Turnout outdoor equipment monitoring method and device based on neural network
By using a neural network-based monitoring method, intelligent monitoring of outdoor turnout equipment is achieved, solving the problems of low efficiency, poor accuracy, and poor reliability in existing technologies, and realizing comprehensive monitoring of outdoor turnout equipment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHENGDU TANGYUAN ELECTRICAL APPLIANCE
- Filing Date
- 2023-06-30
- Publication Date
- 2026-06-12
AI Technical Summary
Existing outdoor turnout monitoring technologies are inefficient, inaccurate, and unreliable, and cannot achieve comprehensive monitoring.
A neural network-based monitoring method is adopted, which monitors all key equipment through an outdoor equipment monitoring device for turnouts. It combines deep learning algorithms, image processing algorithms, and data fitting algorithms to achieve intelligent monitoring.
It reduces manual labor, improves the accuracy and reliability of monitoring, and enables comprehensive monitoring of outdoor turnout equipment.
Smart Images
Figure CN117011781B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of rail transit technology, specifically relating to a method and device for monitoring outdoor turnout equipment based on neural networks. Background Technology
[0002] A turnout is a track connection device that allows locomotives and rolling stock to switch from one track to another. It is also one of the weakest links in the track and is usually laid in large quantities in stations and marshalling yards.
[0003] Because turnouts are numerous, complex in structure, short in service life, restrict train departure speed, have low train safety, and require large maintenance and repair costs, they are considered one of the three weakest links in the track, along with curves and joints. Therefore, it is necessary to monitor turnout equipment to ensure train operation safety.
[0004] Existing technologies mainly focus on monitoring a specific part of the outdoor equipment of turnouts, and are primarily based on manual monitoring. However, manual monitoring requires a large amount of manpower, has low monitoring efficiency, and the monitoring results are easily affected by subjective factors, resulting in poor accuracy and reliability. Furthermore, there is currently no complete monitoring technology for outdoor equipment of turnouts. Summary of the Invention
[0005] To address the problems of low efficiency, poor accuracy and reliability, and the inability to achieve comprehensive monitoring of outdoor turnout equipment in existing monitoring technologies, this invention provides a neural network-based monitoring device and method for outdoor turnout equipment. This invention aims to monitor all key outdoor turnout equipment and achieves intelligent monitoring through neural network-based monitoring technology, significantly reducing manual labor and improving the accuracy and reliability of monitoring.
[0006] This invention is achieved through the following technical solution:
[0007] A neural network-based outdoor equipment monitoring device for turnouts, comprising: turnout switch rail monitoring equipment, monitoring sub-unit, and internal monitoring equipment for switch machines;
[0008] The internal monitoring equipment of the switch machine is used to monitor the automatic switch, turnout gap, turnout operating current, turnout switching force and oil level inside the switch machine.
[0009] The switch point rail monitoring equipment is used to monitor the switch point rail opening and the switch point rail creep.
[0010] The monitoring unit is used to control the turnout switch rail monitoring equipment and the internal monitoring equipment of the switch machine.
[0011] A method for monitoring outdoor turnout equipment based on neural networks, the method comprising:
[0012] The monitoring data of each monitoring point of the turnout outdoor equipment is obtained through the turnout outdoor equipment monitoring device. The monitoring data includes image data and digital data.
[0013] Deep learning algorithms are used to locate and analyze the acquired image data, identify images of key components of the turnout outdoor equipment, and monitor external anomalies of key components of the turnout outdoor equipment.
[0014] Image processing algorithms are used to process the images of the key components of the turnout outdoor equipment, and the edges of the components are extracted from the images of the key components of the turnout outdoor equipment to perform distance anomaly monitoring of the turnout outdoor equipment.
[0015] A data fitting algorithm is used to perform data fitting calculations on the acquired digital data to monitor internal anomalies in the outdoor equipment of the turnout.
[0016] In a preferred embodiment, the present invention employs a deep learning algorithm to perform localization analysis on the acquired image data and identify images of key components of the turnout's outdoor equipment, specifically including:
[0017] The image data is used to construct training sample data, which includes imaging of the automatic switch opening and closing area and the turnout gap area.
