A method and system for identifying road hidden diseases by three-dimensional radar map

Abstract

The application discloses a three-dimensional radar map road hidden disease identification method and system, the method comprising: constructing an initial improved model based on YOLOv8, adding a target Cut_SimAM attention mechanism at the end of the head network, the feature extraction module being C2f_EFAttention, the loss function being a SlideLoss sliding weighted function, and the boundary box loss function being a D-IoU boundary box loss function; obtaining a radar disease map transmitted by a three-dimensional ground penetrating radar, and preprocessing the radar disease map to obtain a training data set; training the initial improved model based on the training data set to obtain a target improved model; inputting a to-be-tested radar image of a to-be-tested road into the target improved model to identify hidden disease information of the to-be-tested road. The application improves the YOLOv8 model through the Cut_SimAM attention mechanism, C2f_EFAttention, SlideLoss sliding weighted function and D-IoU boundary box loss function to obtain a lightweight model applied to the identification of road hidden diseases, so that high-precision and high-efficiency real-time detection and identification are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of radar monitoring, in particular to a three-dimensional radar map road hidden disease identification method and system. BACKGROUND

[0002] At present, the highway network in China has been gradually improved, and with the increasing service life, under the action of heavy load and complex environment, the asphalt pavement structure gradually appears damage, such as structural cracks, interlayer defects, loose and other diseases, and the transportation infrastructure has gradually entered the large-scale maintenance stage. With the development of new nondestructive testing technology, ground penetrating radar provides a solution for the fine identification of internal road diseases. The A-scan and B-scan data collected by two-dimensional ground penetrating radar determine the structure and spatial position of underground medium through radio wave waveform, amplitude intensity and other characteristics, while three-dimensional ground penetrating radar (3D Ground Penetrating Radar) uses a multi-channel three-dimensional antenna array to collect C-scan data, combines time domain and frequency domain characteristic parameters, and can generate three-dimensional images of road cross section and horizontal section, which realizes the "point to surface" change, greatly improves the detection efficiency and accuracy, and can achieve the purpose of rapid diagnosis and evaluation of diseases. Compared with traditional core sampling and ultrasonic detection[4]and other methods, the rapid, efficient and nondestructive three-dimensional ground penetrating radar detection technology has become one of the most widely used technical means for road disease detection.

[0003] With the rapid development of deep learning technology, applying digital and intelligent technology to three-dimensional ground penetrating radar data analysis is beneficial to improve the radar detection efficiency and realize real-time detection of road hidden diseases.

[0004] At present, the application of deep learning technology in the research process of road hidden disease identification still faces many problems, such as the lack of real ground penetrating radar map data, the difference between different types of ground penetrating radar due to the difference in hardware equipment parameters and software interpretation method, which leads to the difficulty in matching radar disease map data; the complex ground penetrating radar signal characteristics, such as aliasing wave, noise and clutter interference information, the difference in understanding of road hidden disease form in artificial identification and labeling may lead to misjudgment, and the low efficiency of artificial interpretation in the face of large-scale data; the lack of targeted detection method, data processing and image analysis technology. The recognition accuracy of conventional single-stage algorithm is low, the recognition efficiency of two-stage algorithm is low, the model size is large, and it is difficult to embed the device to realize real-time detection.

[0005] Therefore, how to realize high-precision real-time detection of road hidden diseases by three-dimensional ground penetrating radar is a problem to be solved. SUMMARY

[0006] In order to solve the above problems, the embodiment of the present application provides a three-dimensional radar map road hidden disease identification method and system.

[0007] In a first aspect, in order to solve the above technical problems, the present application provides a three-dimensional radar map road hidden disease identification method, comprising:

[0008] The initial improved model is constructed based on a YOLOv8 model, a target Cut_SimAM attention mechanism is added at the end of the head network of the initial improved model, the feature extraction module is C2f_EFAttention, the loss function is a SlideLoss sliding weighted function, and the boundary box loss function is a D-IoU boundary box loss function;

[0009] A radar disease map transmitted by a three-dimensional ground penetrating radar is acquired, and the radar disease map is preprocessed to obtain a training data set;

[0010] The initial improved model is trained based on the training data set to obtain a target improved model;

[0011] A to-be-detected radar image of a to-be-detected road is input into the target improved model, and hidden disease information of the to-be-detected road is identified.

[0012] Beneficial effects are:

[0013] In the technical scheme provided by the embodiment of the present application, the initial improved model is constructed based on the baseline model YOLOv8 model, a target Cut_SimAM attention mechanism is added at the end of the head network of the initial improved model, the feature extraction module is C2f_EFAttention, the loss function is a SlideLoss sliding weighted function, and the boundary box loss function is a D-IoU boundary box loss function. The target Cut_SimAM attention mechanism can make the edges of feature information clearer, thereby focusing on the disease area in the image, and C2f_EFAttention and the SlideLoss sliding weighted function can improve the detection speed of the model, so that the constructed initial model can improve the detection precision and operation efficiency of the model. Then, a radar disease map transmitted in real time by a three-dimensional ground penetrating radar is acquired, the radar disease map is preprocessed to obtain a training data set, the initial improved model is trained based on the training data set to obtain a target improved model, that is, a lightweight model for three-dimensional radar map road hidden disease identification is obtained, so as to realize high-precision and high-efficiency real-time detection and identification of road hidden diseases.

