Karst anomaly identification method based on improved Faster R-CNN

Abstract

The invention provides a karst anomaly identification method based on an improved Faster R-CNN. The karst anomaly identification method comprises the following steps: step 1, establishing a karst anomaly image data set based on a transverse wave velocity geophysical prospecting image; step 2, an improved Faster R-CNN network is constructed; step 3, the improved Faster R-CNN network is trained; step 4, visualizing a test result; step 5, identifying a karst abnormal area; and step 6, visualizing a karst abnormal region result. According to the method, the improved Faster R-CNN network is constructed to accurately and efficiently identify the karst abnormal region of the transverse wave velocity geophysical prospecting image.

Description

Technical Field

[0001] The present invention relates to a karst anomaly recognition method based on an improved Faster R-CNN, belonging to the technical field of geological engineering exploration. Background Art

[0002] In the field of geological engineering exploration, the identification of karst anomalies is of great significance for assessing the stability of underground geological structures, preventing geological disasters, and developing mineral resources. Currently, the identification of karst anomalies in shear wave velocity maps relies primarily on manual interpretation. However, this method suffers from problems such as strong reliance on expert experience, low interpretation efficiency, and a high risk of misjudgment or omission. In recent years, target detection algorithms have made significant progress in image processing and have gradually been applied to geological exploration. For example, the Faster R-CNN network, a classic deep learning-based target detection algorithm, has demonstrated excellent accuracy and speed in ground-penetrating radar pipeline identification. This model achieves high-precision target recognition through a two-stage detection process and is highly versatile. However, it still suffers from limited small target detection capabilities and insufficient positioning accuracy. Summary of the Invention

[0003] The purpose of the present invention is to overcome the defects of the existing technology and propose a karst anomaly identification method based on an improved Faster R-CNN. The improved Faster R-CNN network is constructed to identify karst anomaly areas in shear wave velocity geophysical images through the improved Faster R-CNN network. The karst anomaly areas identified by the improved Faster R-CNN network are visualized, thereby realizing automatic identification, labeling and visualization of karst anomaly areas based on shear wave velocity geophysical images.

[0004] The object of the present invention is achieved through the following technical solutions: a karst anomaly recognition method based on an improved Faster R-CNN, comprising the following steps: step 1, establishing a karst anomaly image dataset based on shear wave velocity geophysical images; step 2, constructing an improved Faster R-CNN network: changing the feature extraction backbone network of the Faster R-CNN network to a ResNeXt network, setting a CBAM attention mechanism module between the second convolution layer and the batch normalization layer of each defective block of the ResNeXt network, retaining the region proposal network RPN and classification and regression layers of the Faster R-CNN network, adding a feature pyramid network FPN between the ResNeXt network and the region proposal network RPN, and changing the region of interest pooling ROIPooling of the Faster R-CNN network to region of interest alignment ROI Align to establish an improved Faster R-CNN network; step 3, training the improved Faster R-CNN network: training the improved Faster R-CNN network based on the karst anomaly image dataset and outputting a test result, wherein the test result is a test set based on the karst anomaly image dataset, and the improved Faster R-CNN network is used for the training of the improved Faster R-CNN network. The classification and regression results output by the R-CNN network; Step 4, visualizing the test results: input the test results and the shear wave velocity geophysical images corresponding to the test results in the shear wave velocity geophysical image set into OpenCV, draw annotation boxes and labels to the shear wave velocity geophysical images corresponding to the test results in the shear wave velocity geophysical image set to visualize the test results; Step 5, identifying karst anomaly areas: input the shear wave velocity geophysical images of the karst anomaly areas to be identified into the improved Faster R-CNN network that has completed training to obtain the classification and regression results output by the improved Faster R-CNN network, that is, the karst anomaly area results; Step 6, visualizing the karst anomaly area results: input the karst anomaly area results identified in step 5 and the corresponding shear wave velocity geophysical images of the karst anomaly areas to be identified into OpenCV, draw annotation boxes and labels to the corresponding shear wave velocity geophysical images of the karst anomaly areas to be identified, to obtain the karst anomaly area result map.