[0018] The automatic switch opening and closing device area, key components of the automatic switch opening and closing device, and turnout gaps in the training sample data were labeled using the labeling tool.
[0019] A deep learning model was trained using labeled training sample data to obtain the positioning models of the automatic switch machine opener, key components of the automatic switch machine opener, and turnout gaps.
[0020] The trained positioning model is used to locate the automatic switch opener, key components of the automatic switch opener, and turnout gaps in real-time acquired image data, and the positioning results are output.
[0021] As a preferred embodiment, the deep learning model of the present invention uses YOLOv5.
[0022] In a preferred embodiment, the present invention employs an image processing algorithm to process the identified key component image of the turnout outdoor equipment, extracting the component edges from the key component image, specifically including:
[0023] Remove lighting factors from images of key components of the identified outdoor turnout equipment;
[0024] An edge detector is used to extract component edges from images of key components of outdoor turnout equipment after removing illumination factors.
[0025] As a preferred embodiment, the calculation formula for removing the light factor in the present invention is as follows:
[0026] I' (i,j) =I (i,j) ≥I mean ? 0:110
[0027] Where I and I′ are the input image data and the result image data after removing illumination factors, respectively, and (i,j) is the image position. mean The formula for calculating the pixel mean of the input image data is as follows:
[0028]
[0029] Where col and row are the width and height of the input image, respectively, I i,j The pixel value in the j-th row and i-th column of the input image.
[0030] In a preferred embodiment, the present invention employs an edge detector to extract component edges from an image of key components of outdoor turnout equipment after removing illumination factors, specifically including:
[0031] Apply Gaussian filtering to the image;
[0032] Calculate the gradient value and gray-level direction of the filtered image;
[0033] Based on the calculated gradient values, filter out non-maximum values;
[0034] Edge detection is performed using upper and lower thresholds: pixels greater than the upper threshold are identified as edges, and pixels less than the lower threshold are identified as non-edges; for pixels greater than or equal to the lower threshold and less than or equal to the upper threshold, if they are adjacent to pixels identified as edges, they are identified as edges, otherwise they are non-edges.
[0035] In a preferred embodiment, the present invention employs a data fitting algorithm to perform data fitting calculations on the acquired digital data, specifically including:
[0036] Construct an nth-degree polynomial:
[0037]
[0038] Among them, a0 to a n The fitting parameters for an nth-degree polynomial;
[0039] x-coordinate of the sample point i Substituting the nth-degree polynomial into the equation, we obtain the ordinate of the nth-degree polynomial at the x-axis:
[0040]
[0041] The sum of squares of the errors between the ordinate of the nth-degree polynomial at the x-axis and the ordinate of the sample points is calculated:
[0042]
[0043] Take the (n+1)th partial derivative with respect to the sum of squares of the errors, and set each partial derivative to 0:
[0044]
[0045] The above system of equations can be expanded to include the fitting coefficients a0 to a n The equation;
[0046] Solving the system of equations containing n+1 equations will yield the fitting coefficients from a0 to a. n The fitting coefficients are then substituted into the nth-degree polynomial to achieve the fitting.
[0047] Secondly, the present invention also proposes a turnout outdoor equipment monitoring device based on a neural network. The turnout outdoor equipment monitoring device is applicable to the monitoring method described above in the present invention. The turnout outdoor equipment monitoring device includes a turnout switch rail monitoring device, a monitoring sub-unit, and a switch machine internal monitoring device.
[0048] The internal monitoring equipment of the switch machine is used to monitor the automatic switch, turnout gap, turnout operating current, turnout switching force and oil level inside the switch machine.
[0049] The switch point rail monitoring equipment is used to monitor the switch point rail opening and the switch point rail creep.
[0050] The monitoring unit is used to control the turnout switch rail monitoring equipment and the internal monitoring equipment of the switch machine.
[0051] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.
[0052] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.
