Railway subgrade layer position intelligent identification method, device, equipment, medium and product

Abstract

The present application provides a railway subgrade horizon intelligent identification method, device, equipment, medium and product. The method comprises longitudinally cutting and sampling a radar image of a railway subgrade to obtain multiple horizontal slice images of the radar image; the multiple horizontal slice images of the radar image are subjected to horizon detection and labeling through a preset target detection model to obtain multiple horizontal slice images of the radar image after labeling and a labeling text file, the labeling text file comprising horizon labeling texts corresponding to the multiple horizontal slice images of the radar image after labeling; horizon point coordinates corresponding to positions of the multiple horizontal slice images of the radar image are determined according to the horizon labeling texts in the labeling text file; and the determined horizon point coordinates corresponding to the positions of the multiple horizontal slice images of the radar image are spliced to obtain a horizon line of the railway subgrade. The present method greatly reduces the manual workload and is more accurate than the results of manual tracking.

Description

Technical Field

[0001] This invention relates to the field of railway subgrade detection and evaluation technology, and in particular to a method, device, equipment, medium and product for intelligent identification of railway subgrade location. Background Technology

[0002] Railway subgrades are typically layered structures, and the thickness and morphology of each layer are important parameters reflecting the condition of the subgrade. Current subgrade structure inspection mainly uses ground-penetrating radar technology to obtain radar images of the subgrade, and then manually traces the layers to obtain the layer lines.

[0003] Ground penetrating radar (GPR) technology works by emitting electromagnetic waves into the ground. These waves propagate differently in different media and are reflected at the interfaces between them. The radar's receiving antenna receives the reflected signals, which are then processed to produce a typical radar image. A radar image is a two-dimensional grayscale matrix, with each column representing one data point. Figure 1 As shown, the amplitude of a radar data point is reflected in the image as the grayscale value of pixels at different depths.

[0004] Manual layer tracing involves manually tracking the continuous reflection phase axes of the roadbed structure on the cross-section of a two-dimensional radar image based on waveform similarity, followed by interpolation to obtain the layer lines. This process is labor-intensive. Furthermore, because manual layer tracing requires both visual observation and mouse operation, it is difficult to achieve highly precise tracking. Summary of the Invention

[0005] To address the technical problems of high manual workload and low accuracy in current manual layer tracking, this invention proposes a method, device, equipment, medium, and product for intelligent identification of railway subgrade locations.

[0006] In a first aspect, embodiments of the present invention provide a method for intelligent identification of railway subgrade locations, comprising:

[0007] Acquire radar images of the railway subgrade;

[0008] The radar image of the railway subgrade is longitudinally segmented and sampled to obtain multiple horizontal slice images of the radar image; the longitudinal direction of the radar image of the railway subgrade represents the depth of the railway subgrade, and the horizontal direction of the radar image of the railway subgrade represents the line mileage of the railway subgrade.

[0009] Multiple horizontal slices of the radar image are subjected to layer detection and annotation using a preset target detection model to obtain multiple horizontal slices of the annotated radar image and an annotation text file. The annotation text file includes the layer annotation text corresponding to the multiple horizontal slices of the annotated radar image.

[0010] Based on the layer labeling text in the labeled text file, determine the layer coordinates of the corresponding positions in multiple horizontal slice images of the radar image;

[0011] The layer coordinates of the corresponding positions in multiple transverse slices of the determined radar image are stitched together to obtain the layer line of the railway subgrade.

[0012] In some implementations, the step of performing layer detection and annotation on multiple horizontal slices of the radar image using a preset target detection model to obtain annotated multiple horizontal slices of the radar image and an annotation text file includes:

[0013] Name the multiple horizontal slice images of the radar image;

[0014] The radar image with the file name is subjected to layer detection and annotation through a preset target detection model to obtain the radar image with multiple horizontal slice images and annotation text file after annotation.

[0015] The step of determining the layer coordinates of corresponding positions in multiple horizontal slice images of the radar image based on the layer annotation text in the annotation text file includes:

[0016] The annotation text file is traversed according to the filenames of the multiple transverse slice images of the radar image. If the corresponding layer annotation text is found in the annotation text file, the layer annotation text is read to obtain the layer coordinates of the corresponding position in the transverse slice image of the radar image. If the corresponding layer annotation text is not found in the annotation text file, the layer coordinates of the corresponding position in the transverse slice image of the radar image are set to zero.