[0014] In a second aspect, the present application provides a three-dimensional radar map road hidden disease identification system, comprising a model construction unit, an acquisition unit, a training unit and an identification unit.

[0015] The model construction unit is configured to construct an initial improved model based on a YOLOv8 model, the head network of the initial improved model is added with a target Cut_SimAM attention mechanism, the feature extraction module is C2f_EFAttention, the loss function is a SlideLoss sliding weighted function, and the boundary box loss function is a D-IoU boundary box loss function.

[0016] The acquisition unit is configured to acquire a radar disease map transmitted by a three-dimensional ground penetrating radar and preprocess the radar disease map to obtain a training data set.

[0017] The training unit is configured to train the initial improved model based on the training data set to obtain a target improved model.

[0018] The recognition unit is configured to input a to-be-tested radar image of a to-be-tested road into the target improved model to recognize hidden disease information of the to-be-tested road.

[0019] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings incorporated in and forming a part of the specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the application. It is explicitly contemplated that the drawings described below are only some embodiments of the present application, and for those of ordinary skill in the art, other drawings can be obtained without creative labor on the basis of these drawings. In the drawings:

[0021] Figure 1 is a flow chart of a road hidden disease recognition method of a three-dimensional radar map according to an exemplary embodiment of the present application;

[0022] Figure 2 is a structural schematic diagram of an initial improved model according to an exemplary embodiment of the present application;

[0023] Figure 3 is a schematic diagram of a target Cut_SimAM attention mechanism according to an exemplary embodiment of the present application;

[0024] Figure 4 is a schematic diagram of C2f_EFAttention according to an exemplary embodiment of the present application;

[0025] Figure 5 is a comparative schematic diagram of the output layer heat map of the verification model;

[0026] Figure 6 is a comparative schematic diagram of the detection effect diagram of the verification model;

[0027] Figure 7 is a block diagram of a road hidden disease identification system of a three-dimensional radar map shown by an example embodiment of the present application;

[0028] Figure 8 is a structural schematic diagram of a computer system of an electronic device suitable for implementing embodiments of the present application. DETAILED DESCRIPTION

[0029] The example embodiments will be described in detail herein with reference to the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following example embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of apparatuses and methods consistent with some aspects of the present application as detailed in the appended claims.

[0030] The block diagrams shown in the drawings are merely functional entities, and do not necessarily have to correspond to physically independent entities. That is, the functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0031] The flowcharts shown in the drawings are merely illustrative, and do not necessarily include all contents and operations / steps, nor do they have to be executed in the order described. For example, some operations / steps can be further divided, and some operations / steps can be combined or partially combined, so the actual execution order can be changed according to the actual situation.

[0032] In the present application, "multiple" refers to two or more. The association relationship of "and / or" describes the associated objects, which means that there can be three relationships, for example, A and / or B can represent the three cases of A alone, A and B together, and B alone. The character " / " generally represents an "or" relationship between the associated objects before and after it.

[0033] In order to solve the problems of low recognition accuracy, low efficiency, large model size, and difficulty in embedding devices to realize real-time detection when applying deep learning technology to road hidden disease identification, the embodiments of the present application propose a three-dimensional radar map road hidden disease identification method and device, electronic device, and computer readable storage medium, which mainly relate to the three-dimensional radar map road hidden disease identification technology included in the road radar monitoring technology. The embodiments will be described in detail below.

[0034] First, please refer to Figure 1 , Figure 1is a flowchart of a road hidden disease identification method of a three-dimensional radar map according to an example embodiment of the present application. The method can be specifically performed by a server, which can be an independent server or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and basic cloud computing services such as big data and artificial intelligence platforms, which are not limited herein.

[0035] As shown in Figure 1 In an example embodiment, the road hidden disease identification method of the three-dimensional radar map can include steps S101 to S104, which are described in detail as follows:

[0036] Step S101, an initial improved model is constructed based on a YOLOv8 model, a target Cut_SimAM attention mechanism is added at the end of the head network of the initial improved model, the feature extraction module is C2f_EFAttention, the loss function is a SlideLoss sliding weighted function, and the boundary box loss function is a D-IoU boundary box loss function.