[0005] In a specific embodiment, a plurality of shear wave velocity geophysical images are obtained by a multi-source frequency domain seismic wave method, and each shear wave velocity geophysical image is preprocessed, each preprocessed shear wave velocity geophysical image is amplified to form a shear wave velocity geophysical image set, and the shear wave velocity geophysical image set is sample labeled, and the labeled shear wave velocity geophysical image set is divided into a training set, a validation set, and a test set in a ratio of 8:1:1 to form a karst anomaly image dataset.

[0006] In a specific embodiment, the sample annotation of the shear wave velocity geophysical image set is specifically as follows: the karst anomaly categories are divided into two categories, the first category is loose and non-dense soil entities, and the second category is karst or broken bedrock or weak interlayers. The loose and non-dense soil entities of each shear wave velocity geophysical image in the shear wave velocity geophysical image set are represented by label S, and the karst or broken bedrock or weak interlayers of each shear wave velocity geophysical image in the shear wave velocity geophysical image set are represented by label Y.

[0007] In a specific embodiment, in step 3, the AdamW optimizer is used to adjust the weights to be trained in the improved Faster R-CNN network, and the weight decay parameter is set to 1×10 -4 , the cosine annealing learning rate scheduler is used to adjust the learning rate.

[0008] In a specific embodiment, the step 3 is specifically as follows: step 3.1, initializing the weights to be trained in the improved Faster R-CNN network, setting the learning rate and batch size; step 3.2, inputting the karst abnormal image dataset into the improved Faster R-CNN network to train it, wherein the training set is used to train the improved Faster R-CNN network to adjust the weights of the improved Faster R-CNN network, the validation set is used to adjust the hyperparameters of the improved Faster R-CNN network and preliminarily evaluate the capabilities of the improved Faster R-CNN network, and the test set is used to finally evaluate the capabilities of the improved Faster R-CNN network.

[0009] In a specific embodiment, the precision rate P, recall rate R, and F1 score F1-score are used to finally evaluate the improved Faster R-CNN network capability; the calculation formula of the precision rate P is: P=TP / (TP+FP), the calculation formula of the recall rate R is: R=TP / (TP+FN), and the calculation formula of the F1 score F1-score is: F1-score=2P*R / (P+R), where TP represents the number of correctly identified targets, FP represents the number of incorrectly identified targets, and FN represents the number of unrecognized targets.

[0010] Compared with the prior art, the beneficial effects of the present invention are as follows: the improved Faster R-CNN network constructed by the present invention uses the ResNeXt network to extract the features of the CBAM attention mechanism of the input shear wave velocity geophysical exploration image, obtains the features of the fusion of low-level detail features and high-level semantic features through the feature pyramid network FPN, and enhances the positioning accuracy of the model through the region of interest calibration ROIAlign. By testing the improved Faster R-CNN network, when identifying loose and non-dense entities in the soil layer, its precision rate is 91.4%, which is 1.7 percentage points higher than that of the Faster R-CNN network, and its recall rate is 89.7%, which is 1.4 percentage points higher than that of the Faster R-CNN network; its F1 score is 90.5%, which is 1.5 percentage points higher than that of the Faster R-CNN network; when identifying karst or bedrock fragments or weak interlayers, its precision rate is 93.2%, which is 1.7 percentage points higher than that of the Faster R-CNN network; its recall rate is 91.8%, which is 1.8 percentage points higher than that of the Faster R-CNN network. The improved Faster R-CNN network achieved an F1 score of 92.5%, 1.7 percentage points higher than the Faster R-CNN network. The improved Faster R-CNN network demonstrated superior performance in identifying karst anomaly regions in shear wave velocity geophysical images. Furthermore, compared to traditional manual interpretation, the improved Faster R-CNN network achieved a 16-fold increase in recognition efficiency. Therefore, the present invention can accurately and efficiently identify karst anomaly regions in shear wave velocity geophysical images, providing effective technical support for the field of geological exploration. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 The figure is a flow chart of the karst anomaly recognition method based on the improved Faster R-CNN of the present invention.

[0012] Figure 2 This is a curve chart of the loss value during the training process of each model.

