Data center machine room equipment indicating lamp identification method, terminal and storage medium

Abstract

The invention provides a data center machine room equipment indicating lamp identification method, a terminal and a storage medium, and relates to the technical field of machine room management, and the method comprises the steps: obtaining real-time image information containing all equipment indicating lamps in a machine room; inputting the real-time image information into a small target detection algorithm model, and obtaining the lighting state of each equipment indicating lamp; according to the lighting state of each equipment indicating lamp and a fault sensing information database, determining an equipment operation state so as to notify a machine room operator on duty of an equipment fault when the equipment operation state is a fault state; the small target detection algorithm model is obtained by training an improved RT-DETR algorithm model based on an enhanced historical image data set; the enhanced historical image data set comprises indicator light images of different sizes and indicator light images which are partially shielded by the shielding object. The small target identification efficiency can be improved, the indicating lamp can be accurately identified in a complex machine room environment with numerous devices and sheltering objects, and the operation state of the devices can be timely and effectively monitored.

Description

Technical Field

[0001] The present application relates to the technical field of computer room management, and in particular to a method, terminal and storage medium for identifying indicator lights of equipment in a computer room of a data center. Background Art

[0002] As the digitalization process continues to advance, the number and complexity of equipment in data center computer rooms continue to increase. To ensure the normal operation of computer room equipment, regular inspections have become an indispensable link. Traditional manual inspection methods mainly rely on personnel to observe the color and status of equipment indicators by naked eyes to judge the working status of equipment. However, when the number of equipment is large, manual inspections are difficult to achieve comprehensive and accurate monitoring. In addition, the proportion of indicator lights in abnormal conditions is usually very small, and manual inspections are inefficient and prone to errors and omissions.

[0003] In recent years, the rapid development of computer vision has promoted the progress of the automatic identification technology of the indicator light status of equipment in the computer room. However, in the process of implementing the examples of this application, it is found that the existing technology has at least the following problems:

[0004] In a complex equipment room environment with numerous devices and obstructions, the recognition efficiency of indicator lights is low. Summary of the invention

[0005] The embodiments of the present application provide a method, terminal and storage medium for identifying indicator lights of equipment in a data center computer room to solve the problem of low efficiency in identifying indicator lights in a complex computer room environment with numerous devices and obstructions.

[0006] In a first aspect, an embodiment of the present application provides a method for identifying indicator lights of equipment in a data center computer room, comprising:

[0007] Obtain real-time image information including indicator lights of all equipment in the equipment room;

[0008] Input the real-time image information into a small target detection algorithm model to obtain the lighting status of indicator lights of each device;

[0009] Determine the equipment operation status according to the lighting status of the indicator lights of each device and the fault perception information database, so as to notify the equipment room staff on duty of the equipment failure when the equipment operation status is a fault state;

[0010] Among them, the small target detection algorithm model is obtained by training the improved RT-DETR algorithm model based on the enhanced historical image data set; the enhanced historical image data set includes indicator light images of different sizes and indicator light images partially obscured by obstructions.

[0011] In a possible implementation, the improved RT-DETR algorithm model includes: an improved EfficientFormerV2, a hybrid encoder and a decoder;

[0012] The hybrid encoder is constructed based on HS-FAN; the decoder is trained based on a one-to-many label assignment method.

[0013] In a possible implementation, the improved EfficientFormerV2 includes: a Local module and a Global Local module;

[0014] The Local module is constructed based on depth-separable convolution and squeeze-excitation blocks, and uses GeLU activation function for feature activation;

[0015] The Global Local module integrates a lightweight attention mechanism to enhance the model’s understanding of image semantics and context information.

[0016] In one possible implementation, the hybrid encoder includes an AIFI-HiLo module, a transposed convolution module, a channel attention module, a GSConv module, and a VOVGSCSP module;

[0017] The AIFI-HiLo module divides the multi-head self-attention into two groups to decouple the high-frequency information and low-frequency information in the feature map of the real-time image information;

[0018] The transposed convolution module upsamples the high-order features to the same size as the low-order features;

[0019] The channel attention module weights and filters the underlying features, and fuses the weighted low-level features with high-level features to supplement the necessary context information for the shallow feature mapping;

[0020] The GSConv module and downsampling are used to merge low-order features and high-order features;

[0021] The VOVGSCSP module is used to suppress conflicting information in the feature fusion process to prevent small target features from being submerged by the conflicting information.

[0022] In one possible implementation, the decoder matches multiple positive sample object queries using a grouping allocation strategy based on a separate self-attention mechanism;

[0023] Based on the multi-scale deformable cross-attention mechanism, deformable attention is performed on the three sets of decoder object queries with the keys and values ​​provided by the hybrid encoder.