[0037] The initial improved model constructed in the present application adds an improved target Cut_SimAM attention mechanism to the head network, replaces the C2f feature extraction module in the original network with C2f_EFAttention in the backbone network, replaces the BCE loss function in the original network with a SlideLoss sliding weighted function, and replaces the C-IoU boundary box loss function in the original network with a D-IoU boundary box loss function. As shown in Figure 2 Figure 2 is a structural diagram of the initial improved model in an example embodiment of the present application.

[0038] Step S102, a radar disease map transmitted by a three-dimensional ground penetrating radar is obtained, and the radar disease map is preprocessed to obtain a training data set.

[0039] In the present embodiment, the radar disease map can be obtained by detecting municipal roads using a vehicle-mounted three-dimensional array radar, and the radar disease map is preprocessed to obtain a training data set.

[0040] Step S103, the initial improved model is trained based on the training data set to obtain a target improved model.

[0041] After the construction of the initial improved model and the acquisition of the training data set, the initial improved model is trained based on the training data set to obtain a target improved model adapted to road hidden disease identification, so that it can be applied to real-time detection of road hidden diseases.​

[0042] In step S104, the to-be-detected radar image of the to-be-detected road is input into the target improved model to identify the hidden disease information of the to-be-detected road.

[0043] In the actual application of the target improved model, the to-be-detected radar image of the to-be-detected road is obtained in real time by the three-dimensional ground penetrating radar, and is input into the obtained target improved model, so as to use the target improved model to identify the road hidden disease of the to-be-detected radar image, and detect the hidden disease information of the to-be-detected road.

[0044] As can be seen from the above, in the method provided in the embodiment, on the one hand, the initial improved model is constructed on the basis of the baseline model YOLOv8 model, the target Cut_SimAM attention mechanism is added at the tail end of the head network of the initial improved model, the feature extraction module is C2f_EFAttention, and the loss function is the SlideLoss sliding weighted function, and the boundary box loss function is the D-IoU boundary box loss function. The target Cut_SimAM attention mechanism can make the edges of the feature information clearer, so as to focus on the disease area in the image, and the C2f_EFAttention and the SlideLoss sliding weighted function can improve the detection speed of the model, so that the constructed initial model can improve the detection accuracy and operation efficiency of the model. On the other hand, after the initial improved model is constructed, the radar disease graph transmitted in real time by the three-dimensional ground penetrating radar is obtained, the radar disease graph is preprocessed to obtain a training data set, the initial improved model is trained based on the training data set, and the target improved model is obtained, that is, a lightweight model for identifying the road hidden disease of the three-dimensional radar graph is obtained, so as to realize the real-time detection and identification of the road hidden disease with high precision and high efficiency.

[0045] In an example embodiment provided in the present application, the SlideLoss sliding weighted function replaces the original BCE (Binary Cross-Entropy Loss, binary cross-entropy loss function) loss function in the network, and the basic formula of the SlideLoss sliding weighted function is:

[0046]

[0047] Among them, the threshold μ is the average value of the intersection over union of all sample boundary boxes, and the positive sample is greater than μ, and the negative sample is less than μ.

[0048] In an example embodiment provided in the present application, the specific steps of preprocessing the radar disease graph to obtain the training data set include:

[0049] The radar disease atlas is filtered, and the filtered radar image is labeled and classified according to the disease type to obtain an initial data set; the disease types include interlayer poor, interlayer water, interlayer loose, and structure loose.

[0050] The initial data set is subjected to data enhancement processing to obtain a training data set.

[0051] In this embodiment, the Swedish ImpulseRadar vehicle-mounted three-dimensional array radar Raptor-80 can be used to collect the radar disease atlas, and then the radar disease atlas is subjected to bandpass filtering, background removal, and spectral analysis to remove clutter and noise. The parameters of the radar device are shown in Table 1.

[0052] Table 1: Parameters of the Raptor-80 vehicle-mounted radar

[0053] Technical index Antenna center frequency Channel number Sampling interval Time window Residence time Detection speed Technical parameter 800 MHz 24 2.5-5 cm 25 ns-75 ns 3 μs 30 km / h

[0054] After the filtering of the radar disease atlas is completed, it is labeled and classified according to the disease type, such as using the labeling tool LabelMe to label the filtered radar disease atlas, generating XML labeling information, and obtaining an initial data set based on the radar disease atlas and the labeling information. Further data enhancement processing is performed on the initial data set to obtain a training data set. The disease types include interlayer poor, interlayer water, interlayer loose, and structure loose, which correspond to the labeling information poor_l, water_l, loose_l, and loose_s, respectively.

[0055] In another exemplary embodiment, the data augmentation processing includes blur noise, sharpening processing, random brightness and contrast, random inversion, and image stitching, so the specific steps are: sequentially performing blur noise, sharpening processing, random brightness and contrast, random inversion, and image stitching on the initial data set to obtain a training data set.

[0056] The specific steps of the blur noise are: using a generalized normal filter with randomly selected parameters to blur the images in the initial data set. This conversion also adds multiplication noise to the generated kernel before convolution, combining blur and noise injection to enhance the data.