[0013] Figure 3 This is a curve diagram of the loss value during the validation process of each model.

[0014] Figure 4 This is the geophysical image of shear wave velocity after sample annotation.

[0015] Figure 5 for Figure 4 Visualization of test results based on the improved Faster R-CNN network.

[0016] Figure 6 for Figure 4 Visualization of test results based on the Faster R-CNN network.

[0017] Figure 7 for Figure 4 Visualization of RetinaNet-based test results.

[0018] Figure 8 for Figure 4 Visualization of the test results based on YOLOv8. DETAILED DESCRIPTION

[0019] The present invention will be described in detail below with reference to the embodiments and accompanying drawings. It should be noted that the embodiments and features of the embodiments of the present invention can be combined with each other without conflict.

[0020] refer to Figure 1 The present invention discloses a karst anomaly recognition method based on an improved Faster R-CNN, comprising the following steps: Step 1, establishing a karst anomaly image dataset based on shear wave velocity geophysical images.

[0021] Specifically, multiple shear wave velocity geophysical images are obtained through the multi-source frequency domain seismic wave method, and each shear wave velocity geophysical image is preprocessed. Each preprocessed shear wave velocity geophysical image is amplified to form a shear wave velocity geophysical image set, and samples of the shear wave velocity geophysical image set are annotated. The annotated shear wave velocity geophysical image set is divided into a training set, a validation set, and a test set in a ratio of 8:1:1 to form a karst anomaly image dataset.

[0022] To identify abnormal karst areas in railway sections, multi-source frequency domain seismic wave method is used to obtain shear wave velocity geophysical images because of the interference of tracks.

[0023] In this example, multi-source frequency-domain seismic wave methods were used to conduct karst exploration along the K2250+220-K2255+100 section of the Beijing-Guangzhou Railway to obtain shear wave velocity geophysical images, covering a total of 22 km of survey line. Karst anomalies characterized by relatively low shear wave velocities and closed or concave velocity contours within the rock formation indicate loose, uncompacted soil. Karst anomalies characterized by relatively low shear wave velocities and significant concave velocity contours within the rock formation indicate karst or fractured bedrock or weak interbeds. Six hundred shear wave velocity geophysical images were cropped using a square window sliding technique from the acquired formation shear wave velocity geophysical images.

[0024] The shear wave velocity geophysical images were expanded from 600 to 1600 by using spatial and pixel-level changes, which made the training data diverse and facilitated the training of the model. The spatial-level changes included horizontal mirroring and local amplification, and the pixel-level changes included increasing brightness, increasing contrast, decreasing brightness, and decreasing contrast.

[0025] Combined with the opinions of many experts, LabelImg was used to mark the karst anomaly areas in 1600 shear wave velocity geophysical images. Figure 4 .

[0026] Specifically, the sample annotation of the shear wave velocity geophysical image set is as follows: the karst anomaly categories are divided into two categories, the first category is loose and incomplete soil entities, and the second category is karst or broken bedrock or weak interlayers. The loose and incomplete soil entities of each shear wave velocity geophysical image in the shear wave velocity geophysical image set are represented by label S, and the karst or broken bedrock or weak interlayers of each shear wave velocity geophysical image in the shear wave velocity geophysical image set are represented by label Y.

[0027] The karst anomaly image dataset is saved in VOC format, and the shear wave velocity geophysical image dataset is saved in JPG format.

[0028] The 1600 annotated shear wave velocity geophysical images were divided into training set, validation set and test set in a ratio of 8:1:1 to form a karst anomaly image dataset.

[0029] Step 2. Build an improved Faster R-CNN network: Change the feature extraction backbone network of the Faster R-CNN network to the ResNeXt network, set a CBAM attention mechanism module between the second convolution layer and the batch normalization layer of each defective block of the ResNeXt network, retain the region proposal network RPN and classification and regression layers of the Faster R-CNN network, add a feature pyramid network FPN between the ResNeXt network and the region proposal network RPN, and change the region of interest pooling ROI Pooling of the Faster R-CNN network to region of interest calibration ROI Align to build an improved Faster R-CNN network.

