Concrete sleeper crack identification method, system and equipment and storage medium

Abstract

The invention discloses a concrete sleeper crack identification method, system and device and a storage medium, and the method comprises the following steps: implanting a CA attention mechanism into a Backbone layer of the CA attention mechanism, and constructing a YOLOv5s-CA model; collecting a test picture data set; dividing the test picture data set into a training set, a verification set and a test set; performing crack labeling on the pictures in the training set by adopting a crack label manufacturing tool; according to the marked training set, a YOLOv5s-CA model is trained; performing model parameter adjustment on the YOLOv5s-CA model according to the verification set; the YOLOv5s-CA model is evaluated according to the test set; a YOLOv5s-CA model is deployed to be used for concrete sleeper crack identification; collecting a real-time picture data set; the real-time picture data set is input into the deployed YOLOv5s-CA model to be processed; according to the method, polygon labeling of the sleeper cracks is achieved through Label labeling software, the accuracy of the data set standard is improved, a CA attention mechanism is implanted into the Backbone layer based on the YOLOv5s model, the accuracy of model target detection is effectively improved, and the concrete sleeper cracks can be accurately recognized.

Description

Technical Field

[0001] The present invention relates to the technical field of railway monitoring, and in particular to a method, system, equipment and storage medium for identifying cracks in concrete sleepers. Background Art

[0002] According to the data of the "2023 Railway Statistical Bulletin" released by the National Railway Administration, by the end of 2023, the total operating mileage of my country's high-speed railways has exceeded 45,000 kilometers. As the basis for installing rails, the performance status of sleepers is directly related to the safe operation of high-speed EMUs. However, due to the long-term exposure to the open air, the performance of sleepers is disturbed by external factors such as rain erosion, sun exposure, and line settlement, which can easily lead to cracks in sleepers. In addition, with the opening of my country's first modern high-speed railway-Beijing-Tianjin Intercity, my country's high-speed railway has been in operation for more than 16 years. The high-speed railway lines opened earlier will gradually enter the inspection and maintenance cycle. Therefore, the performance status detection workload of my country's huge high-speed railway network is huge, which provides broad application prospects for artificial intelligence technology to enable high-speed railway performance detection.

[0003] In fact, AI-enabled target detection has been widely used in many fields. In order to extract the information of molten pool and spatter from the image more accurately and effectively, Mao Yangkun et al. proposed a target detection model based on YOLOv5, which realized the real-time positioning and capture of spatter and molten pool in the forming process image. The experimental results show that the network can accurately locate the spatter and molten pool targets from the original image. While having good detection accuracy, the number of parameters of the network model is very small, which is more in line with the needs of industrial applications. In order to solve the problems of small target ambiguity, large target scale difference, and category imbalance in the detection process of sea surface ships, Ma Gan et al. designed a dynamic "copy-paste" data enhancement method, embedded it into the YOLOv5 framework, and proposed an improved YOLOv5s sea surface target detection algorithm. The experimental results show that the improved algorithm model and the proposed dynamic data enhancement method alleviate the problem of category and size imbalance and improve the accuracy of small target detection. Aiming at the problems of backwardness, low efficiency and insufficient small target recognition ability of traditional strip surface defect detection technology, Fu Qiang et al. proposed an improved YOLOv5s-Tiny target detection model, which improves the detection speed and recognition accuracy of strip surface defects while keeping the model computationally small. In addition, target detection technology based on artificial intelligence technology has also been deeply applied in agriculture, medicine and aviation. In recent years, with the concentrated outbreak of the operation and maintenance cycle of my country's high-speed railways, there is an urgent need for intelligent inspection technology represented by artificial intelligence, and many scholars have begun to focus on the empowerment of artificial intelligence technology in the railway field. Cui Xiaoning et al. applied computer vision technology to the intelligent recognition of cracks in double-block sleepers and established a crack target detection model based on YOLO-v5. The sleeper crack target detection experiment was completed through image annotation, parameter selection, model construction, model training and model prediction. By identifying cracks under different conditions such as normal light, dark light and microcracks, the YOLO-v5 target detection model showed high-precision intelligent recognition, reflecting the good generalization ability and robustness of the model. In view of the difficulty in capturing the global features of tunnel cracks, Zhou Zhong et al. proposed a high-precision intelligent detection algorithm for tunnel lining cracks, ST-YOLO, based on the YOLOv5 network framework by integrating the self-attention mechanism and convolutional neural network. The feature fusion module enhanced by convolutional attention is used to perform multi-scale fusion of the crack features extracted by the two branches to achieve the complementarity of the global features and local detail features of the cracks, and the classification confidence and location results of the cracks are obtained with the help of the decoupling head module output of YOLOv5. The application of the YOLO series of algorithms in the field of engineering practice has greatly improved work efficiency, and its application in bridge crack detection and other aspects has significantly improved the working conditions of engineers. However, how to improve the accuracy and credibility of model prediction has prompted scholars to make many explorations in improving and optimizing the relevant algorithm structure. In addition, there is a lack of identification methods specifically used to identify cracks in concrete sleepers in the existing technology. Summary of the invention