[0053] The present invention has the following advantages and beneficial effects:
[0054] This invention can monitor all key components of turnout outdoor equipment and, combined with machine learning algorithms, achieve real-time monitoring of turnout outdoor equipment, reducing unnecessary manual labor and improving the monitoring accuracy and reliability of turnout outdoor equipment.
[0055] This invention achieves external anomaly monitoring of key components of turnout outdoor equipment (such as anomalies in component integrity, shape, and distance dimensions) through deep learning models and image processing algorithms; at the same time, it achieves internal anomaly monitoring of key components of turnout outdoor equipment (such as anomalies in turnout operating current and turnout switching force) through data fitting algorithms, thereby ensuring comprehensive monitoring of turnout outdoor equipment and improving the accuracy and reliability of monitoring. Attached Figure Description
[0056] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:
[0057] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention.
[0058] Figure 2 This is a block diagram illustrating the principle of the monitoring device used in an embodiment of the present invention.
[0059] Figure 3 This is a schematic diagram of the original data display interface of the switch machine gap and automatic gate, collected by the monitoring device in an embodiment of the present invention.
[0060] Figure 4 This is a schematic diagram of the display interface for the results of data processing according to an embodiment of the present invention. Detailed Implementation
[0061] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.
[0062] Example:
[0063] Existing monitoring technologies for outdoor turnout equipment mainly focus on monitoring a portion of the outdoor turnout equipment, and are primarily manual, requiring significant manpower and suffering from poor accuracy and reliability due to subjective human factors. Therefore, this embodiment proposes a neural network-based method for monitoring outdoor turnout equipment.
[0064] Specifically, such as Figure 1 As shown, the method proposed in this embodiment includes the following steps:
[0065] Step 1: Obtain monitoring data from each monitoring point of the turnout outdoor equipment through the turnout outdoor equipment monitoring device, including image data and digital data (sensor data).
[0066] Step 2: Use deep learning algorithms to locate and analyze the acquired image data, identify the key components of the turnout outdoor equipment, and monitor external anomalies such as the integrity of the key components of the turnout outdoor equipment. The main monitoring items include missing nuts, abnormal status of the automatic switch opener, breakage of the automatic switch opener, arcing, etc.
[0067] Step 3: Use image processing algorithms to process the images of key components of the identified turnout outdoor equipment, extract the component edges from the turnout outdoor component image data, and perform distance anomaly monitoring of the turnout outdoor equipment, mainly including switch machine gap width, switch rail opening, switch rail creep, etc.
[0068] Step 4: Use a data fitting algorithm to perform data fitting calculations on the collected digital data, thereby realizing the internal anomaly monitoring of the turnout outdoor equipment, mainly including turnout switching force monitoring, turnout operating current monitoring, and turnout switch machine vibration monitoring.
[0069] As an optional implementation, the turnout outdoor equipment monitoring device used in this embodiment mainly consists of three parts: turnout switch rail monitoring equipment, monitoring sub-unit, and internal monitoring equipment for the switch machine, as detailed below. Figure 2 As shown in the diagram. The internal monitoring equipment of the switch machine uses cameras and sensors to monitor the automatic switch, turnout gap, turnout operating current, turnout switching force, and oil level inside the switch machine. The turnout switch rail monitoring equipment uses cameras to monitor the turnout switch rail opening and crawling. The monitoring sub-unit is responsible for controlling, supplying power to, and transmitting data to the other two monitoring devices (i.e., the turnout switch rail monitoring equipment and the internal monitoring equipment of the switch machine).
[0070] The monitoring data collected by the outdoor equipment monitoring device for turnouts includes: switch machine operating current, switch machine vibration, turnout switching force, status of the switch machine automatic opener / closer, turnout gap measurement, switch machine oil level, turnout switch rail opening stroke, and turnout switch rail creep.
[0071] As an optional implementation method, deep learning algorithms specifically include:
[0072] Training sample data was constructed using image data (including images of automatic switch openers, switch machine gaps, and switch point rails) acquired by cameras in the outdoor equipment monitoring device for turnouts. The training sample data included images of the automatic switch opener area and the turnout gap area.