[0024] In a possible implementation, the lighting status includes one or more of the indicator light color, the indicator light flashing frequency, and the indicator light color change.

[0025] In a possible implementation, the device operating state includes: a normal operating state, a minor fault state, a medium fault state, and a major fault state.

[0026] In a possible implementation, the fault perception information database includes: the lighting status of the equipment indicator light, the fault cause and the fault level;

[0027] Determining the equipment operation status according to the lighting status of the indicator lights of each equipment and the fault perception information database includes:

[0028] Create multiple verification processes for the lighting status of each device indicator light;

[0029] Based on multiple verification processes, the lighting status of each of the device indicator lights is compared and verified with the fault perception information database to determine the cause of the device failure;

[0030] The fault state type is determined according to the fault cause and fault level of the equipment; wherein the fault state type classification includes a minor fault state, a medium fault state and a major fault state.

[0031] In a second aspect, an embodiment of the present application provides a device for identifying indicator lights of equipment in a data center computer room, comprising:

[0032] An acquisition module is used to obtain real-time image information including indicator lights of all equipment in the computer room;

[0033] A lighting status recognition module is used to input the real-time image information into a small target detection algorithm model to obtain the lighting status of each device indicator light;

[0034] The equipment operation status determination module is used to determine the equipment operation status according to the lighting status of the indicator lights of each equipment and the fault perception information database, so as to notify the equipment room staff on duty of the equipment failure when the equipment operation status is a fault state;

[0035] Among them, the small target detection algorithm model is obtained by training the improved RT-DETR algorithm model based on the enhanced historical image data set; the enhanced historical image data set includes indicator light images of different sizes and indicator light images partially obscured by obstructions.

[0036] In a third aspect, an embodiment of the present application provides a terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method described in the first aspect or any possible implementation of the first aspect are implemented.

[0037] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the method described in the first aspect or any possible implementation method of the first aspect.

[0038] The embodiment of the present application provides a data center computer room equipment indicator light recognition method, terminal and storage medium, which obtains real-time image information of all equipment indicator lights in the computer room and inputs this information into a small target detection algorithm model to identify the lighting status of each indicator light. By analyzing the lighting status of each indicator light and a preset fault perception information database, the operating status of the equipment is determined, and the computer room on-duty personnel are notified in a timely manner when a fault state is detected. The small target detection algorithm model is obtained by training an improved RT-DETR algorithm model based on an enhanced historical image data set. The enhanced historical image data set covers indicator light images of different sizes and partially obscured indicator light images. The RT-DETR algorithm adopts a lightweight network structure, which effectively reduces the computational complexity, thereby supporting a faster reasoning speed. By enhancing indicator light images of different sizes and partially obscured indicator light images, and training an improved RT-DETR algorithm model based on the enhanced data set, the robustness and accuracy of the model are improved. The embodiment of the present application significantly improves the efficiency of small target recognition, ensuring that in a complex computer room environment with dense equipment and obstructions, the indicator light status can be accurately identified and the operating status of the equipment can be monitored in a timely and effective manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0040] Figure 1 This is a flowchart of a method for identifying indicator lights of equipment in a data center computer room provided in an embodiment of the present application;

[0041] Figure 2 This is the overall architecture diagram of the improved RT-DETR algorithm model;

[0042] Figure 3It is a diagram of the hybrid encoder network architecture;

[0043] Figure 4 It is the GSConv structure diagram;

[0044] Figure 5 It is a structural schematic diagram of a device for identifying indicator lights of equipment in a data center computer room provided by an embodiment of the present application;

[0045] Figure 6 It is a schematic diagram of the structure of a terminal provided in one embodiment of the present application. DETAILED DESCRIPTION

[0046] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0047] The terms "first", "second", etc. in the specification and claims of the embodiments of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the embodiments of the present application described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions.

[0048] Unless otherwise specified, the term "plurality" means two or more. The character " / " indicates that the objects before and after are in an "or" relationship. For example, A / B means: A or B. The term "and / or" is a description of the association relationship between objects, indicating that there can be three relationships. For example, A and / or B means: A or B, or, A and B.

[0049] The words used in this application are only used to describe the embodiments and are not used to limit the claims. As used in the description of the embodiments and claims, unless the context clearly indicates, the singular forms of "a", "an" and "the" are intended to include the plural forms as well. Similarly, the term "and / or" as used in this application refers to any and all possible combinations of one or more associated listings. In addition, when used in this application, the term "comprise" and its variants "comprises" and / or including (comprising) refer to the existence of stated features, wholes, steps, operations, elements, and / or components, but do not exclude the existence or addition of one or more other features, wholes, steps, operations, elements, components and / or these groups. In the absence of more restrictions, the elements defined by the statement "including one..." do not exclude the existence of other identical elements in the process, method or device including the elements.