[0004] The purpose of the present invention is to provide a method, system, device and storage medium for identifying cracks in concrete sleepers. The method adopts Labelme annotation software to realize polygonal annotation of sleeper cracks, improves the accuracy of data set standards, and based on the YOLOv5s model, implants the CA attention mechanism into its Backbone layer, effectively improves the accuracy of model target detection, and can accurately identify cracks in concrete sleepers.

[0005] To achieve this object, the present invention adopts the following technical solutions: A method for identifying cracks in a concrete sleeper is provided, comprising the following steps: Based on the YOLOv5s model, the CA attention mechanism is implanted into its Backbone layer to build the YOLOv5s-CA model; Collect test image datasets; Dividing the test image dataset into a training set, a validation set and a test set; Use the crack labeling tool to annotate cracks in the images in the training set; Train the YOLOv5s-CA model based on the labeled training set; Adjust the model parameters of the YOLOv5s-CA model according to the validation set; The YOLOv5s-CA model is evaluated based on the test set and the evaluation results are obtained; Determine whether the evaluation results meet the requirements; If yes, return to execute the step of "collecting test image data set"; If yes, then deploy the YOLOv5s-CA model for concrete sleeper crack identification; Collect real-time image datasets; Input the real-time image dataset into the deployed YOLOv5s-CA model for processing; Output the recognition result.

[0006] As a preferred solution of the concrete sleeper crack identification method, in the step of using a crack labeling tool to label the images in the training set, the crack labeling tool is Labelme, and Labelme is used to create a polygonal labeling box to label the sleeper cracks.

[0007] As a preferred solution for the concrete sleeper crack recognition method, the evaluation indicators of the evaluation results include: detection accuracy Precision, recall rate Recall, mAP@0.5 and mAP@0.5:0.95; wherein the detection accuracy Precision is the ratio of the number of correctly predicted objects to the number of all predicted objects, the recall rate Recal is the ratio of the number of correctly predicted objects to the number of all objects, the mAP@0.5 is the average AP when the IOU threshold is 0.5, and mAP@0.5:0.95 represents the average mAP at different IOU thresholds (from 0.5 to 0.95, with a step length of 0.05); the calculation formulas are as follows: Where: C is the number of categories in the target detection task; k is the number of pictures; N The total number of pictures.