[0073] The automatic switch opening and closing device area, key components of the automatic switch opening and closing device, and turnout gaps in the training sample data were labeled using the annotation tool.
[0074] A deep learning model was trained using labeled training sample data to obtain the positioning models of the automatic switch machine opener, key components of the automatic switch machine opener, and turnout gaps.
[0075] Using the trained positioning model, the system locates the automatic switch opener, key components of the automatic switch opener, and turnout gaps in real-time captured images and outputs the positioning results, including target category information and positioning information, to monitor the integrity of key components of the turnout's outdoor equipment.
[0076] The deep learning model used in this embodiment is, but is not limited to, YOLOv5.
[0077] As an optional implementation method, the image processing algorithm specifically includes:
[0078] Remove illumination factors from images of key components of outdoor turnout equipment identified by deep learning algorithms.
[0079] An edge detector is used to extract the edges of key components from images of outdoor turnout equipment after removing illumination factors, which are then used for subsequent component location calculations.
[0080] Specifically, the calculation formula for removing the light factor is as follows:
[0081] I' (i,j) =I (i,j) ≥I mean ? 0:110
[0082] Where I and I′ are the input image data and the result image data after removing illumination factors, respectively, and (i,j) is the image position. mean The formula for calculating the pixel mean of the input image data is as follows:
[0083]
[0084] Where col and row are the width and height of the input image, respectively, I i,j The pixel value in the j-th row and i-th column of the input image.
[0085] In this embodiment, removing illumination factors aims to address the issue of inconsistent illumination intensity in the images of key components of the turnout outdoor equipment identified by the deep learning model, thereby providing image data with uniform illumination intensity for subsequent Canny edge detection.
[0086] The edge detector uses, but is not limited to, the Canny edge detector. The specific detection process includes:
[0087] Gaussian filtering is applied to an image primarily for noise reduction, a common step in most image processing algorithms. The Gaussian function is similar to a normal distribution, characterized by a high center and low extremes. For a pixel at position (m, n), its grayscale value (considering only binary images) is f(m, n). After Gaussian filtering, the grayscale value will become:
[0088]
[0089] That is, multiply each pixel and its neighborhood by a Gaussian matrix, and take the weighted average value as the final gray value.
[0090] Calculate the gradient value and gray-level direction of the filtered image. Edges are sets of pixels with significant gray-level value changes. An edge is the area between a black and a white border, where the gray-level value change is greatest. In an image, the gradient represents the degree and direction of gray-level value change. It can be obtained by dot-multiplying by a Sobel operator or other operators to obtain gradient values g in different directions. x (m,n),g y (m,n). The combined gradient is calculated using the following formulas to determine the gradient value and direction:
[0091]
[0092]
[0093] Where m is the x-coordinate of the current pixel, n is the y-coordinate of the current pixel, x represents the x-axis of the pixel coordinate system, and y represents the y-axis of the pixel coordinate system.
[0094] Based on gradient values, non-maximum values are filtered out. During Gaussian filtering, edges may be amplified. Therefore, it is necessary to filter out points that are not edges, making the edge width as close to one pixel as possible: if a pixel belongs to an edge, then the gradient value of this pixel in the gradient direction is the largest; otherwise, it is not an edge, and its grayscale value is set to 0.
[0095]
[0096] Where M is the gradient value of the current pixel, T is the filtering threshold, and M T This is the new gradient value for the current pixel.
[0097] Edge detection is achieved using upper and lower thresholds. Normally, a single threshold is used, but this can lead to false detections. Therefore, a heuristic approach is used to determine an upper threshold and a lower threshold. Pixels above the lower threshold are considered edges, thus improving accuracy. Specifically, two thresholds are set: `maxVal` and `minVal`. Pixels greater than `maxVal` are detected as edges, while those below `minVal` are detected as non-edges. For intermediate pixels, if they are adjacent to pixels identified as edges, they are considered edges; otherwise, they are non-edges.