[0050] In this application, each embodiment may focus on the differences from other embodiments, and the same or similar parts between the embodiments may refer to each other. For the methods, products, etc. disclosed in the embodiments, if they correspond to the method part disclosed in the embodiments, then the relevant parts may refer to the description of the method part.

[0051] The existing technology for identifying the status of indicator lights of equipment in computer rooms has the following main problems and shortcomings:

[0052] First, there are many types of equipment in the computer room, and the data set of the indicator lights of the equipment in the computer room is small and accompanied by high noise. In addition, there may be inconsistencies in the data annotation process. These factors cause the model to face data quality problems during training and evaluation, which in turn affects the generalization ability and actual application effect of the target detection model.

[0053] Second, most target detection algorithms perform well when detecting large targets, but equipment indicator lights are usually small targets that occupy fewer image pixels and are often densely distributed. Coupled with the complex environment of the computer room, the interference of different light sources, and the indicator lights being easily blocked, it becomes more difficult to accurately identify and locate these small targets.

[0054] Third, in order to be applied in industrial inspection robots, the small target detection algorithm not only needs to run efficiently in the limited memory environment of embedded devices, but also needs to meet the strict requirements of real-time detection while ensuring low resource usage.

[0055] In recent years, the rapid development of computer vision has promoted the application of deep learning technology in real-time small target detection, providing an efficient solution to solve the problems faced by small target detection, such as small targets to be detected and easy to be blocked. In the digital computer room, the intelligent inspection robot uses a small target detection method based on deep learning to identify the equipment indicator lights, which significantly improves the accuracy and speed of recognition, while reducing labor costs and significantly improving the automation level of computer room inspection. The DETR series of algorithms simplifies the target detection process by eliminating hand-crafted components such as non-maximum suppression (NMS) and realizes true end-to-end target detection. This type of algorithm uses an encoder-decoder structure to encode image context information and make full use of the global information of the image for prediction. Therefore, compared with the YOLO series of algorithms, the DETR algorithm performs better in detection accuracy. The DETR algorithm is constantly innovating and upgrading. Among them, the RT-DETR (DETRs Beat YOLOs on Real-time Object Detection) algorithm has the following advantages over other DETR algorithms:

[0056] First, the parallel computing mechanism and end-to-end training method adopted by the RT-DETR algorithm allow for faster reasoning speed, thus enabling the RT-DETR algorithm to achieve real-time dynamic recognition of targets. Secondly, the algorithm designs an efficient hybrid encoder to effectively process multi-scale features by decoupling internal scale interactions and cross-scale fusion, and proposes IOU-aware query selection to further improve performance by providing higher quality initial object queries to the decoder. Finally, the detector supports flexible adjustment of reasoning speed by using different decoder layers without retraining, which facilitates practical applications in various real-time scenarios.

[0057] In order to make the purpose, technical solutions and advantages of the present application clearer, specific embodiments will be described below in conjunction with the accompanying drawings.

[0058] Figure 1 FIG. 1 is a flowchart of a method for identifying indicator lights of equipment in a data center computer room provided in an embodiment of the present application. Figure 1 As shown, the method comprises the following steps:

[0059] S101, obtaining real-time image information including indicator lights of all equipment in the computer room.

[0060] The execution subjects of each embodiment of the present application can be servers, processors, microprocessors and other devices with data processing functions. In the actual implementation process, the specific implementation method of the execution subject can be selected according to actual needs. This embodiment does not impose any special restrictions on this, as long as it is a device with data processing functions.

[0061] Inside the computer room, real-time image information is collected through cameras installed at specific locations. To ensure the comprehensiveness and integrity of image collection, these cameras will not only capture the images of indicator lights of all equipment in the computer room, but also image information of equipment control panels and other key parts. In this way, the monitoring system can obtain real-time visual data of all important equipment in the computer room, thereby ensuring comprehensive monitoring and management of the operating status of the computer room.

[0062] In addition, in this way, technicians can discover abnormal situations in time and quickly take corresponding maintenance measures to ensure the normal operation of the computer room equipment and the stability of the system.

[0063] S102, input the real-time image information into the small target detection algorithm model to obtain the lighting status of the indicator lights of each device; wherein the small target detection algorithm model is obtained by training the improved RT-DETR algorithm model based on the enhanced historical image data set; the enhanced historical image data set includes indicator light images of different sizes and indicator light images partially obscured by obstructions.