[0057] The specific steps of the sharpening processing are: sharpening the initial data set and superimposing it with the original image, and the sharpening degree is randomly within a specified range. The edges and details of the image are enhanced, making the overall image clearer.

[0058] The specific steps of the random brightness and contrast are: randomly increasing or decreasing the brightness and contrast of the initial data set within a specified range.

[0059] The specific steps of random flipping are: randomly flipping the initial data set, including left-right flipping and up-down flipping.

[0060] The image stitching process involves randomly selecting four different images from the initial dataset, randomly cropping and scaling them to fit the dimensions of the new stitched image, and then stitching the four processed images together into a single 640×640 image. Based on the stitched image, the coordinates of the target box are corrected to correspond to the new image coordinate system.

[0061] In an exemplary embodiment provided by this application, the specific steps of training the initial improved model to obtain the target improved model include:

[0062] Divide the training data set based on the preset division ratio to obtain the corresponding training set, validation set and test set;

[0063] Based on the training set, validation set and test set, the initial improved model is trained according to the preset training batches and training rounds to obtain the target improved model; among them, the division ratio is 4:1:1, the training batches are 16, and the training rounds are 300.

[0064] In this embodiment, the training dataset is partitioned using a 4:1:1 ratio, and the basic hyperparameters for training the improved model are set to 16 training batches and 300 training rounds. For example, a dataset of 2451 radar damage maps collected by a 3D ground-penetrating radar includes 1600 of these images as a training set, 400 as a validation set, and 451 as a test set. The basic hyperparameters for training the improved model are set to 16 training batches, 300 training rounds, and the AdamW optimizer. All other hyperparameters remain at their default values.

[0065] In the embodiments provided in this application, the experimental environment configuration for training the initial improved model can be seen in Table 2.

[0066] Table 2: Experimental environment configuration

[0067]

[0068]

[0069] In an exemplary embodiment provided by this application, in the process of training the initial improved model based on the training data set to obtain the target improved model, the specific steps of the target Cut_SimAM attention mechanism for data processing of the training data include:

[0070] Using the target Cut_SimAM attention mechanism, the input feature map of the input target Cut_SimAM attention mechanism is cut into four cut blocks;

[0071] The cut picture block is subjected to mean subtraction processing and normalization processing, and the processed cut picture block is activated by a Sigmoid function to obtain an activated picture block;

[0072] The first feature matrix of the cut picture block and the second feature matrix of the activated picture block are obtained, and the second feature matrix and the second feature matrix are direct product to obtain a target picture block;

[0073] The four target body blocks are subjected to feature splicing to obtain the output feature map of the head network in the target improved model.

[0074] In the embodiment, the target Cut_SimAM attention mechanism is a Cut_SimAM attention mechanism improved for road hidden disease identification, please refer to Figure 3 , Figure 3 In an example embodiment of the present application, a schematic diagram of the target Cut_SimAM attention mechanism is shown. As Figure 3 shown, the improved Cut_SimAM attention mechanism in the model, for the input feature map of the input attention mechanism, first cuts the input feature map into four cut picture blocks, then subjects each cut picture block to mean subtraction processing and normalization processing, and activates the processed cut picture block by a Sigmoid function to obtain an activated picture block; the first feature matrix of the cut picture block and the second feature matrix of the activated picture block are obtained, and the second feature matrix and the second feature matrix are direct product to obtain a target picture block; the four target body blocks are subjected to feature splicing to obtain the output feature map of the head network in the target improved model.

[0075] In addition, it should be noted that the target improved model performs detection and identification, and the target Cut_SimAM attention mechanism also follows the above data processing process.

[0076] In this way, through the above embodiments, the improved Cut_SimAM attention mechanism is added to the head network, so that the improved model is more focused on the identification of useful features, that is, the accuracy of the model is improved.

[0077] In an example embodiment provided by the present application, the C2f_EFAttention includes an EFAttention module; in the process of training the initial improved model based on the training data set to obtain the target improved model, the specific steps of the C2f_EFAttention for data processing of the training data include:

[0078] The input of the C2f_EFAttention is convolved once by the C2f_EFAttention to obtain an intermediate feature map;

[0079] The intermediate feature map is split to obtain a first part feature map and a second part feature map;

[0080] The second part feature map is processed by the EFAttention module for feature enhancement, and the processed second part feature map and the first part feature map are spliced for feature to obtain the output of the C2f_EFAttention module.

[0081] In this embodiment, the C2f_EFAttention is used in the backbone network to replace the C2f feature extraction module in the original network, please refer to Figure 4 , Figure 4 In an example embodiment of the present application, the C2f_EFAttention is a schematic diagram. As shown in Figure 4 , the C2f_EFAttention in the model, the input data is convolved to obtain an intermediate feature map, the intermediate feature map is split to obtain a first part feature map and a second part feature map. Then the second part feature map is processed by the EFAttention module for feature enhancement, and the processed second part feature map and the first part feature map are spliced for feature to obtain the output of the C2f_EFAttention module.