[0098] In this embodiment, the role of Canny edge detection is to extract the edges of key components of the turnout outdoor equipment from the images identified by the deep learning model through illumination factor removal and edge detection operations. This is used to calculate component position information data, such as switch machine gap width, switch rail opening, and switch rail creep.
[0099] As an optional implementation method, the data fitting algorithm is as follows:
[0100] Suppose we have a set of data points, containing m points:
[0101] {(x1,y1),(x2,y2),…,(x m ,y m )}, so that a point in this point can be represented as (x i ,y i ), i = 1, 2, 3, ..., m. These sample points can be represented by an nth-degree polynomial:
[0102]
[0103] Among them, the nth degree polynomial has a0 to a n The fitting parameters for these n+1 positions are obtained through data fitting.
[0104] The solution process is as follows:
[0105] x-coordinate of the sample point i Substitute into our assumed polynomial In the given sample point, the ordinate of the nth degree polynomial at the x-coordinate is:
[0106]
[0107] Then we need an indicator to evaluate all of them. and y in the sample points i The difference can be represented by the sum of squares of the errors:
[0108]
[0109] To minimize ∈, we can take the following (n+1)th partial derivative with respect to ∈, and set each partial derivative to 0.
[0110]
[0111] One of the terms in the above equation It can be expanded to read as
[0112]
[0113] The other terms in the above system of equations can also be expanded to include the fitting coefficients a0 to a n The equations are then used to find the (n+1) fitted coefficients. This yields a set of fitted coefficients that minimize ∈ [the set of equations]. Substituting this back into the equations... The expression is then fitted.
[0114] In this embodiment, the data fitting function is to analyze the sensor data collected by the turnout outdoor equipment monitoring device and calculate whether there are any abnormalities in the turnout outdoor equipment that are not visible to the naked eye, such as excessive turnout operating current or excessive / inadequate turnout switching force.
[0115] The monitoring method proposed in this embodiment is used to monitor the outdoor equipment of the turnout in real time. First, the monitoring device collects real-time data on the turnout switch machine and the turnout switch rail. The collected data display interface is as follows: Figure 3 As shown, the left side displays the raw data of the turnout switch machine gap, and the right side displays the raw data of the automatic switch. Then, deep learning algorithms, image processing algorithms, and data fitting algorithms are used to analyze the collected data, outputting the current status of the turnout switch machine and turnout switch rail, as well as any abnormal information. The display interface is as follows. Figure 4 As shown, where, Figure 4 The upper center coordinates are used to display the dimensions of the left and right turnout gaps after the switch machine operates. The vertical axis represents the turnout gap dimensions in mm, and the horizontal axis represents the turnout operation time. The lower left side shows the real-time image of the left turnout gap, and the lower right side shows the real-time image of the right turnout gap. This enables real-time monitoring of the turnout switch machine and turnout switch rails without human intervention, and provides data support for the maintenance of the turnout switch machine.
[0116] This embodiment also proposes a computer device for performing the methods described above in this embodiment.
[0117] Computer devices include a processor, internal memory, and a system bus; various device components, including internal memory and the processor, are connected to the system bus. A processor is hardware used to execute computer program instructions through basic arithmetic and logical operations within the computer system. Internal memory is a physical device used for temporary or permanent storage of computational programs or data (e.g., program state information). The system bus can be any of several types of bus architectures, including a memory bus or memory controller, a peripheral bus, and a local bus. The processor and internal memory can communicate via the system bus. Internal memory includes read-only memory (ROM) or flash memory, and random access memory (RAM), which typically refers to the main memory loaded with the operating system and computer programs.
[0118] Computer devices typically include an external storage device. The external storage device can be selected from a variety of computer-readable media, which are any usable media accessible by a computer device, including both removable and fixed media. Examples of computer-readable media include, but are not limited to, flash memory (microSD cards), CD-ROMs, digital versatile optical discs (DVDs) or other optical disc storage, magnetic tape cartridges, magnetic tapes, disk storage or other magnetic storage devices, or any other media that can be used to store desired information and is accessible by a computer device.