[0064] Among them, the enhanced historical image dataset not only covers indicator light images of different sizes, but also includes indicator light images that are partially obscured by occluders, in order to improve the robustness and accuracy of the model.

[0065] In order to construct this historical image dataset, first, a large amount of equipment indicator light image data was crawled from the Internet, and equipment images were taken in computer rooms in schools and other practical application scenarios. Images were collected through a variety of image acquisition methods to ensure the diversity and richness of sample data, thereby improving the generalization ability of the model. Then, after acquiring the image data, these images were enhanced, including adjusting the brightness, contrast, and resolution of the images to simulate different shooting conditions and environmental changes. At the same time, the enhanced image data was cleaned, for example: blurry, repeated, or low-quality images were removed to ensure the quality of the dataset and reduce image noise. In addition, the images were annotated in detail, with information such as the location, size, and whether the indicator light was on, so that the model could accurately identify indicator lights of different sizes and obscured. Finally, the enhanced dataset was divided into training, test, and validation sets to facilitate the evaluation and optimization of the model during the training process to ensure its performance in practical applications.

[0066] In this embodiment, real-time image information is input into the small target detection algorithm model, which can overcome the problems of a large number of equipment indicator light targets, small target sizes, and interference from obstructions such as wires, and effectively identify the lighting status of the equipment indicator lights.

[0067] S103, determining the equipment operation status according to the lighting status of the indicator lights of each device and the fault perception information database, so as to notify the equipment room attendant of the equipment failure when the equipment operation status is a fault state.

[0068] Among them, the system can automatically identify the operating status of the equipment based on the lighting status of the equipment indicator lights and the fault diagnosis standards. In addition, through the analysis of real-time image information, the system can quickly detect the change of the equipment status from normal operation to failure, and immediately implement corresponding countermeasures. The system will promptly notify the on-duty personnel in the computer room about the equipment failure, thereby preventing the failure to timely discover the equipment failure due to delays in manual inspections or inaccurate identification due to obstructions.

[0069] In this embodiment, the real-time image information of all the indicator lights of the equipment in the computer room is obtained, and this information is input into the small target detection algorithm model to identify the lighting status of each indicator light. By analyzing the lighting status of each indicator light and the preset fault perception information database, the operating status of the equipment is determined, and the computer room staff on duty is notified in time when the fault state is detected. The small target detection algorithm model is obtained by training the improved RT-DETR algorithm model based on the enhanced historical image data set. The enhanced historical image data set covers indicator light images of different sizes and partially obscured indicator light images. The RT-DETR algorithm adopts a lightweight network structure, which effectively reduces the computational complexity, thereby supporting a faster reasoning speed. By enhancing the indicator light images of different sizes and partially obscured indicator light images, and training the improved RT-DETR algorithm model based on the enhanced data set, the robustness and accuracy of the model are improved. The embodiment of the present application significantly improves the efficiency of small target recognition, ensuring that in a complex computer room environment with dense equipment and obstructions, the indicator light status can be accurately identified and the operating status of the equipment can be monitored in a timely and effective manner.

[0070] In a possible implementation, the improved RT-DETR algorithm model includes: an improved EfficientFormerV2, a hybrid encoder and a decoder;

[0071] Among them, the hybrid encoder is built based on HS-FAN; the decoder is trained based on the multi-label assignment method.

[0072] In this embodiment, the improved EfficientFormerV2 is selected as the backbone network for small target detection, the size of the input feature map is increased, and a lightweight attention mechanism is introduced to improve the model's ability and efficiency in extracting key features and enhance the model's performance when processing small target images. In addition, a hybrid encoder is constructed based on HS-FAN to solve the multi-scale problem in small target datasets. In actual implementation, small targets often have different sizes and shapes, which makes them appear differently in images. Through the hybrid encoder, we can better identify and process these small targets of different scales, thereby improving the robustness and accuracy of the model in practical applications. Finally, a decoder is designed and trained through a one-to-many label assignment method to provide more supervision information for the model, thereby enhancing the model's ability to detect small objects.

[0073] In a possible implementation, the improved EfficientFormerV2 includes: a Local module and a Global Local module;

[0074] Among them, the Local module is built based on depth-separable convolution and squeeze-excitation blocks, and uses the GeLU activation function for feature activation;

[0075] The Global Local module integrates a lightweight attention mechanism to enhance the model’s understanding of image semantics and context information.

