Railway overhead line system icing detection method and system based on polarization imaging and storage medium

Abstract

The invention discloses a railway catenary icing detection method and system based on polarization imaging and a storage medium, and belongs to the technical field of catenary icing detection, and the method comprises the steps: obtaining a to-be-detected railway catenary image containing polarization characteristics; inputting the to-be-detected railway overhead line system image into a pre-trained icing detection model to obtain an icing detection result; the acquisition method of the pre-trained icing detection model comprises the steps that a contact network icing sample set is acquired, the contact network icing sample set comprises contact network icing images, and the contact network icing images comprise polarization characteristics; inputting the overhead line system icing sample set into a pre-constructed image segmentation network to obtain a trained icing detection model, and using the image segmentation network to perform segmentation of different types of regions on the railway overhead line system image based on the difference of polarization characteristics of an overhead line system region and an icing region to obtain a real-time detection result of the icing region, accurate icing state measurement is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of contact network icing detection, and in particular relates to a railway contact network icing detection method, system and storage medium based on polarization imaging. Background Art

[0002] Catenary icing will affect the interaction between the pantograph and the catenary, which may lead to problems such as poor current collection by the pantograph from the catenary and arc burns on power supply equipment. In serious cases, it may even lead to catenary disconnection and railway traffic interruption, which seriously threatens the safe operation of electrified railways and affects railway transportation safety. In order to provide early warning of the catenary icing status during railway operation, it is necessary to detect the icing status.

[0003] At present, in addition to manual visual inspection, there are other methods for detecting the icing state of the contact network, such as infrared imaging, weighing, and meteorological monitoring. Infrared imaging uses infrared cameras to obtain real-time infrared images and visible light images of the target contact network, and judges the icing situation based on the difference between the two types of spectral images; the weighing method monitors the tension and gravity of the contact network of a fixed length. If there are large changes in a short period of time, it is combined with real-time meteorological data to determine whether there is icing; the meteorological detection method is to obtain real-time meteorological data, such as temperature, humidity, rainfall, snowfall, and wind speed, to evaluate whether it will cause icing on the contact network. Among them, the visual method is still the most direct detection method, but due to the small size and transparent color of ice, it is difficult to identify and distinguish using visible light, while infrared cameras can make a certain distinction, but because the temperature of the contact network and ice is also very close, it is actually difficult to accurately distinguish. Summary of the invention

[0004] The present invention provides a method, system and storage medium for detecting ice on a railway contact network based on polarization imaging. The method collects railway contact network images containing polarization characteristics, and uses an image segmentation network to classify the railway contact network images based on the differences in polarization characteristics between contact network areas and ice areas, thereby obtaining the final ice detection result, thereby achieving accurate and efficient real-time ice detection effects.

[0005] To achieve the above object, the present invention is implemented by adopting the following technical solutions: In a first aspect, the present invention provides a method for detecting icing of a railway contact network based on polarization imaging, comprising: Acquire an image of the railway contact network to be tested including polarization characteristics; Inputting the railway contact network image to be tested into a pre-trained ice detection model to obtain an ice detection result; The method for obtaining the pre-trained icing detection model includes: Acquire a contact network icing sample set, wherein the contact network icing sample set includes a contact network icing image, and the contact network icing image contains polarization characteristics; The contact network icing sample set is input into a pre-built image segmentation network to obtain a trained icing detection model.

[0006] Optionally, obtaining the contact network icing sample set includes: Acquire multiple images of catenary icing under different backgrounds; Segmenting and labeling the catenary icing images into categories, and generating labels corresponding to the catenary icing images, wherein the categories of the catenary icing images segmented include catenary areas, ice-covered areas, and background areas; The contact network icing images and their corresponding labels are combined into a contact network icing sample set.

[0007] Optionally, the acquiring of multiple images of contact network icing under different backgrounds includes: acquiring images of the contact network through a polarization imaging system on top of a train during the operation of the train, and screening the acquired images to obtain images of contact network icing under multiple backgrounds, wherein the polarization imaging system includes a camera and a polarizer installed in front of the camera.

[0008] Optionally, the process of installing the polarizer includes: The reflected light from the contact network to the camera lens is divided into multiple channels by a beam splitting element; Install a polarizing film in front of each channel and adjust the polarizing film in front of each channel to the optimal position.

