Railway perimeter intrusion detection method and system with super-resolution reconstruction of railway

Abstract

This invention provides a railway perimeter intrusion detection method and system based on railway super-resolution reconstruction, belonging to the field of rail transit safety status identification technology. Based on rail transit perimeter video, a dynamic detection algorithm is used to coarsely detect intrusion targets in specific railway areas, and a region growing method is used to expand the detection area, recording the coordinates of the detection area. Based on the detection area coordinates, a self-built RailVSR super-resolution reconstruction module is used to enlarge the resolution of this area, obtaining an enlarged local area image, and the coordinates of the enlarged area are recorded. Based on the enlarged local area image and the original image, a target detector is used to detect small targets at a distance. The detection results of the enlarged local area are mapped back to the original image using coordinate mapping, and the detection image is output. Intrusion is determined based on the detection results. This invention effectively solves the problem of missed and false alarms in railway perimeter scenes caused by insufficient feature information due to excessive shooting distance.

Description

Technical Field

[0001] This invention relates to the field of rail transit safety status identification technology, specifically to a railway perimeter intrusion detection method and system based on railway super-resolution reconstruction. Background Technology

[0002] With the increase in train speeds and the expansion of the railway network, the importance of train safety has further increased. Railway lines are long and the security situation along the lines is complex; relying solely on human intervention is insufficient to meet the basic requirements of intrusion detection. Intrusions into the railway perimeter can seriously endanger train safety and affect the operational efficiency of the entire network. Railway video surveillance is currently an effective method to prevent significant losses caused by railway perimeter intrusions. With the development of computer vision and CV technologies in recent years, intelligent video surveillance technology has gradually entered the railway video surveillance field and played a significant role. However, most railway cameras are currently standard definition (SD) and are installed at high locations, resulting in wide and deep coverage areas for individual surveillance, leading to blurred and jagged features. This severely affects the final super-resolution results and the final detection. Furthermore, due to the confidentiality and special nature of railways, it is difficult to establish a relevant high-definition target detection database; therefore, high requirements are placed on the algorithm's feature and semantic recognition capabilities.

[0003] The difficulties mentioned above are pain points for the railway industry, especially in terms of databases and the detection of low-resolution small target features, which have long been difficult to solve and effectively prevent.

[0004] To address the above pain points, numerous scholars have conducted extensive research in visual fields and begun combining object detection with super-resolution reconstruction techniques for precise perimeter intrusion prevention, achieving good detection results. However, they have failed to fully exploit feature information to create a super-resolution reconstruction network suitable for object detection. Zhao et al. constructed a new super-resolution object detection network by sequentially connecting super-resolution reconstruction and object detection networks to adapt to the tunnel detection environment. Their network showed good detection performance for small object features, but it cannot be compared with the current state-of-the-art object detection networks, and is even inferior to deep learning networks with added attention mechanisms. Haris et al. further introduced object perception loss into the training of super-resolution networks based on cascaded network construction, exploring the relationship between the two and confirming the feasibility and necessity of SR super-resolution reconstruction. However, the paper did not explore the main connection between the two and the computation time was too long. Yang et al. achieved a significant breakthrough in pedestrian detection by focusing on high-frequency edge information and low-frequency structure. However, their work did not point out the comparison with real high-resolution image detection. There was a phenomenon that the number of detected features was much greater than that of the labeled image, and the authors did not explore the reasons for this phenomenon. This is actually fatal in the railway industry, as it means that a large number of false alarms may occur, affecting decision-making. Summary of the Invention

[0005] The purpose of this invention is to provide a railway perimeter intrusion detection method and system for railway super-resolution reconstruction, so as to solve at least one of the technical problems existing in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a railway perimeter intrusion detection method based on super-resolution reconstruction, comprising the following steps:

[0008] Step 1: Based on the perimeter video of the rail transit, a dynamic detection algorithm is used to perform coarse detection of intrusion targets in specific areas of the railway, and the detection area is expanded using the region growing method, and the coordinates are recorded.

[0009] Step 2: Based on the coordinates of the area detected in Step 1, the area is magnified by 4 times using a self-built RailVSR super-resolution reconstruction module, and the magnified coordinates are recorded.