[0082] Among them, the EFAttention module is in the channel layer, the feature map is reduced to the channel descriptor by global average pooling (GAP, Global average pooling), then 1D convolution is performed to process the channel information, the attention weight is output by the Sigmoid function, and the input feature map is weighted; In the spatial layer, 1x1 convolution is applied to calculate the spatial weight, and then the attention weight is output by the Sigmoid function, and the input feature map is weighted. Finally, the weighted results of the channel layer and the spatial layer are added to enhance the useful features.

[0083] In addition, it should be noted that when the target improved model detects and identifies, the C2f_EFAttention also follows the above data processing process.

[0084] In this way, through the above embodiments, the C2f_EFAttention is used in the backbone network to replace the C2f feature extraction module in the original network, and the characteristics of the C2f_EFAttention are applied to the data processing process of the feature map to construct an improved model to improve the detection speed of the model.

[0085] In an example embodiment of the present application, after the target improved model is trained, the model performance needs to be detected to make the target improved model reach the expected detection accuracy and efficiency, and the specific steps include:

[0086] an evaluation index of the target improved model, the evaluation index including a detection accuracy evaluation index, a detection efficiency evaluation index, and a lightweight evaluation index;

[0087] obtaining a model performance of the target improved model based on the detection accuracy evaluation index, the detection efficiency evaluation index, and the lightweight evaluation index.

[0088] In another exemplary embodiment, the detection efficiency evaluation index is a frame per second (FPS), the lightweight evaluation index is a parameter quantity (Params, referring to a total number of parameters in the network model that need to be trained) and a calculation quantity (FLOPs, floating point operations, meaning floating point operations) of the target improved model, and the detection accuracy evaluation index includes a recall R e (Recall), a precision P r (Precision), and a multi-class average precision (mAP).

[0089]

[0090] wherein TP is a true positive, FP is a false positive, FN is a false negative, C is a class number of the disease type, and C=4.

[0091] In this way, through the above embodiments, after the target improved model is trained, the model performance thereof is detected by obtaining the frame per second, the parameter quantity, the calculation quantity, the recall, the precision, and the multi-class average precision, so that the target improved model can be further optimized based on the detected model performance of the target improved model, and high-precision and high-efficiency real-time detection and recognition of the road latent disease are realized.

[0092] In an example embodiment provided by the present application, in order to verify the effectiveness of the target improvement model and the influence of each improvement module in the model, seven groups of verification models are designed for ablation experiments based on the baseline model YOLOv8. The module application of each group of verification models is shown in Table 3 below. The model YOLOv8-Sim represents adding a SimAM attention mechanism at the end of the backbone layer of YOLOv8. The model YOLOv8-C represents replacing the SimAM attention mechanism in YOLOv8-SimAM with a Cut_SimAM module with a cutting function. YOLOv8-E represents replacing the C2f module in the original network structure with a C2f_EFAttention module at the neck layer. The model YOLOv8-S represents replacing the original network using a BCE loss function with a SlideLoss. The model YOLOv8-CE represents the comprehensive use of the Cut_SimAM module and the C2f_EFAttention module. The model YOLOv8-CES represents the comprehensive use of the Cut_SimAM module, the C2f_EFAttention module and the SlideLoss sliding loss function.

[0093] In the training phase, compared with the baseline model YOLOv8, the YOLOv8-Sim model with the added SimAM attention mechanism and the YOLOv8-C model with the added Cut_SimAM module, the maximum value of each channel feature of the disease atlas of the three models is taken as the heat map pixel value, and the visualized results are shown in the form of heat maps as shown in Figure 5 Figure 5 is a comparison diagram of the output layer heat map of the verification model, wherein red represents high attention of the model to the area, and blue represents low attention of the model to the area.

[0094] Therefore, compared with the baseline model, the SimAM attention mechanism reassigns the weight value to the global feature, enhances the feature extraction capability of the model, and enhances the features of the target that is not originally noticed. Compared with the SimAM attention mechanism, the Cut_SimAM module makes the edge of the feature information clearer, focuses on the disease area in the image, and ignores the non-disease area information, but at the same time, it also enhances the attention to the unlabeled area, i.e., the suspected disease area.

[0095] In the verification process of the present embodiment, 451 radar disease atlases in the verification set are used to perform ablation experiments on the weight file saved by the training result, and the results are shown in Table 3.