[0030] The improved Faster R-CNN network includes the ResNeXt network, the Feature Pyramid Network (FPN), the Region Proposal Network (RPN), the Region of Interest (ROI) Alignment (ROI Align), and the classification and regression layers. The output of the ResNeXt network is connected to the Feature Pyramid Network (FPN), which in turn is connected to the Region Proposal Network (RPN), which in turn is connected to the Region of Interest (ROI) Alignment (ROI Align), which in turn is connected to the classification and regression layers.

[0031] The second convolutional layer of each incomplete block in the ResNeXt network uses grouped convolution to perform preliminary feature extraction and transformation, resulting in a feature map that contains relatively rich information. Inserting a CBAM attention mechanism module between the second convolutional layer and the batch normalization layer can better capture key channel and spatial information in the feature map, effectively improving the model's focus on the target region and enhancing feature extraction accuracy. Inserting the CBAM attention mechanism module after the first convolutional layer increases its computational load due to the large number of channels in the feature map. Inserting it after the third convolutional layer may miss some characteristic patterns already formed during the grouped convolution. The specific operation is as follows: First, assuming the output feature map of the second convolutional layer is f, f is input into the CBAM attention mechanism module. Channel-wise attention weights are obtained using CAM, and spatial attention weights are obtained using SAM. Next, feature map f is multiplied by the channel-wise attention weights and then by the spatial attention weights to obtain a feature map F weighted by the channel-wise and spatial-wise attention weights. F is then batch normalized.

[0032] A feature pyramid network (FPN) is added between the ResNeXt network and the region proposal network (RPN). The FPN fuses low-level detail features with high-level semantic features to enhance the model's small object detection capabilities. Specifically, the multi-scale feature maps C2, C3, C4, and C5 extracted by the feature extraction network are input into the FPN. Through lateral connections (1×1 convolution to adjust the number of channels) and a top-down path (upsampling + feature map addition), multi-scale feature maps P2, P3, P4, and P5 are generated. P2 is used to detect small objects, and P5 is used to detect large objects. Subsequently, the multi-scale feature maps P2, P3, P4, and P5 output by the FPN are input into the RPN to generate candidate regions, and the final candidate regions are screened through non-maximum suppression.

[0033] Region of interest calibration (ROI Align) avoids quantization errors through bilinear interpolation to enhance the positioning accuracy of the model.

[0034] Step 3: Train and improve the Faster R-CNN network: Train and improve the Faster R-CNN network based on the karst anomaly image dataset and output test results. The test results are the classification and regression results output by the improved Faster R-CNN network based on the test set of the karst anomaly image dataset.

[0035] Specifically, in step 3, the AdamW optimizer is used to adjust and improve the weights to be trained in the Faster R-CNN network, and the weight decay parameter is set to 1×10 -4, which can improve the convergence speed and generalization ability of the improved Faster R-CNN network. The cosine annealing learning rate scheduler is used to adjust the learning rate, which helps the improved Faster R-CNN network to learn quickly in the early stage of training and make fine adjustments in the later stage of training, thereby improving the convergence speed and final performance of the improved Faster R-CNN network.

[0036] The formula for adjusting the learning rate of the cosine annealing learning rate scheduler is as follows: , where: η t is the learning rate of the tth cycle during training, η min is the minimum value of the learning rate, η max is the maximum learning rate, T cur is the index of the current training cycle, T max Represents the length of a complete cosine annealing cycle.

[0037] Specifically, the step 3 is as follows: Step 3.1, initialize the weights to be trained in the improved Faster R-CNN network, and set the learning rate and batch size.

[0038] Step 3.2: Input the karst anomaly image dataset into the improved Faster R-CNN network for training. The training set is used to train the improved Faster R-CNN network to adjust the weights of the improved Faster R-CNN network. The validation set is used to validate the improved Faster R-CNN network to adjust the hyperparameters of the improved Faster R-CNN network and preliminarily evaluate the capabilities of the improved Faster R-CNN network. The test set is used to test the improved Faster R-CNN network to ultimately evaluate the capabilities of the improved Faster R-CNN network, namely, generalization ability, reliability, and practicality.