[0010] Step 3: Based on the magnified local area image from Step 2 and the original image, a target detector is used to detect small targets at a distance.

[0011] Step 4: Map the detection results of the local area after super-resolution magnification back to the original image using coordinates and output the detection image; then determine whether an intrusion has occurred based on the detection results.

[0012] Optionally, in step 1, RailVSR is a video super-resolution reconstruction network suitable for railway scenarios, which is based on an improved GAN network and draws on the downsampling degradation strategy of BasicVSR.

[0013] Optionally, in step 1, the dataset used for training RailVSR is based on video data of real railway monitoring scenes collected over a long period of time, which is highly relevant to railways.

[0014] Optionally, RailVSR establishes a connection with the target detector through perceptual loss, enabling the two to enhance each other.

[0015] Optionally, the target detector in step 3 is not limited to the YOLO series and DETR series, but is only used in conjunction with RailVSR.

[0016] Optionally, the RailVSR network includes a feature separation module, a feature interaction enhancement module, a feature alignment module, an amplification module, and a generator discrimination module. The generator discrimination module includes a perceptual loss function, a Gann loss function, and a pixel loss function. The definitions of each loss function are shown in the following formulas.

[0017] RailVSR reconstruction loss function:

[0018] VSRLoss=pixel_loss+cleaning_loss+perceptual_loss+(5×10 -2 )×gan_loss

[0019] The pixel_loss function and the feature cleaning loss function are defined using the L1 loss function, as shown in the following formula:

[0020]

[0021] The perceptual loss function is shown in the following formula:

[0022]

[0023] BCEWithLogitsLoss(x,y)=max(x,0)-x*y+log(1+exp(-abs(x)))

[0024] Joint loss function:

[0025] α is the RT-DETR loss function; α and β are weighting coefficients, representing the weight ratio between reconstruction loss and detection loss, and the coefficients can be dynamically adjusted to ensure the balance between the two.

[0026] Secondly, the present invention provides a railway perimeter intrusion detection system for railway super-resolution reconstruction, comprising:

[0027] The coarse detection module is used to perform coarse detection of intrusion targets in specific areas of the railway based on the perimeter video of the rail transit, using a dynamic detection algorithm, and expanding the detection area using a region growing method, and recording the coordinates of the detection area;

[0028] The magnification module is used to magnify the resolution of the detection area based on the coordinates of the detection area using a self-built RailVSR super-resolution reconstruction module, to obtain a magnified local area map, and to record the magnified coordinates.

[0029] The target detection module is used to detect small targets at a distance based on a magnified local area image and the original image, using a target detector.

[0030] The mapping and judgment module is used to map the detection results of local areas after super-resolution magnification back to the original image using coordinates, and output the detection image. Based on the detection results, it is determined whether an intrusion has occurred.

[0031] Thirdly, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the railway perimeter intrusion detection method for railway super-resolution reconstruction as described in the first aspect.

[0032] Fourthly, the present invention provides a computer device including a memory and a processor, wherein the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the railway perimeter intrusion detection method for railway super-resolution reconstruction as described in the first aspect.

[0033] Fifthly, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the railway perimeter intrusion detection method for railway super-resolution reconstruction as described in the first aspect.

[0034] The beneficial effects of this invention are as follows: In response to the problem of small target intrusion into the perimeter of rail transit, a method combining super-resolution reconstruction and target detector is proposed, taking into account the actual application scenario of rail transit intrusion. This solves the problem of high false alarm rate and high false alarm rate of existing target detectors for detecting small targets at long distances in railways. At the same time, it verifies the benefits of collaboration between various computer vision tasks for anomaly perception of railway perimeter.

[0035] The advantages of additional aspects of the invention will be set forth more clearly in the following description or will be learned by practice of the invention. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a schematic diagram of the railway perimeter intrusion detection method for railway super-resolution reconstruction according to an embodiment of the present invention.

[0038] Figure 2 This is a flowchart of the railway perimeter intrusion detection method for railway super-resolution reconstruction according to an embodiment of the present invention.

[0039] Figure 3 This is a flowchart of the dynamic monitoring algorithm for extracting feature regions from railway surveillance videos according to an embodiment of the present invention.