[0009] Optionally, adjusting the polarizer before each channel to an optimal orientation includes: Adjust the orientation of the polarizer in front of each channel to obtain the difference in texture information between the ice area and the contact network area in the image under different polarizer orientations; The polarizer is fixed in the orientation where the difference in texture information is the largest.

[0010] Optionally, the step of inputting the contact network icing sample set into a pre-built image segmentation network to obtain a trained icing detection model includes: Extracting features from the input sample set image through the encoder of the image segmentation network; Upsampling the feature map extracted by the encoder through the decoder of the image segmentation network to obtain a feature map with restored resolution; Converting the feature map restored by the resolution into a category segmentation map through the output layer of the image segmentation network; The parameters of the detection model are adjusted according to the loss between the category segmentation map and the corresponding label to obtain a trained ice detection model.

[0011] In a second aspect, the present invention provides a railway contact network icing detection system based on polarization imaging, comprising: The contact network image acquisition module to be tested is used to obtain the image of the contact network to be tested including polarization characteristics; An icing detection result acquisition module is used to input the railway contact network image to be tested into a pre-trained icing detection model to obtain an icing detection result; The method for obtaining the pre-trained icing detection model includes: Acquire a contact network icing sample set, wherein the contact network icing sample set includes a contact network icing image, and the contact network icing image contains polarization characteristics; The contact network icing sample set is input into a pre-built image segmentation network to obtain a trained icing detection model.

[0012] In a third aspect, the present invention provides a computer storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the railway contact network icing detection method based on polarization imaging as described in any step of the first aspect is implemented.

[0013] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: by collecting contact network images containing polarization characteristics, utilizing the differences in polarization directions of reflected sunlight by objects of different materials, and distinguishing the shapes of the contact network and ice in the image, the contact network area and the ice area in the image are further distinguished in the texture of the imaging, which is convenient for the subsequent image processing algorithm to perform accurate segmentation and extraction, and is conducive to achieving more accurate measurement of the icing state. At the same time, by utilizing target detection and segmentation technology, real-time detection of the icing state is achieved, and the ice coverage and ice size are obtained in real time, which is convenient for timely judgment on whether deicing is needed, thereby ensuring the normal operation of the train; by adding a polarizing plate in front of each reflective channel of the visible light camera and adjusting the angular direction of the polarizing plate, the difference between the ice imaging texture and the contact network imaging texture on each polarizing plate is maximized, thereby collecting contact network images containing all polarization characteristics, and further improving the detection effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The figure is a flow chart of a railway contact network icing detection method based on polarization imaging in one embodiment of the present invention;

[0015] Figure 2 Shown is a schematic structural diagram of a polarization imaging system in an embodiment of the present invention. DETAILED DESCRIPTION

[0016] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0017] Example 1

[0018] As shown in the figure, this embodiment provides a railway contact network icing detection method based on polarization imaging, including: Acquire an image of the railway contact network to be tested including polarization characteristics; Inputting the railway contact network image to be tested into a pre-trained ice detection model to obtain an ice detection result; The method for obtaining the pre-trained icing detection model includes: Acquire a contact network icing sample set, wherein the contact network icing sample set includes a contact network icing image, and the contact network icing image contains polarization characteristics; The contact network icing sample set is input into a pre-built image segmentation network to obtain a trained icing detection model.

[0019] Example 2

[0020] Based on Example 1, this example also makes the following design.

[0021] The railway contact network icing detection method is divided into the following steps.

[0022] S1. Install a polarization imaging system on the roof.

[0023] A polarization imaging system is used to obtain real-time images of the railway overhead contact network, which can capture the changes in polarization characteristics caused by icing. Figure 2 As shown, the polarization imaging system can be realized by adding a polarizing plate in front of the original camera lens on the top of the train.

[0024] In this embodiment, in order to further improve the detection effect, an amplitude-split polarization imaging system may be used, and a beam splitter element is used to split the reflected light into multiple channels, and a separate polarizer is added in front of each channel.

[0025] S2. Adjust the direction and angle of the polarizer.

[0026] To achieve the best effect, the angle and direction of the polarizer can be continuously adjusted. After each adjustment, the contact network in the same icy state is photographed and observed. When the difference in texture information such as color and brightness between the ice area and the contact network area in the imaging image is the largest, the angle and direction of the polarizer at this time is fixed.