[0096] Table 3: Ablation experiment

[0097]

[0098] Please refer to Figure 6 , Figure 6 ​is a contrast diagram of the detection effect diagram of the verification model. As shown in Figure 6 As shown in Table 3, the YOLOv8-Sim model added with the SimAM attention mechanism has improved recognition accuracy, and the mAP is improved by 1.3% compared with the YOLOv8na baseline model, but the FPS is reduced by about 15, which indicates that the processing speed of the model is thus improved by 8%. The YOLOV8-C model added with the Cut_SimAM attention mechanism does not improve the processing speed, but the mAP is improved by 0.8% compared with the YOLOv8-Sim. The YOLOv8-C model using the C2f_EFAttention module reduces the parameter amount by 0.5M and the calculation amount by 1GFLOPs, and improves the detection speed of the model. The YOLOv8-S model using the SlideLoss sliding weighting function greatly improves the FPS without reducing the parameter amount. This shows that the C2f_EFAttention module and the SlideLoss are both helpful to improve the operation efficiency of the model.

[0099] Therefore, the three-dimensional radar map road hidden disease identification method provided by the embodiment of the application, by adding the improved target Cut_SimAM attention mechanism in the head network, using the C2f_EFAttention instead of the C2f feature extraction module in the original network in the backbone network, using the SlideLoss sliding weighting function to replace the BCE loss function in the original network, using the D-IoU bounding box loss function to replace the C-IoU bounding box loss function in the original network, and after training, the target improved model YOLOv8-CES model is constructed, the YOLOv8-CES model integrates the above-mentioned Cut_SimAM attention mechanism, C2f_EFAttention and SlideLoss sliding weighting function these three optimization methods, achieves the highest recognition accuracy, and the mAP is improved by 3.6% compared with the baseline model, and the processing speed is also improved by 16.7%, wherein the C2f_EFAttention module reduces the parameter amount and the number of floating point operations per second, and the lightweight model is more advantageous for the embedded device of the vehicle-mounted ground penetrating radar.

[0100] Figure 7 is a block diagram of a three-dimensional radar map road hidden disease identification system 700 according to an example embodiment of the application. As shown in Figure 7 The system includes:

[0101] The model construction unit 701 is configured to construct an initial improved model based on a YOLOv8 model, the initial improved model has a target Cut_SimAM attention mechanism added at the end of a head network, a feature extraction module is C2f_EFAttention, a loss function is a SlideLoss sliding weighted function, and a bounding box loss function is a D-IoU bounding box loss function.

[0102] The acquisition unit 702 is configured to acquire a radar disease map transmitted by a three-dimensional ground penetrating radar, and preprocess the radar disease map to obtain a training data set.

[0103] The training unit 703 is configured to train the initial improved model based on the training data set to obtain a target improved model.

[0104] The identification unit 704 is configured to input a to-be-detected radar image of a to-be-detected road into the target improved model to identify implicit disease information of the to-be-detected road.

[0105] The system applies the road implicit disease identification method of the three-dimensional radar map provided in the present application, and the model construction unit 701 constructs an initial improved model based on the baseline model YOLOv8 model, the initial improved model has a target Cut_SimAM attention mechanism added at the end of a head network, a feature extraction module is C2f_EFAttention, a loss function is a SlideLoss sliding weighted function, and a bounding box loss function is a D-IoU bounding box loss function. Then, the acquisition unit 702 acquires a radar disease map transmitted in real time by a three-dimensional ground penetrating radar, and preprocesses the radar disease map to obtain a training data set, the training unit 703 trains the initial improved model based on the training data set to obtain a target improved model, that is, a lightweight model for identifying road implicit diseases of a three-dimensional radar map is obtained, so that the identification unit 704 can realize real-time detection and identification of road implicit diseases with high precision and high efficiency.

[0106] In another exemplary embodiment, the acquisition unit 702 is further configured to perform filtering processing on the radar disease map, and label and classify the filtered radar image according to disease types to obtain an initial data set; the disease types include interlayer defects, interlayer water, interlayer looseness, and structural looseness; and the initial data set is subjected to data enhancement processing to obtain the training data set.

[0107] In another exemplary embodiment, the acquisition unit 702 is further configured to sequentially perform fuzzy noise, sharpening, random brightness and contrast, random inversion, and image splicing on the initial data set to obtain the training data set.

[0108] In another example embodiment, the training unit 703 is further configured to divide the training data set based on a preset division ratio to obtain a corresponding training set, a validation set and a test set; and train the initial improved model based on the training set, the validation set and the test set according to a preset training batch and a training round to obtain the target improved model; wherein the division ratio is 4:1:1, the training batch is 16, and the training round is 300.

[0109] In another example embodiment, the training unit 703 is further configured to use the target Cut_SimAM attention mechanism to cut an input feature map input into the target Cut_SimAM attention mechanism into four cut patches; perform mean subtraction processing and normalization processing on each cut patch, and activate the processed cut patch through a Sigmoid function to obtain an activated patch; obtain a first feature matrix of the cut patch and a second feature matrix of the activated patch, and perform a direct product of the second feature matrix and the second feature matrix to obtain a target patch; and perform feature splicing on the four target patches to obtain an output feature map of the head network in the target improved model.