[0039] During the training of the improved Faster R-CNN network, the loss value of each training process and the loss value of each verification process are recorded in real time. When the loss value of the training process and the loss value of the verification process converge, the improved Faster R-CNN network weights are saved, and the test set is input to test the improved Faster R-CNN network to finally evaluate the improved Faster R-CNN network capabilities.

[0040] Specifically, the precision rate P, recall rate R, and F1 score F1-score are used to finally evaluate the improved Faster R-CNN network capabilities; the calculation formula of the precision rate P is: P=TP / (TP+FP), the calculation formula of the recall rate R is: R=TP / (TP+FN), and the calculation formula of the F1 score F1-score is: F1-score=2P*R / (P+R).

[0041] Among them, TP represents the number of targets that are correctly identified, FP represents the number of targets that are incorrectly identified, and FN represents the number of targets that are not identified.

[0042] Precision indicates the proportion of samples predicted to be positive that are actually positive, recall indicates the proportion of samples that are correctly predicted to be positive, and the F1 score is the harmonic average of precision and recall, balancing the performance of the two.

[0043] Step 4. Visualize the test results: Input the test results and the shear wave velocity geophysical images corresponding to the test results in the shear wave velocity geophysical image set into OpenCV, draw annotation boxes and labels to the shear wave velocity geophysical images corresponding to the test results in the shear wave velocity geophysical image set to visualize the test results.

[0044] Because the classification and regression results output by the improved Faster R-CNN network are numerical data, including bounding box coordinates and labels, visualization is often required in practical applications for easy viewing and statistical analysis. OpenCV (Open Source Computer Vision Library) is a cross-platform computer vision library that implements many common algorithms in image processing and computer vision. Therefore, OpenCV was used to visualize the classification and regression results output by the improved Faster R-CNN network.

[0045] In this embodiment, a karst anomaly image dataset including 1,600 shear wave velocity geophysical images is input into the improved Faster R-CNN network, the Faster R-CNN network, the RetinaNet, and the YOLOv8, respectively, to train the improved Faster R-CNN network, the Faster R-CNN network, the RetinaNet, and the YOLOv8, and output test results.

[0046] The initial batch size and learning rate settings for each model were determined by the grid search method. The batch size and learning rate of the improved Faster R-CNN network were 16 and 0.001, respectively; the batch size and learning rate of the Faster R-CNN network were 8 and 0.001, respectively; the batch size and learning rate of RetinaNet were 16 and 0.001, respectively; and the batch size and learning rate of YOLOv8 were 16 and 0.01, respectively.

[0047] During the training of each model, the loss value of each training process and the loss value of each verification process are recorded in real time. When the loss value of the training process and the loss value of the verification process converge, the weight of the corresponding model is saved, and the test set is input into each model respectively to finally evaluate the ability of each model.

[0048] See Figure 2 and Figure 3 After 100 iterations, the loss values ​​of each model in the training process and the loss values ​​in the validation process converged. The loss values ​​of the improved Faster R-CNN network, Faster R-CNN network, RetinaNet, and YOLOv8 in the training process were 0.063, 0.081, 0.089, and 0.114, respectively, and the loss values ​​in the validation process were 0.056, 0.072, 0.083, and 0.113, respectively. The difference between the loss values ​​of each model in the training process and the validation process was very small, indicating that none of the models had overfitting. The two loss values ​​of the improved Faster R-CNN were lower than those of the other three models, indicating that the improved Faster R-CNN had the highest accuracy in identifying karst anomaly areas in shear wave velocity geophysical maps on both the training and validation sets.