[0017] In some implementations, the radar image of the railway subgrade is longitudinally and uniformly segmented and sampled to obtain multiple transverse slice images of the radar image.

[0018] In some implementations, the radar image of the railway subgrade is longitudinally and uniformly segmented and sampled to obtain multiple transverse slice images of the radar image, including:

[0019] The distance of each sampling slide is determined based on the preset number of sampling points and the preset aspect ratio of the horizontal slice image;

[0020] Based on the determined distance of each sampling slide, the radar image is slid laterally along the radar image and the radar image of the railway subgrade is longitudinally segmented and sampled until the segmentation and sampling are completed, resulting in a horizontal slice image of multiple radar images.

[0021] In some implementations, the construction of the target detection model includes:

[0022] Obtain radar image samples of railway subgrade, wherein the radar image samples of railway subgrade include radar images of railway subgrade or laterally segmented images of radar images of railway subgrade;

[0023] Obtain the layer label text corresponding to the radar image sample of the railway subgrade;

[0024] Multiple sets of samples are formed by combining the radar image samples of the railway subgrade and the corresponding layer labeling text of the radar image samples of the railway subgrade. Each set of samples includes a radar image sample of the railway subgrade and the corresponding layer labeling text of the radar image sample of the railway subgrade.

[0025] The multiple sets of samples are divided into a training set and a test set;

[0026] The target detection model is obtained by training the model based on the training set and the test set.

[0027] In some implementations, multiple horizontal slices of the radar image are named digitally.

[0028] In a second aspect, embodiments of the present invention provide a railway subgrade location intelligent identification device, comprising:

[0029] The acquisition module is used to acquire radar images of the railway subgrade;

[0030] The sampling module is used to perform longitudinal segmentation sampling on the radar image of the railway subgrade to obtain multiple horizontal slice images of the radar image; the longitudinal direction of the radar image of the railway subgrade represents the depth of the railway subgrade, and the horizontal direction of the radar image of the railway subgrade represents the line mileage of the railway subgrade.

[0031] The layer detection module is used to perform layer detection and annotation on multiple horizontal slice images of the radar image through a preset target detection model, so as to obtain multiple horizontal slice images of the annotated radar image and an annotation text file. The annotation text file includes the layer annotation text corresponding to the multiple horizontal slice images of the annotated radar image.

[0032] The identification module is used to determine the coordinates of the layer positions corresponding to the positions in multiple horizontal slice images of the radar image based on the layer labeling text in the labeled text file.

[0033] The stitching module is used to stitch together the layer coordinates of corresponding positions in multiple transverse slices of the determined radar image to obtain the layer line of the railway subgrade.

[0034] Thirdly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by at least one processor, implements the method described in the first aspect.

[0035] Fourthly, embodiments of the present invention provide a computer program product that executes the method described in the first aspect when running on a processor.

[0036] Fifthly, embodiments of the present invention provide an electronic device, including a memory and at least one processor, wherein the memory stores a computer program, and the computer program, when executed by the at least one processor, implements the method as described in the first aspect.

[0037] One or more embodiments of the present invention provide at least the following beneficial effects:

[0038] This invention constructs a target detection model to perform layer detection and annotation on radar images of railway subgrades, automatically identifying the coordinates of layer points at each layer of the railway subgrade. The identified layer point coordinates are then stitched together to obtain the layer lines of the railway subgrade. This significantly reduces manual workload and provides more precise results than manual tracking. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope.