[0027] For each reflected light channel in the amplitude-divided polarization imaging system, the polarization direction needs to be adjusted, and different polarization modulation methods are implemented in each channel to obtain the maximum difference in the imaging effects of ice and contact network on the image.

[0028] S3. Collect a large number of photos of contact network icing under different backgrounds, including sunny backgrounds, cloudy backgrounds, rainy backgrounds, snowy backgrounds, and night backgrounds, with no less than 100 photos under each background.

[0029] The collected photos of contact network icing are preprocessed to remove invalid photos and obtain a contact network icing sample set.

[0030] S4. Use annotation tools to segment and annotate the images in the contact network icing sample set and extract the icing area.

[0031] In this embodiment, LabelImg is used to annotate images of the contact network icing sample set. After starting LabelImg, select the contact network icing image file to be annotated and then click the "Polygon" option in the dot product menu bar. A draggable polygon box is displayed on the contact network icing image. Click and drag the polygon box to add vertices so that the polygon box coincides with the boundary of the contact network area, the icing area, or the background area. After the polygons of each area boundary are drawn, enter the corresponding category name in the "Label" panel, and finally save the annotated data.

[0032] S5. Use the deep learning image segmentation network to detect and segment the image and complete the image segmentation network training.

[0033] 1. Dataset division

[0034] The labeled catenary icing sample set is divided into a training set, a validation set, and a test set. For example, ice morphology images of a specific year can be selected as the training set and the test set.

[0035] 2. Model selection

[0036] Select model architecture: Depending on the task requirements, you can choose a deep learning model suitable for image segmentation, such as DeepLab, U-net, etc. U-net consists of an encoder module, a decoder module, and an output layer.

[0037] The encoder module is mainly responsible for reducing the dimension of the input feature map of the catenary icing image, and usually consists of the following parts:

[0038] Max pooling layer: used to reduce the spatial dimension of the input feature map.

[0039] Double convolutional layer: It consists of two convolutional layers, and each convolutional layer is followed by batch normalization BatchNorm and ReLU activation function. The first convolutional layer is used to extract features from the feature map after dimensionality reduction, and the second convolutional layer is used to further extract and refine the feature map extracted by the first convolutional layer.

[0040] The decoder module is responsible for restoring the feature map after dimensionality reduction to the same resolution as the input image, and usually consists of the following parts:

[0041] Upsampling layer: You can use transposed convolution or bilinear interpolation to achieve upsampling of feature maps.

[0042] Double convolutional layer: It has the same structure as the double convolutional layer in the encoder module and is used to process the feature maps output by the upsampling layer.

[0043] Feature fusion: Through skip connections, the feature map output by the upsampling layer is concatenated with the feature map of the corresponding resolution output by the encoder module, and the final feature map is output to enhance the feature expression ability of the model.

[0044] The output layer is responsible for converting the final feature map into a segmentation map, which usually consists of the following parts:

[0045] 1x1 convolution: Use a 1x1 convolution kernel to reduce the number of channels of the final feature map to the preset number of categories, implement the classification task, and output the segmentation result of the ice area.

[0046] 3. Model training

[0047] Configuration file settings: Modify the dataset configuration file to specify the paths to the training, validation, and test data, as well as the category names.

[0048] Training process: Use the training set data to train the model. During the training process, monitor indicators such as loss value Loss, pixel accuracy PA, and mean intersection over union (MIoU).

[0049] Validation and adjustment: Use validation set data to evaluate model performance and adjust model parameters based on the calculated indicators to optimize model performance.

[0050] 4. Model Evaluation

[0051] Evaluation indicators: Evaluate the model's pixel accuracy PA, average intersection over union (MIoU) value, and intersection over union (IoU) value of each category on the test set to determine the segmentation effect of the model.

[0052] S6. Use the polarization imaging system to collect real-time images of the contact network to be tested, and use the trained network and polarization imaging system to perform real-time detection of the contact network icing state on the images to be tested and determine the icing area. For different images formed by the amplitude-based polarization imaging system, a deep learning segmentation network is used to detect icing. The results of different images can be fused and the result with the highest confidence is selected.