[0049] When identifying loose and incomplete entities in soil layers, the improved Faster R-CNN network has a precision of 91.4%, which is 1.7, 4.7, and 7.7 percentage points higher than the Faster R-CNN network, RetinaNet, and YOLOv8, respectively; the recall rate is 89.7%, which is 1.4, 4.5, and 7.6 percentage points higher than the Faster R-CNN network, RetinaNet, and YOLOv8, respectively; the F1 score is 90.5%, which is 1.5, 4.6, and 7.6 percentage points higher than the Faster R-CNN network, RetinaNet, and YOLOv8, respectively; when identifying karst or bedrock fragments or weak interlayers, the improved Faster R-CNN has a precision of 93.2%, which is 1.7, 3.9, and 6.5 percentage points higher than the Faster R-CNN network, RetinaNet, and YOLOv8, respectively; the recall rate is 91.8%, which is 1. The improved Faster R-CNN network outperformed the RetinaNet network by 1.7, 3.7, and 6.5 percentage points, respectively. Its F1 score reached 92.5%, surpassing the Faster R-CNN network, RetinaNet, and YOLOv8 by 1.7, 3.8, and 6.5 percentage points, respectively. These analyses demonstrate that the improved Faster R-CNN performs best when identifying karst anomaly areas on the test set.

[0050] The test results obtained by each model and the shear wave velocity geophysical images corresponding to the test results in the shear wave velocity geophysical image set containing 1600 shear wave velocity geophysical images are input into OpenCV, and annotation boxes and labels are drawn on the shear wave velocity geophysical images corresponding to the test results in the shear wave velocity geophysical image set to obtain a visualization of the test results based on each model. Figures 4 to 8 ,from Figure 4 It can be seen that the marked soil layer is loose and not dense, and there are two places with karst or broken bedrock or weak interlayers. Figure 5 In the middle, there is one loose and loose soil layer, and two karst or broken bedrock or weak interlayers. The identification is completely correct. Figure 6 There are two loose and incomplete soil entities, and two karst or broken bedrock or weak interlayers. One of the loose and incomplete soil entities is mislabeled. Figure 7 There is one loose and incomplete soil entity and two karst or broken bedrock or weak interlayers. Among them, the loose and incomplete soil entity is incorrectly identified. Figure 8 There are two places with karst or broken bedrock or weak interlayers. One of them has a smaller identification area and the loose and incomplete soil layer is not identified.

[0051] Compared to other models, the improved Faster R-CNN network significantly reduces missed and misidentified karst anomalies, and is more consistent with expert annotations. Using the improved Faster R-CNN network improves recognition efficiency by 16 times compared to traditional manual recognition.

[0052] Step 5: Identify karst anomalies: Input the shear wave velocity geophysical exploration image that needs to identify karst anomalies into the trained improved Faster R-CNN network to obtain the classification and regression results output by the improved Faster R-CNN network, that is, the karst anomaly area results.

[0053] The shear wave velocity geophysical image is input into the ResNeXt network to extract features that introduce the CBAM attention mechanism. The features extracted by the ResNeXt network are input into the feature pyramid network FPN to output features that integrate low-level detail features and high-level semantic features. The features obtained by the feature pyramid network FPN are input into the region proposal network RPN to generate candidate regions. The candidate regions generated by the region proposal network RPN are input into the region of interest calibration ROI Align. The features obtained from the region of interest calibration ROI Align are input into the classification and regression layer to determine whether the candidate region belongs to the target category and to fine-tune the position of the candidate region to obtain the final target detection result, that is, the karst abnormal area result.

[0054] Step 6: Visualize the results of the karst anomaly area: Input the results of the karst anomaly area identified in step 5 and the corresponding shear wave velocity geophysical images of the karst anomaly area that needs to be identified into OpenCV, draw annotation boxes and labels to the corresponding shear wave velocity geophysical images of the karst anomaly area that needs to be identified to obtain the karst anomaly area result map.

[0055] The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, several simple deductions and substitutions can be made without departing from the concept of the present invention, and all of these should be considered to fall within the scope of protection of the present invention.