[0040] Figure 1 A radar image schematic diagram of a railway subgrade provided by the present invention;

[0041] Figure 2 This is a schematic diagram of the results of manual layer tracing provided by the present invention;

[0042] Figure 3 This is a schematic diagram of the final result of the stratum under the ideal state provided by the present invention;

[0043] Figure 4 A comparison chart of the results of manual and ideal stratigraphic tracking provided by this invention;

[0044] Figure 5 A flowchart of a railway subgrade location intelligent identification method provided in an embodiment of the present invention;

[0045] Figure 6 A schematic diagram of a horizontal slice of a radar image of a railway subgrade provided in an embodiment of the present invention;

[0046] Figure 7 A flowchart of a railway subgrade location intelligent identification method provided in another embodiment of the present invention;

[0047] Figure 8 This is a schematic diagram of the output results of the target detection model provided in an embodiment of the present invention;

[0048] Figure 9 This is a visualization result of the stitched layer lines provided in an embodiment of the present invention;

[0049] Figure 10 This is a comparison chart of the target detection model recognition results trained using the YOLOv5 model and the results of manual tracking provided in an embodiment of the present invention.

[0050] Figure 11 This is a block diagram of a railway subgrade location intelligent identification device provided in an embodiment of the present invention. Detailed Implementation

[0051] The technical solutions of the embodiments of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0052] Manual stratigraphic tracking involves manually tracing the continuous reflection phase axes of the roadbed structure on the cross-section of a two-dimensional radar image based on waveform similarity, followed by interpolation to obtain stratigraphic lines. This process is labor-intensive. The results of manual stratigraphic tracking are as follows: Figure 2 As shown. The ideal goal of layer tracking is to find a feature point, such as a peak, trough, or zero-crossing point, for each radar data track in the radar image, forming a layer line. That is, ideal layer tracking should be the peak or trough point at an appropriate location in each data track. The ideal effect of layer tracking is as follows: Figure 3 As shown in the figure. The comparison between the results of manual and ideal stratum tracking is as follows. Figure 4 As shown, Figure 4 It is known that the final accuracy of manual stratification is low because manual stratification tracking requires visual observation and mouse operation, making it difficult to track very precisely.

[0053] Example 1

[0054] like Figure 5 As shown, this embodiment provides a method for intelligent identification of railway subgrade locations, including:

[0055] Step S510: Obtain radar images of the railway subgrade.

[0056] Radar images are two-dimensional grayscale matrices, with each column representing one data point. The amplitude of a radar data point is reflected in the image as the grayscale value of pixels at different depths.

[0057] Step S520: The radar image of the railway subgrade is longitudinally segmented and sampled to obtain multiple horizontal slice images of the radar image; the longitudinal direction of the radar image of the railway subgrade represents the depth of the railway subgrade, and the horizontal direction of the radar image of the railway subgrade represents the line mileage of the railway subgrade.

[0058] In this embodiment, it is preferable to perform longitudinal uniform segmentation sampling on the radar image of the railway subgrade to obtain multiple transverse slice images of the radar image. Specifically, this may include:

[0059] Step S520a: Determine the distance of each sampling slide based on the preset number of sampling points and the preset aspect ratio of the horizontal slice image.

[0060] Step S520b: Based on the determined distance of each sampling slide, slide horizontally along the radar image and perform longitudinal segmentation sampling on the radar image of the railway subgrade until the segmentation sampling is completed, to obtain horizontal slice images of multiple radar images.

[0061] Preferably, adjacent horizontal slice images repeat by 50%.

[0062] For example, such as Figure 6 As shown, each slide involves 100 lines, and 2000 lines are captured as a horizontal slice image.

[0063] Step S530: Multiple horizontal slice images of the radar image are subjected to layer detection and annotation using a preset target detection model to obtain multiple horizontal slice images of the annotated radar image and an annotation text file. The annotation text file includes the layer annotation text corresponding to the multiple horizontal slice images of the annotated radar image.

[0064] The target detection model identifies the positions of peaks and troughs in multiple horizontal slices of the radar image and uses bounding boxes to mark the "chengwei" and the confidence level of the layer.

[0065] In this embodiment, the construction of the target detection model includes:

[0066] Step S530a: Obtain radar image samples of the railway subgrade, wherein the radar image samples of the railway subgrade include radar images of the railway subgrade or laterally segmented images of radar images of the railway subgrade.

[0067] Step S530b: Obtain the layer label text corresponding to the radar image sample of the railway subgrade.

[0068] For example, layer label text can be created using image annotation tools, such as labelimg software.

[0069] Step S530c: The radar image samples of the railway subgrade and the corresponding layer labeling text of the radar image samples of the railway subgrade are formed into multiple sets of samples, wherein each set of samples includes a radar image sample of the railway subgrade and the corresponding layer labeling text of the radar image sample of the railway subgrade.