[0076] Figure 2 This is the overall architecture diagram of the improved RT-DETR algorithm model. Figure 2 As shown in the figure, considering the focus on small target detection, the P5 layer originally used to detect larger targets is removed from the backbone network, and the output features of the P2, P3, and P4 layers are used in the HSFAN module. The purpose of this adjustment is to use high-resolution feature maps to improve the detection accuracy of small targets and retain rich target edge and texture information to reduce missed detection and false detection rates. At the same time, in order to reduce the number of model parameters and the amount of calculation, the self-attention module in the global local module is reduced to 1.

[0077] In one possible implementation, the hybrid encoder includes an AIFI-HiLo module, a transposed convolution module, a channel attention module, a GSConv module, and a VOVGSCSP module;

[0078] The AIFI-HiLo module divides the multi-head self-attention into two groups of high-frequency information and low-frequency information in the feature map of the decoupled real-time image information, which is used to promote the intra-scale interaction of high-order features;

[0079] Transposed convolution module, which upsamples high-order features to the same size as low-order features;

[0080] The channel attention module weights and filters the underlying features, and fuses the weighted underlying features with the high-level features to supplement the necessary contextual information for the shallow feature mapping;

[0081] Use GSConv module and downsampling to merge low-order features with high-order features;

[0082] The VOVGSCSP module is used to suppress the conflicting information in the feature fusion process to prevent the small target features from being submerged by the conflicting information.

[0083] In this embodiment, the hybrid encoder uses the HiLo-AiFi module to promote the scale interaction of high-order features and effectively capture the complex relationship between image features. A cross-scale feature fusion network is constructed through the channel attention mechanism and the GSConv module to enhance the model's ability to detect and recognize objects of different scales.

[0084] Figure 3 This is the network architecture diagram of the hybrid encoder (HS-FAN). The hybrid encoder network mainly consists of two parts: intra-scale feature interaction and cross-scale feature fusion.

[0085] Specifically, Figure 3 As shown in the figure, in the intra-scale feature interaction stage, the AIFI-HiLo module divides the multi-head self-attention into two groups to decouple the high-frequency and low-frequency information in the feature map. The high-frequency path (Hi-Fi) captures the fine-grained information in the image through local window self-attention, while the low-frequency path (Lo-Fi) obtains the low-frequency signal through average pooling and uses global attention to capture the low-frequency information of the image, thereby improving the computational efficiency and performance of the model.

[0086] Subsequently, high-frequency features are combined with low-frequency features to enhance the model's ability to extract key feature information. In the cross-scale feature fusion stage, transposed convolution (TPConv) is used to upsample high-order features to the same size as low-order features. A channel attention module (CA) is used to weight and filter the underlying features, and the weighted underlying features are fused with high-level features to supplement the necessary contextual information for shallow feature mapping.

[0087] Finally, if Figure 4 As shown in Figure 1, the GSConv module and downsampling are used to achieve the fusion of low-order and high-order features. The VOVGSCSP module is used to suppress the conflicting information in the feature fusion process to prevent the small target features from being submerged by the conflicting information.

[0088] In one possible implementation, the decoder matches multiple positive object queries using a group assignment strategy based on a separate self-attention mechanism;

[0089] Based on the multi-scale deformable cross-attention mechanism, deformable attention is performed on the three sets of decoder object queries with the keys and values ​​provided by the hybrid encoder.

[0090] like Figure 1 As shown. This component receives N queries as input and divides them into three groups: Q1, Q2, and Q3. In each group, a separate self-attention mechanism (SA) is performed to dynamically adjust the attention to different parts of the feature sequence to reduce false detection and missed detection of small objects. Subsequently, a multi-scale deformable cross-attention mechanism is used to perform deformable attention on the three sets of decoder object queries with the keys and values ​​provided by the hybrid encoder. Since it only focuses on a few key sampling points near the reference point, the deformable attention mechanism uses the query to calculate K offsets, which are then added to the reference point to determine the coordinates of the K "attention points", thereby reducing the amount of computation.

[0091]

[0092] In the formula, z q represents the original input feature, wherein the original feature corresponds to the image feature of the indicator light of the equipment in the computer room of this application, p q represents the normalized coordinates of the current reference point, represents the position of the device indicator light image feature in the feature map, x is the feature map index, W m represents multi-head attention, m is the attention number (the maximum value in this paper is 8), l is the dimension number of the feature pyramid, k is the number of the current sampling point, Δp mqk It represents the position offset of the kth sampling point in the mth attention head, that is, the relative distance of the coordinates of the point with the strongest relationship with the current reference point, which needs to be trained through the network. The complete execution process of the decoder can be expressed by the following formula:

[0093] Decoder(X,Q1)→Q1,Predictor(Q1)→Y1,

[0094] Decoder(X,Q2)→Q2,Predictor(Q2)→Y2,

[0095] Decoder(X,Q3)→Q3,Predictor(Q3)→Y3,

[0096] Through the above steps, we get three sets of prediction results: Y1, Y2, and Y3. Next, we use bipartite graph matching to match each set of prediction results with the corresponding ground truth objects (Y K,GT) and calculate the loss of each group during training. In the inference stage, the model only uses the first set of queries to predict the location and category of the small target of the device indicator light. This method significantly improves the inference speed of the model without changing the decoder structure.