[0008] As a preferred solution of the concrete sleeper crack identification method, the ratio between the training set, the validation set and the test set is 6:2:2. The present invention also provides a concrete sleeper crack identification system, comprising: The model building module is used to build the YOLOv5s-CA model by embedding the CA attention mechanism into its Backbone layer based on the YOLOv5s model; The acquisition module is used to collect the test image data set; A partitioning module, used to divide the test image data set into a training set, a validation set and a test set; The labeling module is used to label cracks in the images in the training set using the crack labeling tool; The training module is used to train the YOLOv5s-CA model based on the labeled training set; The verification module is used to adjust the model parameters of the YOLOv5s-CA model according to the verification set; The evaluation module is used to evaluate the YOLOv5s-CA model based on the test set and obtain the evaluation results; A judgment module, used to judge whether the evaluation result meets the requirements; A loop module, used to return to execute the step of "collecting a test image data set"; A deployment module, used to deploy the YOLOv5s-CA model for concrete sleeper crack identification; The acquisition module is also used to acquire real-time image data sets; Running a recognition module to input the real-time image dataset into the deployed YOLOv5s-CA model for processing; The output module is used to output the recognition results.

[0009] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of any one of the above-mentioned methods for identifying cracks in concrete sleepers when executing the computer program.

[0010] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of any of the above-mentioned methods for identifying cracks in concrete sleepers are implemented.

[0011] Beneficial effects of the present invention: The concrete sleeper crack recognition method, system, device and storage medium proposed in the present invention use Labelme labeling software to realize polygonal labeling of sleeper cracks, improve the accuracy of data set standards, and based on the YOLOv5s model, implant the CA attention mechanism into its Backbone layer, effectively improve the accuracy of model target detection, and can accurately identify concrete sleeper cracks. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0013] Figure 1 is a flow chart of a method for identifying cracks in a concrete sleeper according to an embodiment of the present invention; Figure 2 is a structural block diagram of the CA attention mechanism according to an embodiment of the present invention; Figure 3 It is a network structure diagram after YOLOv5sCA according to an embodiment of the present invention adds CA attention mechanism; Figure 4 is a picture of a sleeper crack collected in one embodiment of the present invention; Figure 5 is a picture of a sleeper crack marking process in one embodiment of the present invention; Figure 6 It is a comparison chart of mAP@0.5 results of the YOLOv5s model and the YOLOv5s-CA model in one embodiment of the present invention; Figure 7 It is a schematic block diagram of the structure of a concrete sleeper crack identification system according to an embodiment of the present invention; Figure 8 is the sleeper crack detection result of the YOLOv5s model described in an embodiment of the present invention; Fig. 9 It is the sleeper crack detection result of the YOLOv5s-CA model described in an embodiment of the present invention; Fig.10 It is a schematic diagram of the structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0015] The following describes the embodiments of the present disclosure through specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0016] Reference Figure 1 An embodiment of the present invention provides a method for identifying cracks in a concrete sleeper, comprising the following steps: S1: Based on the YOLOv5s model, the CA attention mechanism is implanted into its Backbone layer to build the YOLOv5s-CA model; S2: Collect test image dataset; S3: Divide the test image dataset into a training set, a validation set and a test set; S4: Use the crack labeling tool to annotate cracks in the images in the training set; S5: Train the YOLOv5s-CA model based on the labeled training set; S6: Adjust the model parameters of the YOLOv5s-CA model according to the validation set; S7: Evaluate the YOLOv5s-CA model based on the test set and obtain the evaluation results; S8: Determine whether the evaluation result meets the requirements; S9: If yes, then return to execute the step of “collecting the test image data set”; S10: If yes, deploy the YOLOv5s-CA model for concrete sleeper crack identification; S11: collect real-time image data sets; S12: Input the real-time image dataset into the deployed YOLOv5s-CA model for processing; S13: Output recognition results.

[0017] As described in step S1 above, when humans observe things, the visual system automatically selects to focus on the key information of the things and automatically screens out the relatively unimportant parts. The attention mechanism in machine vision technology is similar to the human visual system, which is committed to focusing on certain important features of the detection target and its position information. The specific content of the CA module is divided into two parts: coordinate information embedding and coordinate attention generation. The CA module decomposes the input feature tensor into two one-dimensional feature encodings in the horizontal and vertical directions, enabling the model to capture long-distance dependencies between channels along one of the directions and maintain position information along the other direction. At this point, the target position information is saved in the generated attention map and then added to the input feature map through multiplication for the next convolution operation.