[0070] Step S530d: Divide the multiple sets of samples into a training set and a test set.

[0071] Step S530e: Train the model based on the training set and the test set to obtain the target detection model.

[0072] For example, based on the training set and test set, the target detection model is obtained by training the model using the YOLO model, including:

[0073] The first step is to put the weight file used for training the YOLO model and the resulting sets of samples into the project folder.

[0074] The second step is to write a YAML file (YAML is a format for expressing data serialization), set the file paths for the training set, validation set, and test set, and set the target categories;

[0075] The third step is to modify the path of the weight file of the training function (the training function is a machine learning function used to build a model) and the path of the corresponding yaml file, and at the same time import the prepared yaml file, set the number of training iterations and set the image processor (GPU) for training.

[0076] The fourth step is to use the `python train.py` command to run the training function, and train the YOLO v5s model and the YOLO v5l model respectively to obtain the object detection model.

[0077] Step S540: Based on the layer annotation text in the annotation text file, determine the layer coordinates of the corresponding positions in multiple horizontal slices of the radar image. Store the determined layer coordinates of the corresponding positions in the multiple horizontal slices of the radar image into a text file.

[0078] Step S550: The layer coordinates of the corresponding positions of multiple transverse slices of the radar image are stitched together to obtain the layer line of the railway subgrade.

[0079] This embodiment utilizes a target detection model to perform layer detection and annotation on radar images of the railway subgrade, automatically identifying the coordinates of layer points at each layer of the railway subgrade. The identified layer point coordinates are then stitched together to obtain the layer lines of the railway subgrade. This significantly reduces manual workload and provides more precise results than manual tracking.

[0080] Example 2

[0081] like Figure 7 As shown, this embodiment provides a method for intelligent identification of railway subgrade locations, including:

[0082] Step S710: Obtain radar images of the railway subgrade.

[0083] In a specific example, the radar image is a grayscale image of a two-dimensional matrix, with each column representing one data point. The amplitude of one radar data point is reflected in the image as pixel values ​​at different depths.

[0084] Step S720: Vertically segment and sample the radar image of the railway subgrade to obtain multiple horizontal slice images of the radar image; the vertical direction of the radar image of the railway subgrade represents the depth of the railway subgrade, and the horizontal direction of the radar image sample of the railway subgrade represents the line mileage of the railway subgrade.

[0085] In this embodiment, it is preferable to perform longitudinal uniform segmentation sampling on the radar image of the railway subgrade to obtain multiple transverse slice images of the radar image. Specifically, this may include:

[0086] Step S720a: Determine the distance of each sampling slide based on the preset number of sampling points and the preset aspect ratio of the horizontal slice image.

[0087] Step S720b: Based on the determined distance of each sampling slide, slide horizontally along the radar image and perform longitudinal segmentation sampling on the radar image of the railway subgrade until the segmentation sampling is completed, to obtain horizontal slice images of multiple radar images.

[0088] Preferably, adjacent horizontal slice images repeat by 50%.

[0089] For example, such as Figure 6 As shown, each slide involves 100 lines, and 2000 lines are captured as a horizontal slice image.

[0090] Step S730: Name the files of multiple horizontal slice images of the radar image.

[0091] In some implementations, multiple horizontal slices of the radar image are named digitally. This facilitates subsequent traversal of the labeled text file to locate horizontal slices that have not been detected by the target detection model and perform special processing.

[0092] For example, the filenames of horizontally sliced ​​images can be test0, test1, test2, etc.

[0093] Step S740: The multiple horizontal slice images of the radar image after file naming are subjected to layer detection and annotation through a preset target detection model to obtain multiple horizontal slice images of the annotated radar image and an annotation text file. The annotation text file includes the layer annotation text corresponding to the multiple horizontal slice images of the annotated radar image.