[0110] In another example embodiment, the C2f_EFAttention includes an EFAttention module; and the training unit 703 is further configured to use the C2f_EFAttention to perform one convolution on an input of the C2f_EFAttention to obtain an intermediate feature map; split the intermediate feature map to obtain a first part feature map and a second part feature map; perform feature enhancement processing on the second part feature map through the EFAttention module, and perform feature splicing on the processed second part feature map and the first part feature map to obtain an output of the C2f_EFAttention module.

[0111] In another example embodiment, the system further includes:

[0112] The evaluation unit is configured to obtain an evaluation index of the target improved model, the evaluation index including a detection accuracy evaluation index, a detection efficiency evaluation index and a lightweight evaluation index; and obtain a model performance of the target improved model based on the detection accuracy evaluation index, the detection efficiency evaluation index and the lightweight evaluation index.

[0113] It should be noted that the road hidden disease identification system of the three-dimensional radar map provided in the above embodiments and the road hidden disease identification method of the three-dimensional radar map provided in the above embodiments belong to the same concept, wherein the specific manner in which each module and unit performs operations has been described in detail in the method embodiments, and will not be described here. The road hidden disease identification device of the three-dimensional radar map provided in the above embodiments can be used in actual application, and the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above, and this is not limited herein.

[0114] Embodiments of the present application also provide an electronic device, comprising: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the electronic device implements the road hidden disease identification method of the three-dimensional radar map provided in each of the above embodiments.

[0115] Figure 8 The structure of the computer system of the electronic device suitable for implementing the embodiments of the present application is shown. It should be noted that, Figure 8 The computer system 800 of the electronic device shown is only an example, and should not limit the functions and use range of the embodiments of the present application.

[0116] As Figure 8 shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 802 or programs loaded from a storage portion 808 into a random access memory (RAM) 803, such as performing the method in the above embodiments. In the RAM 803, various programs and data required for system operation are also stored. The CPU 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0117] The following components are connected to the I / O interface 805: an input part 806 including a keyboard, a mouse, etc.; an output part 807 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), etc., and a speaker, etc.; a storage part 808 including a hard disk, etc.; and a communication part 809 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication part 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as necessary. A removable medium 811 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is attached to the drive 810 as necessary, so that a computer program read out therefrom is installed in the storage part 808 as necessary.

[0118] In particular, according to embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing a computer program for executing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network by the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the central processing unit (CPU) 801, various functions defined in the system of the present application are executed.

[0119] It should be noted that the computer-readable medium in the embodiments shown in the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may, for example, be an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination of the above. More specific examples of computer-readable storage media can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (Compact Disc Read-Only Memory, CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, the computer-readable signal medium can include a data signal propagating in a baseband or as a carrier wave in a propagated data signal, in which the computer-readable computer program is carried. Such a propagated data signal can take various forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer-readable signal medium can also be any computer-readable medium other than the computer-readable storage medium, which can send, propagate or transmit programs for use by or in connection with an instruction execution system, device or component. The computer program contained on the computer-readable medium can be transmitted by any suitable medium, including but not limited to wireless, wired, or the like, or any suitable combination of the above.

[0120] The flowcharts and block diagrams in the drawings illustrate the possible implementation architectures, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In the flowcharts or block diagrams, each block can represent a module, a program segment or a part of code, which contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different order than that shown in the drawings. For example, two blocks represented in succession can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and the combination of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0121] Another aspect of the present application also provides a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the method for identifying road hidden diseases from a three-dimensional radar map as described above. The computer readable storage medium can be included in the electronic device as described in the above embodiments, or can exist separately without being assembled into the electronic device.

[0122] Another aspect of the present application also provides a computer program product or a computer program, which includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to make the computer device perform the method for identifying road hidden diseases from a three-dimensional radar map as provided in the above embodiments.

[0123] The above merely describes the preferred embodiments of the present application, and is not used to limit the present application. Any modification, equivalent replacement or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for identifying hidden road defects using three-dimensional radar images, characterized in that: The method comprises: The initial improved model is constructed based on the YOLOv8 model. The target Cut_SimAM attention mechanism is added to the end of the head network of the initial improved model. The feature extraction module is C2f_EFAttention, the loss function is the SlideLoss sliding weighted function, and the bounding box loss function is the D-IoU bounding box loss function. The expression of the SlideLoss sliding weighted function is: Among them, the threshold μ is the average value of the intersection-union ratio of all sample bounding boxes. Samples greater than μ are positive samples, and samples less than μ are negative samples. Acquire a radar damage map transmitted by a three-dimensional ground penetrating radar, and preprocess the radar damage map to obtain a training data set; Training the initial improved model based on the training data set to obtain a target improved model; Inputting a radar image of the road to be tested into the target improvement model to identify hidden disease information of the road to be tested; The C2f_EFAttention includes an EFAttention module; in the step of training the initial improved model based on the training data set to obtain a target improved model, the method includes: Using the target Cut_SimAM attention mechanism, the input feature map input to the target Cut_SimAM attention mechanism is cut into four cut blocks; performing mean subtraction and normalization processing on each of the clipped image blocks, and activating the processed clipped image blocks through a Sigmoid function to obtain activated image blocks; Obtaining a first characteristic matrix of the cut block and a second characteristic matrix of the activated block, and performing a direct product of the first characteristic matrix and the second characteristic matrix to obtain a target block; Perform feature splicing on the four target blocks to obtain an output feature map of the head network in the target improved model; Using the C2f_EFAttention, convolve the input of the C2f_EFAttention to obtain an intermediate feature map; Splitting the intermediate feature map to obtain a first partial feature map and a second partial feature map; The second part of the feature map is subjected to feature enhancement processing by the EFAttention module, and the processed second part of the feature map and the first part of the feature map are subjected to feature splicing to obtain the output of the C2f_EFAttention module.