[0018] like Figure 3 As shown, C × H × W The main channels of the input feature map of shape perform avgpool, where C , H , W are the number of channels, height, and width of the feature map respectively. Use ( H , 1) and (1, W ) are respectively along X Axis and Y The pooling is performed in the axis direction to encode each channel and produce C × H ×1 and C ×1× W This enables the CA attention mechanism to capture long-distance dependencies within a channel and retain precise location information, allowing the network structure to obtain positioning information more accurately.

[0019] The YOLOv5s model mainly consists of three modules: Backbone, Neck and Head. Backbone is the main network, which mostly refers to the network for extracting features. Its function is to extract information from the image for use by the subsequent network. Neck is between Backbone and Head, and it further uses the features extracted by Backbone to improve the robustness of the model; the Head layer obtains the network output and can use the previously extracted features to make predictions.

[0020] Embedding the CA attention mechanism module into the Backbone layer of the YOLOv5s structure helps to process and utilize feature information more finely, especially in improving the sensitivity to spatial position and enhancing the ability to distinguish features, thereby improving the accuracy of target detection. The improved YOLOv5s-CA model network structure is shown in the figure. Figure 4 shown.

[0021] As described in step S2 above, collect a test image data set; Figure 5 As shown in Figure 1, the test image dataset is collected under different lighting and interference conditions, which can effectively test the robustness and generalization ability of the target detection model.

[0022] As described in the above step S3, the test picture data set is divided into a training set, a validation set and a test set; specifically, the test picture data set is divided into a training set, a validation set and a test set according to a ratio of 6:2:2.

[0023] As described in step S4 above, crack labeling tools are used to label cracks in the images in the training set; specifically, Figure 6 As shown, the crack labeling tool is Labelme. Through a large number of trial calculations, it is found that the sleeper crack labeling box cannot be too large or too small: if the labeling box is too large, the deep learning model will have poor recognition results for sleeper cracks; if the labeling box is too small, the deep learning model will easily learn the noise and interference in the crack image. Therefore, the present invention uses Labelme to create a polygonal labeling box to label sleeper cracks, providing a data basis for sleeper crack detection. After labeling with Labelme, a file in json format will be automatically generated (one json file for each image). After obtaining the json file, it needs to be converted into txt format before training the model.

[0024] As in steps S5-S7 above, the YOLOv5s-CA model is trained and fitted with the model weight parameters according to the labeled training set, the validation set is used to adjust the model parameters in order to obtain the optimal model, and the test set uses the obtained optimal model to perform final output prediction and evaluation to obtain the evaluation results; specifically, the prediction results of the model are divided into four types: true positive (TP), false positive (FP), true negative (TN), and false negative (FN). The evaluation indicators of the evaluation results include: detection accuracy Precision, recall rate Recall, mAP@0.5 and mAP@0.5:0.95; wherein, the detection accuracy Precision is the ratio of the number of correctly predicted objects to the number of all predicted objects, the recall rate Recal is the ratio of the number of correctly predicted objects to the number of all objects, mAP (mean average precision) refers to the average accuracy, which is used to measure the quality of the model in all categories, and the mAP@0.5 is the average AP when the IOU threshold is 0.5, which is mainly used to reflect the recognition ability of the model; mAP@0.5:0.95 represents the average mAP at different IOU thresholds (from 0.5 to 0.95, step size 0.05), which is mainly used to reflect the positioning effect and boundary regression ability. The calculation formulas are as follows: Where: C is the number of categories in the target detection task; k is the number of pictures; N The total number of pictures.

[0025] Test and compare the YOLOv5s model with the YOLOv5s-CA model: The experimental environment parameters are shown in Table 1 below. Table 1 Experimental environment parameters like Figure 7 The following table shows the training results of mAP@0.5 of the YOLOv5s model and the YOLOv5s-CA model at each epoch. The comparison of the evaluation indicators of the two models is shown in Table 2.