[0094] For example, such as Figure 8 As shown, the labels file is the output labeled text file, containing layer labeling text for the horizontal slice images of peaks and troughs identified by the target detection model. Images such as test0.png, test1.png, and test2.png are output horizontal slice images. The target detection model identifies and marks the positions of peaks and troughs in multiple horizontal slice images of the radar image using bounding boxes, and marks "chengwei" and the layer confidence score next to the bounding boxes. In this case, the labels file contains the corresponding layer labeling text. For example, in the horizontal slice image test1.png, the bounding box is marked, and "chengwei" is marked, and the layer confidence score is 0.76. The labels file contains the corresponding layer labeling text test1.txt. If the output horizontal slice image does not have a bounding box, "chengwei", or layer confidence score, it means that it was not detected by the target detection model, as in the horizontal slice image test6.png.

[0095] In this embodiment, the construction of the target detection model includes:

[0096] Step S740a: Obtain radar image samples of the railway subgrade, wherein the radar image samples of the railway subgrade include radar images of the railway subgrade or laterally segmented images of radar images of the railway subgrade.

[0097] Step S740b: Obtain the layer label text corresponding to the radar image sample of the railway subgrade.

[0098] For example, layer label text can be created using image annotation tools, such as labelimg software.

[0099] Step S740c: The radar image samples of the railway subgrade and the corresponding layer labeling text of the radar image samples of the railway subgrade are formed into multiple sets of samples, wherein each set of samples includes a radar image sample of the railway subgrade and the corresponding layer labeling text of the radar image sample of the railway subgrade.

[0100] Step S740d: Divide the multiple sets of samples into a training set and a test set.

[0101] Step S740e: Train the model based on the training set and the test set to obtain the target detection model.

[0102] For example, based on the training set and test set, the target detection model is obtained by training the model using the YOLO model, including:

[0103] The first step is to put the weight file used for training the YOLO model and the resulting sets of samples into the project folder.

[0104] The second step is to write a YAML file, setting the file paths for the training set, validation set, and test set, and specifying the target categories;

[0105] The third step is to modify the path of the training function weight file and the path of the corresponding yaml file, import the prepared yaml file, set the number of training iterations, and set the image processor (GPU) for training.

[0106] The fourth step is to use the `python train.py` command to run the training function, and train the YOLO v5s model and the YOLO v5l model respectively to obtain the object detection model.

[0107] Step S750: Traverse the annotation text file according to the filenames of the multiple horizontal slice images of the radar image. If a corresponding layer annotation text is found in the annotation text file, read the layer annotation text to obtain the layer coordinates of the corresponding position in the horizontal slice image of the radar image. If no corresponding layer annotation text is found in the annotation text file, set the layer coordinates of the corresponding position in the horizontal slice image of the radar image to zero. Store the determined layer coordinates of the multiple horizontal slice images of the radar image into a text file.

[0108] Step S760: The layer coordinates of the corresponding positions of multiple transverse slices of the radar image are stitched together to obtain the layer line of the railway subgrade.

[0109] The visualization result of the stitched stratigraphic lines is as follows Figure 9 As shown in the figure, the comparison between the object detection model trained with the YOLOv5 model and the results of manual tracking is as follows. Figure 10 As shown.

[0110] In this embodiment, multiple transverse slice images of the radar image of the railway subgrade are named. The annotation text file is traversed according to the filenames of the transverse slice images of the radar image input to the target detection model. If a corresponding layer annotation text is found in the annotation text file, the layer coordinates of the corresponding position in the transverse slice image of the corresponding radar image are read from the layer annotation text. If no corresponding layer annotation text is found in the annotation text file, the layer coordinates of the corresponding position in the transverse slice image of the corresponding radar image are set to zero, thereby automatically obtaining the layer coordinates of the corresponding position in the radar image of the railway subgrade. These obtained layer coordinates are then stitched together to obtain the layer line of the railway subgrade. This significantly reduces the workload and provides more precise results than manual tracking.

[0111] Example 3

[0112] like Figure 11 As shown, this embodiment provides a railway subgrade location intelligent identification device, including:

[0113] The acquisition module 310 is used to acquire radar images of the railway subgrade;

[0114] The sampling module 320 is used to perform longitudinal segmentation sampling on the radar image of the railway subgrade to obtain multiple horizontal slice images of the radar image; the longitudinal direction of the radar image of the railway subgrade is the depth of the railway subgrade, and the horizontal direction of the radar image of the railway subgrade is the line mileage of the railway subgrade.