[0053] Example 3

[0054] This embodiment provides a railway contact network icing detection system based on polarization imaging, which is characterized by comprising: The contact network image acquisition module to be tested is used to obtain the image of the contact network to be tested including polarization characteristics; An icing detection result acquisition module is used to input the railway contact network image to be tested into a pre-trained icing detection model to obtain an icing detection result; The method for obtaining the pre-trained icing detection model includes: Acquire a contact network icing sample set, wherein the contact network icing sample set includes a contact network icing image, and the contact network icing image contains polarization characteristics; The contact network icing sample set is input into a pre-built image segmentation network to obtain a trained icing detection model.

[0055] Example 4

[0056] This embodiment provides a computer storage medium on which a computer program is stored. When the computer program is executed by a processor, the railway contact network icing detection method based on polarization imaging as described in any step of Example 2 is implemented.

[0057] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0058] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0059] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0060] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0061] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A method for detecting icing of a railway contact network based on polarization imaging, characterized in that: include: Acquire an image of the railway contact network to be tested including polarization characteristics; Inputting the railway contact network image to be tested into a pre-trained ice detection model to obtain an ice detection result; The method for obtaining the pre-trained icing detection model includes: Acquire a contact network icing sample set, wherein the contact network icing sample set includes a contact network icing image, and the contact network icing image contains polarization characteristics; The contact network icing sample set is input into a pre-built image segmentation network to obtain a trained icing detection model.

2. The method for detecting icing of a railway contact network based on polarization imaging according to claim 1, characterized in that: The step of obtaining a contact network icing sample set comprises: Acquire multiple images of catenary icing under different backgrounds; Segmenting and labeling the catenary icing images into categories, and generating labels corresponding to the catenary icing images, wherein the categories of the catenary icing images segmented include catenary areas, ice-covered areas, and background areas; The contact network icing images and their corresponding labels are combined into a contact network icing sample set.

3. The method for detecting icing of a railway contact network based on polarization imaging according to claim 2, characterized in that: The method of acquiring multiple images of contact network icing under different backgrounds includes: acquiring images of the contact network through a polarization imaging system on the top of a train during train operation, and screening the acquired images to obtain images of contact network icing under multiple backgrounds, wherein the polarization imaging system includes a camera and a polarizing plate installed in front of the camera.

4. The method for detecting icing of a railway contact network based on polarization imaging according to claim 3 is characterized in that: The process of installing the polarizer includes: The reflected light from the contact network to the camera lens is divided into multiple channels by a beam splitting element; Install a polarizing film in front of each channel and adjust the polarizing film in front of each channel to the optimal position.

5. The method for detecting icing of a railway contact network based on polarization imaging according to claim 4, characterized in that: The step of adjusting the polarizers in front of each channel to the optimal position comprises: Adjust the orientation of the polarizer in front of each channel to obtain the difference in texture information between the ice area and the contact network area in the image under different polarizer orientations; The polarizer is fixed in the orientation where the difference in texture information is the largest.

6. The method for detecting icing of a railway contact network based on polarization imaging according to claim 2, characterized in that: The step of inputting the contact network icing sample set into a pre-built image segmentation network to obtain a trained icing detection model comprises: Extracting features from the input sample set image through the encoder of the image segmentation network; Upsampling the feature map extracted by the encoder through the decoder of the image segmentation network to obtain a feature map with restored resolution; Converting the feature map restored by the resolution into a category segmentation map through the output layer of the image segmentation network; The parameters of the detection model are adjusted according to the loss between the category segmentation map and the corresponding label to obtain a trained ice detection model.

7. A railway contact network icing detection system based on polarization imaging, characterized in that: include: The contact network image acquisition module to be tested is used to obtain the image of the contact network to be tested including polarization characteristics; An icing detection result acquisition module is used to input the railway contact network image to be tested into a pre-trained icing detection model to obtain an icing detection result; The method for obtaining the pre-trained icing detection model includes: Acquire a contact network icing sample set, wherein the contact network icing sample set includes a contact network icing image, and the contact network icing image contains polarization characteristics; The contact network icing sample set is input into a pre-built image segmentation network to obtain a trained icing detection model.

8. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the railway contact network icing detection method based on polarization imaging as described in any one of claims 1 to 6 is implemented.

Images