[0026] Table 2 Comparison of detection results of different models From the above results, we can know that: (1) The YOLOv5s and YOLOv5s-CA models fluctuated greatly in the early stage of training, but gradually became stable after 200 training times. The average detection accuracy of the YOLOv5s-CA model for concrete sleeper cracks was higher than that of the YOLOv5s.

[0027] (2) The detection accuracy of the YOLOv5s model for concrete sleeper cracks is 90.88%, and the average detection accuracy is 86.87%. After adding the CA attention mechanism module, the feature extraction capability of the model has been improved, the detection accuracy of sleeper cracks has increased by 1.8%, mAP@0.5 has increased by 3.54%, mAP@0.5:0.95 has increased by 9.39%, and the recall rate has increased by 4.44%.

[0028] (3) By adding the CA attention mechanism, the evaluation indicators of the YOLOv5s model are comprehensively improved, indicating that the improved YOLOv5s-CA model has good applicability for sleeper crack target detection.

[0029] In order to demonstrate the generalization ability and strong robustness of the model, the detection results of sleeper cracks under different lighting conditions and different color conditions are selected for display. The results are as follows: Figure 8 and Fig. 9 shown.

[0030] Depend on Figure 8 and Fig. 9 It can be seen that the confidence of the original crack detection of the YOLOv5s model under different conditions is low, and there are missed detections and over-detection phenomena. The detection rate and confidence of the sleeper cracks using the YOLOv5s-CA model are higher, which reflects the good generalization ability and robustness of the YOLOv5s-CA model. The model of the present invention can play an important role in the actual detection and intelligent identification of concrete sleeper cracks.

[0031] As described in the above steps S8-S10: determine whether the evaluation results meet the requirements; if not, return to execute the step of "collecting the test image data set"; re-train the model; if so, deploy the YOLOv5s-CA model for concrete sleeper crack identification. As in the above steps S11-S13: using a drone or other equipment carrying a camera to collect a real-time image data set; inputting the real-time image data set into the deployed YOLOv5s-CA model for processing; identifying whether there are cracks, outputting the identification results, and realizing the function of accurately identifying cracks in concrete sleepers.

[0032] In this embodiment, Labelme annotation software is used to realize polygonal annotation of sleeper cracks, improve the accuracy of the data set standard, and based on the YOLOv5s model, the CA attention mechanism is implanted into its Backbone layer, which effectively improves the accuracy of model target detection and can accurately identify concrete sleeper cracks.

[0033] like Figure 2 As shown, an embodiment of the present invention further provides a concrete sleeper crack identification system, comprising: The model building module is used to build the YOLOv5s-CA model by embedding the CA attention mechanism into its Backbone layer based on the YOLOv5s model; The acquisition module is used to collect the test image data set; A partitioning module, used to divide the test image data set into a training set, a validation set and a test set; The labeling module is used to label cracks in the images in the training set using the crack labeling tool; The training module is used to train the YOLOv5s-CA model based on the labeled training set; The verification module is used to adjust the model parameters of the YOLOv5s-CA model according to the verification set; The evaluation module is used to evaluate the YOLOv5s-CA model based on the test set and obtain the evaluation results; A judgment module, used to judge whether the evaluation result meets the requirements; A loop module, used to return to execute the step of "collecting a test image data set"; A deployment module, used to deploy the YOLOv5s-CA model for concrete sleeper crack identification; The acquisition module is also used to acquire real-time image data sets; Running a recognition module to input the real-time image dataset into the deployed YOLOv5s-CA model for processing; The output module is used to output the recognition results.

[0034] Reference Fig.10 In an embodiment of the present application, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Fig.10As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a storage medium and an internal memory. The storage medium stores an operating system, a computer program and a database. The memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store test image data sets, real-time image data sets, etc. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the concrete sleeper crack method described in any of the above embodiments can be implemented.

[0035] Those skilled in the art will understand that Fig.10 The structure shown in is merely a block diagram of a portion of the structure related to the present application solution and does not constitute a limitation on the computer device to which the present application solution is applied.