[0115] The layer detection module 330 is used to perform layer detection and annotation on multiple horizontal slice images of the radar image through a preset target detection model, so as to obtain multiple horizontal slice images of the annotated radar image and an annotation text file. The annotation text file includes layer annotation text corresponding to the multiple horizontal slice images of the annotated radar image.

[0116] The identification module 340 is used to determine the layer position coordinates of multiple horizontal slice images corresponding to the radar image based on the layer labeling text in the labeled text file.

[0117] The stitching module 350 is used to stitch together the layer coordinates of corresponding positions of multiple transverse slices of the radar image to obtain the layer line of the railway subgrade.

[0118] This embodiment utilizes a target detection model to perform layer detection and annotation on radar images of the railway subgrade, automatically identifying the coordinates of layer points at each layer of the railway subgrade. The identified layer point coordinates are then stitched together to obtain the layer lines of the railway subgrade. This significantly reduces manual workload and provides more precise results than manual tracking.

[0119] In another embodiment, the sampling module 320 is also used to file names for multiple horizontal slice images of the radar image.

[0120] The layer detection module 330 is used to perform layer detection and annotation on multiple horizontal slice images of the radar image after file naming through a preset target detection model, so as to obtain multiple horizontal slice images of the annotated radar image and an annotation text file.

[0121] The identification module 340 is used to traverse the annotation text file according to the file names of multiple horizontal slice images of the radar image. If the corresponding layer annotation text is found in the annotation text file, the layer point coordinates of the corresponding position of the horizontal slice image of the radar image are obtained by reading the annotation text. If the corresponding layer annotation text is not found in the annotation text file, the layer point coordinates of the corresponding position of the horizontal slice image of the radar image are set to zero.

[0122] This embodiment names multiple transverse slices of radar images of the railway subgrade. Based on the filenames of these transverse slices input to the target detection model, it iterates through the annotation text file. If a corresponding layer annotation text is found in the annotation text file, the layer coordinates of the corresponding position in the transverse slice of the radar image are obtained by reading the layer annotation text. If no corresponding layer annotation text is found, the layer coordinates of the corresponding position in the transverse slice of the radar image are set to zero, thus automatically obtaining the layer coordinates of the corresponding position in the radar image of the railway subgrade. These obtained layer coordinates are then stitched together to obtain the layer line of the railway subgrade. This significantly reduces manual workload and provides more precise results than manual tracking.

[0123] Example 4

[0124] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by at least one processor, it implements the intelligent identification method for railway subgrade location described in the foregoing embodiment.

[0125] The computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0126] Example 5

[0127] This embodiment provides an electronic device, including a memory and at least one processor. The memory stores a computer program, which, when executed by the at least one processor, implements the method of the aforementioned embodiment.

[0128] The processor can be implemented using an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a controller, a Microcontroller Unit (MCU), a microprocessor, or other electronic components, and is used to execute the intelligent identification method for railway roadbed locations in the above embodiments.

[0129] Example 6

[0130] This embodiment provides a computer program product that executes the railway subgrade location intelligent identification method described in the preceding embodiment when the computer program product is run on a processor.

[0131] In practical applications, computer program products can run on electronic devices.

[0132] In the several embodiments provided in this invention, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus and method embodiments described above are merely illustrative.

[0133] It should be noted that, in this document, the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0134] While the embodiments disclosed in this invention are as described above, the content is merely for the purpose of facilitating understanding of the invention and is not intended to limit the invention. Any person skilled in the art to which this invention pertains may make any modifications and variations in form and detail of the implementation without departing from the spirit and scope disclosed herein; however, the scope of patent protection for this invention shall still be determined by the scope defined in the appended claims.

