the clustering space coordinates includes: Construct a hash table with the clustering space coordinates as the key and the point-of-interest array corresponding to the clustering space coordinates as the value; Wherein, during the construction of the hash table, the load factor of the hash table is monitored in real time. If the load factor exceeds a preset load threshold, perform an expansion operation on the hash table; the load factor is the ratio of the number of elements already stored in the hash table to the capacity of the hash table.

8. The target recognition method based on dynamic clustering of points of interest according to claim 1, characterized in that Performing multi-dimensional sorting on each of the POI arrays and generating cluster centers of the POI data set according to the results of the multi-dimensional sorting includes: Sorting the rank information of each POI data in the POI array to obtain a first sorting result; wherein the rank information reflects the importance or priority of each POI data; Obtain each first interest point data with the same sorting in the first sorting result, calculate the distance information between each first interest point data and the clustering center, and update the first sorting result based on the distance information to obtain a second sorting result; where the clustering center is expressed as , represents the clustering space coordinates, represents the dynamic clustering parameter; Obtaining each second POI data with the same sorting in the second sorting result, and sorting according to the unique ID information of each second POI data to obtain a third POI sequence; Obtaining the cluster centers of the POI data set according to the third POI sequence.

9. An object recognition system based on dynamic clustering of points of interest, characterized in that, Including: A first acquisition module, configured to, in response to a target recognition request, acquire a POI data set to be clustered, and acquire the physical value and pixel value of the screen width of the target display device to be recognized; A calculation module, configured to calculate the screen resolution of the target display device based on the physical value of the screen width and the pixel value of the screen width, and calculate a dynamic clustering parameter according to the screen resolution; A second acquisition module, configured to scale each POI data in the POI data set based on the dynamic clustering parameter to obtain the cluster space coordinates corresponding to the POI data, wherein each POI data includes coordinate information, rank information, and unique ID information; A construction module, configured to acquire a POI array constructed by each POI data corresponding to the cluster space coordinates, and construct a hash table according to the cluster space coordinates and the POI array corresponding to the cluster space coordinates; A sorting module, configured to perform multi-dimensional sorting on each POI array, generate cluster centers of the POI data set according to the results of the multi-dimensional sorting, and perform target recognition based on the generated cluster centers, wherein the multi-dimensional sorting is designed to perform sorting in sequence based on the rank information, coordinate information, and unique ID information of each POI data.

10. The target recognition system based on dynamic clustering of points of interest according to claim 9, characterized in that, The sorting module is further configured to: Sort the rank information of each POI data in the POI array to obtain a first sorting result; wherein the rank information reflects the importance or priority of each POI data; Obtain each first interest point data with the same sorting in the first sorting result, calculate the distance information between each first interest point data and the clustering center, and update the first sorting result based on the distance information to obtain a second sorting result; wherein the clustering center is expressed as , represents the clustering space coordinates, represents the dynamic clustering parameter; Obtain each second POI data with the same sorting in the second sorting result, and sort according to the unique ID information of each second POI data to obtain a third POI sequence; Obtain the cluster centers of the POI data set according to the third POI sequence.

Images