[0097] In addition, in this embodiment, the bounding box regression loss function used by the original RT-DETR algorithm is GIOU, and its calculation formula is as follows:

[0098]

[0099] In the GIOU loss function, B is the area of ​​the predicted bounding box, B gt is the area of ​​the true bounding box, and C is the area that contains both B and B gt The GIOU loss function takes into account the relationship between the size and position of the bounding box, guiding the model to more accurately capture the position and size of the target. However, when using GIOU as the loss function for detecting small objects, the calculated value is more sensitive to changes in size and position due to the small size of the predicted object, which may lead to poor performance of the model in learning and predicting the position of small objects.

[0100] The EIOU loss function applies the aspect ratio factor to the width and height of the target bounding box and the true bounding box respectively to accelerate the convergence of the model. In addition, it also combines the concept of Focal Loss to reduce the contribution of many predicted bounding anchor boxes that have little overlap with the target anchor box, thereby solving the sample imbalance problem. The calculation is shown in the following formula

[0101]

[0102] L FocalEIOU =IOU γ L EIOU

[0103] In order to address the challenges of small object detection, such as small object size, easy clustering, and frequent occlusion, we use Inner-EIOU as the loss function. The calculation results are shown in the following formula:

[0104]

[0105] union=(w gt *h gt )*(ratio) 2 +(w*h)*(ratio) 2 -inter

[0106]

[0107] The width and height of the anchor box are denoted by w and h respectively, and the width and height of the anchor box (GT) are denoted by w gt and h gt Indicates that this ratio is used as an auxiliary factor to control the size of the auxiliary anchor box. The Inner-EIOU loss function effectively alleviates the imbalance between easy and difficult samples, making the loss calculation of high IoU targets more accurate, thereby improving the detection accuracy of small targets in complex background environments. The loss of the model during the training phase is the sum of the losses of each decoder group, as shown in the following formula:

[0108]

[0109] in, is the optimal arrangement of N indicators for the kth decoder.

[0110] In this embodiment, Inner-EIOU is used as the bounding box loss function to improve the convergence speed and positioning accuracy of the model. The location and type of indicator lights of equipment in the computer room can be accurately and quickly identified to detect faulty equipment early.

[0111] In a possible implementation, the lighting status includes one or more of the indicator light color, the indicator light flashing frequency, and the indicator light color change.

[0112] In different embodiments, different device types have different ways of indicating the lighting status of different device operating states, and therefore, different lighting statuses exist according to different device types.

[0113] In the actual implementation process, the flashing frequency of the indicator light and the color change of the indicator light are determined by identifying multiple frames of real-time image information.

[0114] In a possible implementation, the device operating status includes: a normal operating status, a minor fault status, a medium fault status, and a major fault status.

[0115] Among them, equipment failure information will be classified and managed, and the indicator light status of the equipment in the computer room will be monitored and maintained. Based on the classification management, corresponding graded processing measures will be taken according to the type and severity of the failure.

[0116] Optionally, a minor fault does not affect the normal operation of the equipment in the computer room and can be handled by on-duty personnel; an intermediate fault does not affect the normal operation of the equipment in the short term, but requires maintenance by maintenance personnel; a major fault requires immediate maintenance to avoid affecting the normal operation of the equipment.

[0117] In this embodiment, through hierarchical management of faults, the present invention can take corresponding countermeasures according to the severity of the fault, further improve processing efficiency, optimize human resource allocation, and significantly reduce the time and cost required for equipment maintenance.

[0118] In a possible implementation, the fault perception information database includes: the lighting status of the equipment indicator light, the fault cause and the fault level;

[0119] Determine the equipment operation status based on the lighting status of each equipment indicator and the fault perception information database, including:

[0120] Create multiple verification processes for the lighting status of each device indicator light;

[0121] Based on multiple verification processes, the lighting status of each device indicator light and the fault perception information database are compared and verified to determine the cause of the equipment failure;

[0122] The fault status type is determined according to the cause of the equipment failure and the fault level; wherein the fault status type classification includes a minor fault status, a medium fault status and a major fault status.

[0123] In this embodiment, the fault perception information database collects a wealth of fault-related information, including detailed data on the lighting status of the equipment indicator lights, fault causes, and fault levels. This information helps to more accurately understand the operating status and fault types of the equipment.