Claims

1. A method for intelligent identification of railway subgrade location, characterized in that, include: Acquire radar images of the railway subgrade; The radar image of the railway subgrade is longitudinally segmented and sampled to obtain multiple transverse slice images of the radar image; The vertical axis of the radar image of the railway subgrade represents the depth of the railway subgrade, and the horizontal axis of the radar image of the railway subgrade represents the line mileage of the railway subgrade. The radar image is divided into multiple horizontal slices and named; the named radar image is then subjected to layer detection and annotation using a preset target detection model to obtain the annotated radar image and an annotation text file, wherein the annotation text file includes the layer annotation text corresponding to the annotated radar image. Based on the layer annotation text in the annotation text file, determine the layer point coordinates corresponding to the positions of multiple horizontal slice images of the radar image. This includes traversing the annotation text file according to the filenames of the multiple horizontal slice images of the radar image. If the corresponding layer annotation text is found in the annotation text file, then the layer annotation text is read to obtain the layer point coordinates of the corresponding position of the horizontal slice image of the radar image. If the corresponding layer annotation text is not found in the annotation text file, then the layer point coordinates of the corresponding position of the horizontal slice image of the radar image are set to zero. The layer coordinates of the corresponding positions in multiple transverse slices of the determined radar image are stitched together to obtain the layer line of the railway subgrade.

2. The method according to claim 1, characterized in that, Multiple horizontal slice images of the radar image are obtained by longitudinally and uniformly segmenting and sampling the radar image of the railway subgrade.

3. The method according to claim 2, characterized in that, The radar image of the railway subgrade is longitudinally and uniformly segmented and sampled to obtain multiple transverse slice images of the radar image, including: The distance of each sampling slide is determined based on the preset number of sampling points and the preset aspect ratio of the horizontal slice image; Based on the determined distance of each sampling slide, the radar image is slid laterally along the radar image and the radar image of the railway subgrade is longitudinally segmented and sampled until the segmentation and sampling are completed, resulting in multiple horizontal slice images of the radar image.

4. The method according to claim 1, characterized in that, The construction of the target detection model includes: Obtain radar image samples of railway subgrade, wherein the radar image samples of railway subgrade include radar images of railway subgrade or laterally segmented images of radar images of railway subgrade; Obtain the layer label text corresponding to the radar image sample of the railway subgrade; Multiple sets of samples are formed by combining the radar image samples of the railway subgrade and the corresponding layer labeling text of the radar image samples of the railway subgrade. Each set of samples includes a radar image sample of the railway subgrade and the corresponding layer labeling text of the radar image sample of the railway subgrade. The multiple sets of samples are divided into a training set and a test set; The target detection model is obtained by training the model based on the training set and the test set.

5. The method according to claim 3, characterized in that, The multiple horizontal slice images of the radar image are named digitally.

6. A railway subgrade location intelligent identification device, characterized in that, include: The acquisition module is used to acquire radar images of the railway subgrade; The sampling module is used to perform longitudinal segmentation sampling on the radar image of the railway subgrade to obtain multiple horizontal slice images of the radar image; the longitudinal direction of the radar image of the railway subgrade represents the depth of the railway subgrade, and the horizontal direction of the radar image of the railway subgrade represents the line mileage of the railway subgrade. The layer detection module is used to name multiple horizontal slice images of the radar image; and to perform layer detection and annotation on the multiple horizontal slice images of the radar image after file naming through a preset target detection model to obtain multiple horizontal slice images of the radar image after annotation and annotation text file, wherein the annotation text file includes layer annotation text corresponding to the multiple horizontal slice images of the radar image after annotation. The identification module is used to determine the layer coordinates of the corresponding positions in multiple transverse slice images of the radar image based on the layer annotation text in the annotation text file; the identification module is also used to traverse the annotation text file according to the filenames of the multiple transverse slice images of the radar image; if the corresponding layer annotation text is found in the annotation text file, the layer annotation text is read to obtain the layer coordinates of the corresponding position in the transverse slice image of the radar image; if the corresponding layer annotation text is not found in the annotation text file, the layer coordinates of the corresponding position in the transverse slice image of the radar image are set to zero. The stitching module is used to stitch together the layer coordinates of corresponding positions in multiple transverse slices of the determined radar image to obtain the layer line of the railway subgrade.

7. An electronic device, characterized in that, It includes a memory and at least one processor, wherein the memory stores a computer program that, when executed by the at least one processor, implements the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by at least one processor, implements the method as described in any one of claims 1 to 5.

9. A computer program product, characterized in that, The computer program product executes the method as described in any one of claims 1 to 5 when it is run on a processor.

Images