[0040] Figure 4 This is a flowchart of the RailVSR super-resolution reconstruction network with RT-DETR target detector described in an embodiment of the present invention.

[0041] Figure 5 This is a super-resolution result image of the feature region of the RailVSR super-resolution reconstruction of railway surveillance video according to an embodiment of the present invention.

[0042] Figure 6 This is a diagram showing the intrusion detection results of railway surveillance video according to an embodiment of the present invention. Detailed Implementation

[0043] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0044] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0045] It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as here.

[0046] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, and / or groups thereof.

[0047] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0048] To facilitate understanding of the present invention, the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. However, the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0049] Those skilled in the art should understand that the accompanying drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily essential for implementing the present invention.

[0050] This invention combines super-resolution reconstruction technology with target detection technology to achieve an integrated super-resolution-detection solution for railway perimeter intrusion prevention, meeting the security maintenance needs of rail transit perimeters and enabling omnidirectional, blind-spot-free perception of objects entering the perimeter. This invention applies super-resolution reconstruction algorithms to rail transit scenarios, addressing the pain point of existing rail transit areas' inability to effectively identify long-distance monitoring targets. It primarily uses a dynamic monitoring algorithm to initially define feature regions, then employs parallel detection. One path sends the feature regions to a RailVSR network for super-resolution processing before inputting them into the target detector; the other path directly inputs the original image into the target detector. The detection results are then fused using coordinate mapping and displayed in the original image. This method effectively solves the problem of missed and false alarms in railway perimeter scenarios caused by insufficient feature information due to excessive shooting distance.

[0051] Example 1

[0052] In this embodiment 1, a railway perimeter intrusion detection system based on railway super-resolution reconstruction is first provided, including: a coarse detection module, used to perform coarse detection of intrusion targets in a specific area of ​​the railway based on the perimeter video of the rail transit, using a dynamic detection algorithm, and expanding the detection area using a region growing method, and recording the coordinates of the detection area; an amplification module, used to amplify the resolution of the area based on the coordinates of the detection area using a self-built RailVSR super-resolution reconstruction module, to obtain an amplified local area map, and recording the coordinates after amplification; a target detection module, used to perform long-distance small target detection based on the amplified local area map and the original image using a target detector; and a mapping judgment module, used to map the detection results of the amplified local area back to the original image using coordinates, output the detection image, and determine whether an intrusion has occurred based on the detection results.

[0053] In this embodiment 1, the railway perimeter intrusion detection method based on railway super-resolution reconstruction is implemented using the above-described system, including the following steps:

[0054] Step 1: Based on the perimeter video of the rail transit, a dynamic detection algorithm is used to perform coarse detection of intrusion targets in specific areas of the railway, and the detection area is expanded using the region growing method, and the coordinates are recorded.

[0055] Step 2: Based on the coordinates of the area detected in Step 1, the area is magnified by 4 times using a self-built RailVSR super-resolution reconstruction module, and the magnified coordinates are recorded.

[0056] Step 3: Based on the magnified local area image from Step 2 and the original image, a target detector is used to detect small targets at a distance.

[0057] Step 4: Map the detection results of the local area after super-resolution magnification back to the original image using coordinates and output the detection image; then determine whether an intrusion has occurred based on the detection results.

[0058] In Step 1, RailVSR is a video super-resolution reconstruction network suitable for railway scenarios, based on an improved GAN network and incorporating the downsampling degradation strategy of BasicVSR. The training dataset used by RailVSR in Step 1 is based on long-term collected video data of real railway monitoring scenes, which has a strong correlation with railways. RailVSR establishes a connection with the object detector through perceptual loss, achieving a combined enhancement function between the two. The object detector in Step 3 is not limited to the YOLO series or DETR series; it is only used in conjunction with RailVSR.