Claims

1. A karst anomaly recognition method based on improved Faster R-CNN, characterized in that: The following steps are involved: Step 1: Establish a karst anomaly image dataset based on shear wave velocity geophysical images; Step 2. Build an improved Faster R-CNN network: Change the feature extraction backbone network of the Faster R-CNN network to the ResNeXt network, set the CBAM attention mechanism module between the second convolution layer and the batch normalization layer of each defective block of the ResNeXt network, retain the region proposal network RPN and classification and regression layers of the Faster R-CNN network, add a feature pyramid network FPN between the ResNeXt network and the region proposal network RPN, and change the region of interest pooling ROI Pooling of the Faster R-CNN network to region of interest alignment ROI Align to build an improved Faster R-CNN network; Step 3: Train and improve the Faster R-CNN network: Train and improve the Faster R-CNN network based on the karst anomaly image dataset and output test results. The test results are the classification and regression results output by the improved Faster R-CNN network on the test set of the karst anomaly image dataset. Step 4: Visualize the test results: Input the test results and the shear wave velocity geophysical images corresponding to the test results in the shear wave velocity geophysical image set into OpenCV, draw annotation boxes and labels on the shear wave velocity geophysical images corresponding to the test results in the shear wave velocity geophysical image set to visualize the test results; Step 5: Identify karst anomaly areas: Input the shear wave velocity geophysical exploration image of the karst anomaly area to be identified into the trained improved Faster R-CNN network to obtain the classification and regression results output by the improved Faster R-CNN network, i.e., the karst anomaly area results; Step 6. Visualize the results of the karst anomaly area: Input the results of the karst anomaly area identified in step 5 and the corresponding shear wave velocity geophysical images of the karst anomaly area that needs to be identified into OpenCV, draw annotation boxes and labels to the corresponding shear wave velocity geophysical images of the karst anomaly area that needs to be identified to obtain the karst anomaly area result map.

2. The karst anomaly identification method based on improved Faster R-CNN according to claim 1, characterized in that: Multiple shear wave velocity geophysical images were obtained by the multi-source frequency domain seismic wave method, and each shear wave velocity geophysical image was preprocessed. Each preprocessed shear wave velocity geophysical image was amplified to form a shear wave velocity geophysical image set. Samples of the shear wave velocity geophysical image set were annotated, and the annotated shear wave velocity geophysical image set was divided into a training set, a validation set, and a test set in the ratio of 8:1:1 to form a karst anomaly image dataset.

3. The karst anomaly identification method based on improved Faster R-CNN according to claim 2, characterized in that: The sample annotation of the shear wave velocity geophysical image set is specifically as follows: the karst anomaly categories are divided into two categories, the first category is loose and non-dense soil entities, and the second category is karst or broken bedrock or weak interlayers. The loose and non-dense soil entities of each shear wave velocity geophysical image in the shear wave velocity geophysical image set are represented by label S, and the karst or broken bedrock or weak interlayers of each shear wave velocity geophysical image in the shear wave velocity geophysical image set are represented by label Y.

4. The karst anomaly identification method based on improved Faster R-CNN according to claim 3, characterized in that: In step 3, the AdamW optimizer is used to adjust and improve the weights to be trained in the Faster R-CNN network, and the weight decay parameter is set to 1×10 -4 , the cosine annealing learning rate scheduler is used to adjust the learning rate.

5. The karst anomaly identification method based on improved Faster R-CNN according to claim 4, characterized in that: The step 3 is specifically as follows: Step 3.1: Initialize the weights to be trained in the improved Faster R-CNN network and set the learning rate and batch size. Step 3.2: Input the karst anomaly image dataset into the improved Faster R-CNN network for training. The training set is used to train the improved Faster R-CNN network to adjust the weights of the improved Faster R-CNN network. The validation set is used to adjust the hyperparameters of the improved Faster R-CNN network and preliminarily evaluate the capabilities of the improved Faster R-CNN network. The test set is used to finally evaluate the capabilities of the improved Faster R-CNN network.

6. The karst anomaly identification method based on improved Faster R-CNN according to claim 5, characterized in that: The precision rate P, recall rate R, and F1 score F1-score are used to finally evaluate the improved Faster R-CNN network capabilities; the calculation formula of the precision rate P is: P=TP / (TP+FP), the calculation formula of the recall rate R is: R=TP / (TP+FN), and the calculation formula of the F1 score F1-score is: F1-score=2P*R / (P+R), among which TP represents the number of correctly identified targets, FP represents the number of incorrectly identified targets, and FN represents the number of unrecognized targets.

Images