2. The method according to claim 1, characterized in that The preprocessing of the radar disease atlas to obtain a training data set includes: Filtering the radar damage map, and labeling and classifying the filtered radar image according to the damage type to obtain an initial data set; the damage types include poor interlayer quality, interlayer moisture, loose interlayer, and loose structure; Perform data enhancement processing on the initial data set to obtain a training data set.

3. The method according to claim 2, characterized in that The performing data enhancement processing on the initial data set to obtain a training data set includes: The initial data set is subjected to blurring noise, sharpening processing, random brightness and contrast, random inversion and image splicing in sequence to obtain a training data set.

4. The method according to claim 1, wherein The step of training the initial improved model based on the training data set to obtain a target improved model includes: Divide the training data set based on a preset division ratio to obtain corresponding training set, validation set and test set; Based on the training set, the validation set and the test set, the initial improved model is trained according to preset training batches and training rounds to obtain a target improved model; wherein, the division ratio is 4:1:1, the training batches are 16, and the training rounds are 300.

5. The method according to claim 1, wherein The method further comprises: Obtaining evaluation indicators of the target improvement model, wherein the evaluation indicators include a detection accuracy evaluation indicator, a detection efficiency evaluation indicator, and a lightweight evaluation indicator; The model performance of the target improved model is obtained based on the detection accuracy evaluation index, the detection efficiency evaluation index and the lightweight evaluation index.

6. The method according to claim 5, characterized in that The detection efficiency evaluation index is the number of detection frames per second, and the lightweight evaluation index is the number of parameters and the amount of calculation of the target improvement model; The detection accuracy evaluation indicators include recall rate, precision rate and multi-category average precision. The expressions of precision rate, recall rate and multi-category average precision are: AP=∫0 1 P r (R c )dR e ; Among them, R e is the recall rate, P r is the precision, mAP is the multi-category average precision, TP is the true positive, FP is the false positive, FN is the false negative, C is the number of disease type categories, C = 4.

7. A three-dimensional radar map road hidden disease identification system, characterized by: include: The model construction unit is used to obtain an initial improved model based on the YOLOv8 model. The target Cut_SimAM attention mechanism is added to the head network end of the initial improved model, the feature extraction module is C2f_EFAttention, the loss function is SlideLoss sliding weighted function, and the bounding box loss function is D-IoU bounding box loss function; the expression of the SlideLoss sliding weighted function is: Among them, the threshold μ is the average value of the intersection-union ratio of all sample bounding boxes. Samples greater than μ are positive samples, and samples less than μ are negative samples. an acquisition unit, configured to acquire a radar damage map transmitted by a three-dimensional ground penetrating radar, and preprocess the radar damage map to obtain a training data set; A training unit, configured to train the initial improved model based on the training data set to obtain a target improved model; an identification unit, configured to input a radar image of the road to be tested into the target improvement model, and identify hidden disease information of the road to be tested; The C2f_EFAttention includes an EFAttention module; the training unit is further used to use the target Cut_SimAM attention mechanism to cut the input feature map of the target Cut_SimAM attention mechanism into four cut blocks; performing mean subtraction and normalization processing on each of the clipped image blocks, and activating the processed clipped image blocks through a Sigmoid function to obtain activated image blocks; Obtaining a first characteristic matrix of the cut block and a second characteristic matrix of the activated block, and performing a direct product of the first characteristic matrix and the second characteristic matrix to obtain a target block; Perform feature splicing on the four target blocks to obtain an output feature map of the head network in the target improved model; Using the C2f_EFAttention, convolve the input of the C2f_EFAttention to obtain an intermediate feature map; Splitting the intermediate feature map to obtain a first partial feature map and a second partial feature map; The second part of the feature map is subjected to feature enhancement processing by the EFAttention module, and the processed second part of the feature map and the first part of the feature map are subjected to feature splicing to obtain the output of the C2f_EFAttention module.

Images