[0124] In addition, due to the large number of devices in the data center room, checking the status of each device one by one will significantly reduce the efficiency of fault identification and processing. Therefore, multiple verification processes are created for the lighting status of the indicator lights of each device. These processes work in parallel to monitor and analyze the operating status of the equipment in real time to ensure that faults can be discovered in time and corresponding measures can be taken.

[0125] Optionally, multiple verification processes are created for the lighting status of the indicator lights of each device, including: allocating the same verification process to devices of the same type according to the device type.

[0126] Based on these verification processes, the lighting status of the equipment indicator lights and the fault perception information database are carefully compared and verified to determine the cause of the equipment failure. For example, if the indicator light of a certain equipment lights up or the flashing frequency increases, the verification process will compare it with the fault perception information database to identify the specific cause of the failure, so as to implement targeted maintenance and treatment.

[0127] Due to the numerous causes of failure, a fault classification system is also implemented. In specific operations, some failures can be overcome by adjusting the equipment itself. For example, when the operating temperature of the equipment is too high, the equipment can reduce the temperature and resume normal operation by reducing the operating speed. At the same time, when multiple devices fail at the same time due to different reasons and the number of maintenance personnel is limited, the failures need to be handled in order of priority. Therefore, it is necessary to classify the different causes of failure.

[0128] After the cause of the equipment failure is determined, the fault status type will be determined based on the fault cause and fault level. The fault status type is divided into minor fault status, medium fault status and major fault status to more effectively manage and respond to different fault situations.

[0129] Minor fault status usually refers to some minor problems that occur during the operation of the equipment, which may not have much impact on the normal operation of the equipment; intermediate fault status means that there are some serious problems with the equipment, which need to be dealt with in time to avoid further damage; and major fault status means that the equipment has very serious problems and may need to be shut down immediately for maintenance to prevent greater losses and safety hazards. Through this classification, maintenance plans and response measures can be formulated more targeted to ensure the stable operation of the equipment and the safety of production.

[0130] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0131] The following is an embodiment of the device of the present application. For details not described in detail, please refer to the corresponding method embodiment described above.

[0132] Figure 5 is a structural schematic diagram of a device for identifying indicator lights of equipment in a data center computer room provided by an embodiment of the present application, such as Figure 5 For the sake of convenience, only the parts related to the embodiment of the present application are shown. Figure 5 As shown, the device comprises:

[0133] The acquisition module 501 is used to acquire real-time image information including indicator lights of all devices in the computer room;

[0134] The lighting status recognition module 502 is used to input the real-time image information into the small target detection algorithm model to obtain the lighting status of the indicator lights of each device;

[0135] The equipment operation status determination module 503 is used to determine the equipment operation status according to the lighting status of each equipment indicator light and the fault perception information database, so as to notify the equipment room staff on duty of the equipment failure when the equipment operation status is a fault state;

[0136] Among them, the small target detection algorithm model is obtained by training the improved RT-DETR algorithm model based on the enhanced historical image dataset; the enhanced historical image dataset includes indicator light images of different sizes and indicator light images partially obscured by obstructions.

[0137] In this embodiment, the real-time image information of all the indicator lights of the equipment in the computer room is obtained, and this information is input into the small target detection algorithm model to identify the lighting status of each indicator light. By analyzing the lighting status of each indicator light and the preset fault perception information database, the operating status of the equipment is determined, and the computer room staff on duty is notified in time when the fault state is detected. The small target detection algorithm model is obtained by training the improved RT-DETR algorithm model based on the enhanced historical image data set. The enhanced historical image data set covers indicator light images of different sizes and partially obscured indicator light images. The RT-DETR algorithm adopts a lightweight network structure, which effectively reduces the computational complexity, thereby supporting a faster reasoning speed. By enhancing the indicator light images of different sizes and partially obscured indicator light images, and training the improved RT-DETR algorithm model based on the enhanced data set, the robustness and accuracy of the model are improved. The embodiment of the present application significantly improves the efficiency of small target recognition, ensuring that in a complex computer room environment with dense equipment and obstructions, the indicator light status can be accurately identified and the operating status of the equipment can be monitored in a timely and effective manner.

[0138] Figure 6 is a schematic diagram of the structure of a terminal provided by an embodiment of the present application. Figure 6 As shown, the terminal 6 of this embodiment includes: a processor 60, a memory 61, and a computer program 62 stored in the memory 61 and executable on the processor 60. When the processor 60 executes the computer program 62, the steps in the above-mentioned data center room equipment indicator light recognition method embodiments are implemented, for example Figure 1 Alternatively, when the processor 60 executes the computer program 62, the functions of each module / unit in the above-mentioned device embodiments are realized, for example Figure 5 The functions of each module are shown.