[0036] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the concrete sleeper crack treatment method described in any of the above embodiments can be implemented.

[0037] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0038] In the description of the present invention, it is necessary to understand that the terms "middle", "length", "up", "down", "front", "back", "vertical", "horizontal", "inside", "outside", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0039] In the present invention, unless otherwise clearly specified and limited, the first feature "on" the second feature may be that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. "Multiple" means at least two, such as two, three, etc., unless otherwise clearly and specifically limited.

[0040] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral one; it can be a mechanical connection, an electrical connection, or communication with each other; it can be a direct connection, or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0041] The above is only for explaining the implementation mode of the present invention and is not intended to limit the present invention. For those skilled in the art, any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention without creative work should be included in the protection scope of the present invention.

Claims

1. A method for identifying cracks in concrete sleepers, characterized in that: The following steps are involved: Based on the YOLOv5s model, the CA attention mechanism is implanted into its Backbone layer to build the YOLOv5s-CA model; Collect test image datasets; Dividing the test image dataset into a training set, a validation set and a test set; Use the crack labeling tool to annotate cracks in the images in the training set; Train the YOLOv5s-CA model based on the labeled training set; Adjust the model parameters of the YOLOv5s-CA model according to the validation set; The YOLOv5s-CA model is evaluated based on the test set and the evaluation results are obtained; Determine whether the evaluation results meet the requirements; If yes, return to execute the step of "collecting test image data set"; If yes, then deploy the YOLOv5s-CA model for concrete sleeper crack identification; Collect real-time image datasets; Input the real-time image dataset into the deployed YOLOv5s-CA model for processing; Output the recognition result.

2. The concrete sleeper crack identification method according to claim 1, characterized in that: In the step of using a crack labeling tool to label the images in the training set, the crack labeling tool is Labelme, and Labelme is used to create a polygonal labeling box to label the sleeper cracks.

3. The method for identifying cracks in concrete sleepers according to claim 1, characterized in that: The evaluation indicators of the evaluation results include: detection accuracy Precision, recall rate Recall, mAP@0.5 and mAP@0.5:0.95; wherein, the detection accuracy Precision is the ratio of the number of correctly predicted objects to the number of all predicted objects, the recall rate Recal is the ratio of the number of correctly predicted objects to the number of all objects, the mAP@0.5 is the average AP when the IOU threshold is 0.5, and mAP@0.5:0.95 represents the average mAP at different IOU thresholds (from 0.5 to 0.95, with a step size of 0.05); the calculation formulas are as follows: Where: C is the number of categories in the target detection task; k is the number of pictures; N The total number of pictures.

4. The method for identifying cracks in concrete sleepers according to claim 1, characterized in that: The ratio between the training set, validation set and test set is 6:2:

2.

5. A concrete sleeper crack identification system, characterized in that: include: The model building module is used to build the YOLOv5s-CA model by embedding the CA attention mechanism into its Backbone layer based on the YOLOv5s model; The acquisition module is used to collect the test image data set; A partitioning module, used to divide the test image data set into a training set, a validation set and a test set; The labeling module is used to label cracks in the images in the training set using the crack labeling tool; The training module is used to train the YOLOv5s-CA model based on the labeled training set; The verification module is used to adjust the model parameters of the YOLOv5s-CA model according to the verification set; The evaluation module is used to evaluate the YOLOv5s-CA model based on the test set and obtain the evaluation results; A judgment module, used to judge whether the evaluation result meets the requirements; A loop module, used to return to the step of "collecting a test image data set"; A deployment module, used to deploy the YOLOv5s-CA model for concrete sleeper crack identification; The acquisition module is also used to acquire real-time image data sets; Running a recognition module to input the real-time image dataset into the deployed YOLOv5s-CA model for processing; The output module is used to output the recognition results.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the concrete sleeper crack identification method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the concrete sleeper crack identification method according to any one of claims 1 to 4 are implemented.

Images