[0059] The RailVSR network comprises a feature separation module, a feature interaction enhancement module, a feature alignment module, an amplification module, and a generator discrimination module. The generator discrimination module includes a perceptual loss function, a GA loss function, and a pixel loss function. The definitions of each loss function are as follows:

[0060] RailVSR reconstruction loss function:

[0061] VSRLoss=pixel_loss+cleaning_loss+perceptual_loss+(5×10 -2 )×gan_loss

[0062] The pixel_loss function and the feature cleaning loss function are defined using the L1 loss function, as shown in the following formula:

[0063]

[0064] The perceptual loss function is shown in the following formula:

[0065]

[0066] BCEWithLogitsLoss(x,y)=max(x,0)-x*y+log(1+exp(-abs(x)))

[0067] Joint loss function:

[0068] α is the RT-DETR loss function; α and β are weighting coefficients, representing the weight ratio between reconstruction loss and detection loss, and the coefficients can be dynamically adjusted to ensure the balance between the two.

[0069] Example 2

[0070] like Figures 1 to 6 As shown, this embodiment provides a railway perimeter intrusion detection method based on railway super-resolution reconstruction. It applies the super-resolution reconstruction algorithm to the rail transit scenario, solving the pain point problem that existing rail transit areas cannot effectively identify long-distance monitoring targets.

[0071] The railway perimeter intrusion detection method for railway super-resolution reconstruction in this embodiment includes the following steps:

[0072] Step 1: Acquire video images of each section of the rail transit perimeter and extract basic information such as the time and monitoring location of the video.

[0073] Step 2: The improved dynamic monitoring algorithm is used to detect the video, and the feature regions are coarsely extracted based on the detection results, and the mapped coordinates are recorded.

[0074] Step 2 should specifically include the following steps:

[0075] Step 2.1: First, perform dynamic coarse detection on the selected region using the improved three-frame difference method.

[0076] Step 2.2: Next, perform point clustering analysis on the region after coarse detection, preliminarily analyze the region where the features are located, and identify redundant regions.

[0077] Step 2.3: Finally, based on the processed coordinates, extend the area to a size of 416*416, and record the coordinates of the top left corner, as well as the length and width, for subsequent coordinate mapping.

[0078] Step 3: Calculate the relative position of the feature region and determine whether it is a region of special interest based on the assumed camera position;

[0079] Step 4: Transmit the area of ​​special interest to the RailVSR network for super-resolution reconstruction, and magnify it by 2 or 4 times depending on the needs and location;

[0080] The specific super-resolution reconstruction method in step 4 mainly includes the following steps:

[0081] Step 4.1: First, locate the features in the image, and then align the features.

[0082] Step 4.2: Separate and enhance the aligned image features.

[0083] Step 4.3: Upsample the feature-enhanced image to obtain a super-resolution image.

[0084] Step 5: Transmit the magnified region image and the original image in parallel to the target detector (this embodiment provides a feasible method, and the target detector uses YOLOv7);

[0085] Step 6: Perform coordinate transformation mapping based on the detection results, and map the super-resolution detection results back to the original image in coordinate form.

[0086] Step 6 mainly includes the following steps:

[0087] Step 6.1: Calculate the relative position of the detection result coordinates within the feature region based on the feature region coordinates obtained in Step 3;

[0088] Step 6.2: Simultaneously, based on the magnification factor of the RailVSR super-resolution reconstruction network in Step 4, the relative position coordinates are proportionally reduced.

[0089] Step 6.3: Compare the original image detection coordinates with the original image detection coordinates. If the IoU of the original image coordinates and the super-resolution feature region detection results is higher than the threshold, discard the original image detection results and use the super-resolution feature region detection results instead, and cover the original image detection results according to the coordinates.

[0090] In this embodiment, a detection implementation method is provided that is paired with an RT-DETR target detector. Taking a certain rail transit perimeter scene as an example, railway perimeter intrusion detection is performed by super-resolution reconstruction of the railway. Video images of each section of the rail transit perimeter are acquired, and the method proposed in this embodiment is used to detect the surveillance video.

[0091] The example demonstrates RailVSR combined with RT-DETR for railway perimeter intrusion detection. The network architecture of the example algorithm is as follows. Figure 4 As shown. It is worth noting that this embodiment only demonstrates some of the target detectors that can be paired with it, and shows the pairing methods. This can be equivalently transferred to other target detectors, and has equivalent universality. In the daily rail transit perimeter scenario, the images captured by the camera are acquired, encoded and decoded, and sent to the dynamic monitoring algorithm for real-time detection, and the detection results are labeled.