[0139] Exemplarily, the computer program 62 may be divided into one or more modules / units, which are stored in the memory 61 and executed by the processor 60 to complete the present application. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program 62 in the terminal 6. For example, the computer program 62 may be divided into Figure 5 The modules shown.

[0140] The terminal 6 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The terminal 6 may include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art will appreciate that Figure 6It is only an example of terminal 6 and does not constitute a limitation on terminal 6. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal may also include input and output devices, network access devices, buses, etc.

[0141] The processor 60 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0142] The memory 61 may be an internal storage unit of the terminal 6, such as a hard disk or memory of the terminal 6. The memory 61 may also be an external storage device of the terminal 6, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal 6. Further, the memory 61 may also include both an internal storage unit of the terminal 6 and an external storage device. The memory 61 is used to store the computer program and other programs and data required by the terminal. The memory 61 may also be used to temporarily store data that has been output or is to be output.

[0143] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0144] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0145] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0146] In the embodiments provided in the present application, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0147] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0148] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0149] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned data center room equipment indicator light identification method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium.

[0150] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for identifying indicator lights of equipment in a data center computer room, characterized in that: include: Obtain real-time image information including indicator lights of all equipment in the equipment room; Input the real-time image information into a small target detection algorithm model to obtain the lighting status of indicator lights of each device; Determine the equipment operation status according to the lighting status of the indicator lights of each device and the fault perception information database, so as to notify the equipment room staff on duty of the equipment failure when the equipment operation status is a fault state; The small target detection algorithm model is obtained by training the improved RT-DETR algorithm model based on the enhanced historical image dataset; The enhanced historical image data set includes indicator light images of different sizes and indicator light images partially blocked by obstructions.

2. The data center equipment room equipment indicator light identification method according to claim 1, characterized in that: The improved RT-DETR algorithm model includes: an improved EfficientFormerV2, a hybrid encoder and a decoder; The hybrid encoder is constructed based on HS-FAN; the decoder is trained based on a one-to-many label assignment method.

3. The data center equipment room equipment indicator light recognition method according to claim 2, characterized in that: The improved EfficientFormerV2 includes: Local module and Global Local module; The Local module is constructed based on depth-separable convolution and squeeze-excitation blocks, and uses GeLU activation function for feature activation; The Global Local module integrates a lightweight attention mechanism to enhance the model’s understanding of image semantics and context information.

4. The data center equipment room equipment indicator light identification method according to claim 2, characterized in that: The hybrid encoder includes an AIFI-HiLo module, a transposed convolution module, a channel attention module, a GSConv module and a VOVGSCSP module; The AIFI-HiLo module divides the multi-head self-attention into two groups to decouple the high-frequency information and low-frequency information in the feature map of the real-time image information; The transposed convolution module upsamples the high-order features to the same size as the low-order features; The channel attention module weights and filters the low-level features, and fuses the weighted low-level features with the high-level features to supplement the necessary context information for the shallow feature mapping; The GSConv module and downsampling are used to merge low-order features and high-order features; The VOVGSCSP module is used to suppress conflicting information in the feature fusion process to prevent small target features from being submerged by the conflicting information.

5. The data center equipment room equipment indicator light identification method according to claim 2, characterized in that: The decoder is based on a separate self-attention mechanism and adopts a grouping allocation strategy to match multiple positive sample object queries; Based on the multi-scale deformable cross-attention mechanism, deformable attention is performed on the three sets of decoder object queries with the keys and values ​​provided by the hybrid encoder.

6. The data center equipment room equipment indicator light identification method according to claim 1, characterized in that: The lighting status includes one or more of the indicator light color, the indicator light flashing frequency and the indicator light color change.

7. The data center equipment room equipment indicator light identification method according to claim 6, characterized in that: The equipment operation status includes: normal operation status, minor fault status, intermediate fault status and major fault status.

8. The method for identifying indicator lights of equipment in a data center computer room according to claim 7, characterized in that: The fault perception information database includes: the lighting status of the equipment indicator light, the fault cause and the fault level; Determining the equipment operation status according to the lighting status of the indicator lights of each equipment and the fault perception information database includes: Create multiple verification processes for the lighting status of each device indicator light; Based on multiple verification processes, the lighting status of each of the device indicator lights is compared and verified with the fault perception information database to determine the cause of the device failure; The fault state type is determined according to the fault cause and fault level of the equipment; wherein the fault state type classification includes a minor fault state, a medium fault state and a major fault state.

9. A terminal comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images