[0092] The detection coordinates are used as the center to extend the region, and the coordinates are recorded. Then, based on the RailVSR network model loading and training weights, the resolution of this region is appropriately enlarged, and details are enhanced. Here, depending on the available GPU memory, a multi-threaded parallel detection method can be chosen. Multi-threaded parallel detection allows for block loading and processing of the model, which is faster than serial step-by-step loading. After processing, the super-resolution image and the original image are loaded into the RT-DETR object detector for synchronous processing, and the model detection result coordinates are recorded. The detected coordinates are then loaded into the decision layer for processing, and the super-resolution image detection coordinates are mapped back to the original image according to the coordinate mapping relationship.

[0093] Table 1 shows the accuracy comparison results of the railway perimeter intrusion detection method of railway super-resolution reconstruction in this embodiment with other algorithms on the VOC2007+2012 dataset (using a public dataset to demonstrate the superiority of the algorithm in this patent).

[0094] Table 1 Comparison of Method Accuracy

[0095]

[0096] Figure 6In the image, the first and second rows show the detection results of the original three frames of the video, while the third and fourth rows show the detection images of special areas (enlarged view of special areas). In actual railway operation applications, the detection results of the railway perimeter intrusion detection algorithm based on railway super-resolution reconstruction can be processed and decisions can be made. The pixel area size can be calculated based on the detection results, and the distance of the target relative to the camera can be calculated according to the ratio of real and image pixels. This allows for the rapid location of the intruding target based on detection perception.

[0097] Example 3

[0098] This embodiment 3 provides a non-transitory computer-readable storage medium for storing computer instructions. When executed by a processor, the computer instructions implement the railway perimeter intrusion detection method for railway super-resolution reconstruction as described above. The method includes:

[0099] Based on the perimeter video of rail transit, a dynamic detection algorithm is used to coarsely detect intrusion targets in specific areas of the railway, and the detection area is expanded by using a region growing method, and the coordinates of the detection area are recorded.

[0100] Based on the coordinates of the detection area, the resolution of the area is magnified using a self-built RailVSR super-resolution reconstruction module to obtain a magnified local area map, and the coordinates after magnification are recorded.

[0101] Based on the magnified local region image and the original image, a target detector is used to detect small targets at a distance.

[0102] The detection results of the local area after super-resolution magnification are mapped back to the original image using coordinates, and the detection image is output. The detection results are used to determine whether an intrusion has occurred.

[0103] Example 4

[0104] This embodiment 4 provides a computer device, including a memory and a processor, wherein the processor and the memory communicate with each other, and the memory stores program instructions that can be executed by the processor. The processor calls the program instructions to execute the railway perimeter intrusion detection method for railway super-resolution reconstruction as described above, the method including:

[0105] Based on the perimeter video of rail transit, a dynamic detection algorithm is used to coarsely detect intrusion targets in specific areas of the railway, and the detection area is expanded by using a region growing method, and the coordinates of the detection area are recorded.

[0106] Based on the coordinates of the detection area, the resolution of the area is magnified using a self-built RailVSR super-resolution reconstruction module to obtain a magnified local area map, and the coordinates after magnification are recorded.

[0107] Based on the magnified local region image and the original image, a target detector is used to detect small targets at a distance.

[0108] The detection results of the local area after super-resolution magnification are mapped back to the original image using coordinates, and the detection image is output. The detection results are used to determine whether an intrusion has occurred.

[0109] Example 5

[0110] This embodiment 5 provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the railway perimeter intrusion detection method for railway super-resolution reconstruction as described above. The method includes:

[0111] Based on the perimeter video of rail transit, a dynamic detection algorithm is used to coarsely detect intrusion targets in specific areas of the railway, and the detection area is expanded by using a region growing method, and the coordinates of the detection area are recorded.

[0112] Based on the coordinates of the detection area, the resolution of the area is magnified using a self-built RailVSR super-resolution reconstruction module to obtain a magnified local area map, and the coordinates after magnification are recorded.

[0113] Based on the magnified local region image and the original image, a target detector is used to detect small targets at a distance.

[0114] The detection results of the local area after super-resolution magnification are mapped back to the original image using coordinates, and the detection image is output. The detection results are used to determine whether an intrusion has occurred.

[0115] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0116] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0117] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0118] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0119] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solutions disclosed in the present invention, various modifications or variations that can be made by those skilled in the art without creative effort should be included within the scope of protection of the present invention.

Claims

1. A method for detecting railway perimeter intrusion in railway super-resolution reconstruction, characterized in that, include: Based on the perimeter video of rail transit, a dynamic detection algorithm is used to coarsely detect intrusion targets in specific areas of the railway, and the detection area is expanded by using a region growing method, and the coordinates of the detection area are recorded. Based on the coordinates of the detection region, a self-built RailVSR super-resolution reconstruction module is used to upscale the region, obtaining an enlarged local area map, and recording the coordinates after upscaling. RailVSR is a video super-resolution reconstruction network suitable for railway scenarios, based on an improved GAN network combined with a BasicVSR downsampling degradation strategy. RailVSR establishes a connection with the target detector through perceptual loss. The RailVSR network includes a feature separation module, a feature interaction enhancement module, a feature alignment module, an upscaling module, and a generator discrimination module. The generator discrimination module includes a perceptual loss function (perceptual_loss), a GAN loss function (gan_loss), and a pixel loss function (pixel_loss). The RailVSR reconstruction loss function... for: ； The pixel_loss function and the feature cleaning loss function are defined using the L1 loss function, specifically as follows: ； The perceptual loss function is: ； ； The joint loss function is: ; in, The loss function is RT-DETR. , The weighting coefficient represents the weight ratio between reconstruction loss and detection loss, and the coefficient can be dynamically adjusted to ensure a balance between the two. Based on the magnified local region image and the original image, a target detector is used to detect small targets at a distance. The detection results of the local area after super-resolution magnification are mapped back to the original image using coordinates, and the detection image is output. The detection results are used to determine whether an intrusion has occurred.

2. The railway perimeter intrusion detection method for railway super-resolution reconstruction according to claim 1, characterized in that, The training dataset used by RailVSR consists of video data from real-world railway monitoring scenarios.

3. The railway perimeter intrusion detection method for railway super-resolution reconstruction according to claim 1, characterized in that, The target detector is designed for use with RailVSR and is not limited to the YOLO series or DETR series.

4. A railway perimeter intrusion detection system based on super-resolution railway reconstruction, characterized in that, include: The coarse detection module is used to perform coarse detection of intrusion targets in specific areas of the railway based on the perimeter video of the rail transit, using a dynamic detection algorithm, and expanding the detection area using a region growing method, and recording the coordinates of the detection area; The magnification module is used to magnify the resolution of the detection region based on its coordinates using a self-built RailVSR super-resolution reconstruction module, obtaining a magnified local region map and recording the magnified coordinates. RailVSR is a video super-resolution reconstruction network suitable for railway scenarios, based on an improved GAN network combined with a BasicVSR downsampling degradation strategy. RailVSR establishes a connection with the target detector through perceptual loss. The RailVSR network includes a feature separation module, a feature interaction enhancement module, a feature alignment module, a magnification module, and a generator discriminant module. The generator discriminant module contains a perceptual loss function. ,GAN loss function Pixel loss function; RailVSR reconstruction loss function is: ； The pixel_loss function and the feature cleaning loss function are defined using the L1 loss function, specifically as follows: ； The perceptual loss function is: ； ； The joint loss function is: ; in, The loss function is RT-DETR. , The weighting coefficient represents the weight ratio between reconstruction loss and detection loss, and the coefficient can be dynamically adjusted to ensure a balance between the two. The target detection module is used to detect small targets at a distance based on a magnified local area image and the original image, using a target detector. The mapping and judgment module is used to map the detection results of local areas after super-resolution magnification back to the original image using coordinates, and output the detection image. Based on the detection results, it is determined whether an intrusion has occurred.

5. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the railway perimeter intrusion detection method for railway super-resolution reconstruction as described in any one of claims 1-3.

6. A computer device, characterized in that, The system includes a memory and a processor, which communicate with each other. The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the railway perimeter intrusion detection method for railway super-resolution reconstruction as described in any one of claims 1-3.

7. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the railway perimeter intrusion detection method for railway super-resolution reconstruction as described in any one of claims 1-3.

Images