[0119] Computer devices can logically connect to one or more network terminals in a network environment. Network terminals can be personal computers, servers, routers, smartphones, tablets, or other public network nodes. Computer devices connect to network terminals through network interfaces (LAN interfaces). A Local Area Network (LAN) is a computer network interconnected within a limited area, such as a home, school, computer lab, or office building using network media. WiFi and twisted-pair Ethernet are the two most commonly used technologies for building LANs.
[0120] It should be noted that other computer systems, including more or fewer subsystems than computer equipment, are also applicable to the invention.
[0121] As described in detail above, the computer device applicable to this embodiment can perform the specified operations of the switch outdoor equipment monitoring method. The computer device performs these operations through software instructions executed by a processor in a computer-readable medium. These software instructions can be read into memory from a storage device or from another device via a local area network interface. The software instructions stored in memory cause the processor to execute the aforementioned group membership information processing method. Furthermore, the present invention can also be implemented through hardware circuitry or hardware circuitry combined with software instructions. Therefore, implementing this embodiment is not limited to any specific combination of hardware circuitry and software.
[0122] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for monitoring outdoor turnout equipment based on neural networks, characterized in that, The method includes: The monitoring data of each monitoring point of the turnout outdoor equipment is obtained through the turnout outdoor equipment monitoring device. The monitoring data includes image data and digital data. Deep learning algorithms are used to locate and analyze the acquired image data, identify images of key components of the turnout outdoor equipment, and monitor external anomalies of key components of the turnout outdoor equipment. Image processing algorithms are used to process the images of the key components of the turnout outdoor equipment, and the edges of the components are extracted from the images of the key components of the turnout outdoor equipment to perform distance anomaly monitoring of the turnout outdoor equipment. A data fitting algorithm is used to perform data fitting calculations on the acquired digital data to monitor internal anomalies in the outdoor equipment of the turnout. Deep learning algorithms were used to locate and analyze the acquired image data, identifying key components of the turnout's outdoor equipment, specifically including: The image data is used to construct training sample data, which includes imaging of the automatic switch opening and closing area and the turnout gap area. The automatic switch opening and closing device area, key components of the automatic switch opening and closing device, and turnout gaps in the training sample data were labeled using the labeling tool. A deep learning model was trained using labeled training sample data to obtain the positioning models of the automatic switch machine opener, key components of the automatic switch machine opener, and turnout gaps. Using the trained positioning model, the positioning results are output for the automatic switch machine opener and closer, key components of the automatic switch machine opener and closer, and turnout gaps in the real-time acquired image data. The acquired numerical data is fitted using a data fitting algorithm, specifically including: Construct an nth-degree polynomial: ； in, arrive The fitting parameters for an nth-degree polynomial; x-coordinates of sample points Substituting the nth-degree polynomial into the equation, we obtain the ordinate of the nth-degree polynomial at the x-axis: ； The sum of squares of the errors between the ordinate of the nth-degree polynomial at the x-axis and the ordinate of the sample points is calculated: ； Take the (n+1)th partial derivative with respect to the sum of squares of the errors, and set each partial derivative to 0: ； The above system of equations can be expanded to include the fitting coefficients. arrive The equation; The fitting coefficients can be obtained by solving the system of equations containing n+1 equations. arrive Substituting the fitting coefficients into the nth-degree polynomial achieves the fitting result. The digital data includes: turnout operating current, turnout switching force, and oil level in the switch machine.
2. The method for monitoring outdoor turnout equipment based on neural networks according to claim 1, characterized in that, Image processing algorithms are used to process the images of the key components of the turnout outdoor equipment, and the edges of the components are extracted from the images. Specifically, this includes: Remove lighting factors from images of key components of the identified outdoor turnout equipment; An edge detector is used to extract component edges from images of key components of outdoor turnout equipment after removing illumination factors.
3. The method for monitoring outdoor turnout equipment based on neural networks according to claim 2, characterized in that, The formula for calculating the removal of light factors is as follows: ； in, and These are the input image data and the resulting image data after removing illumination factors, respectively. Image location, The formula for calculating the pixel mean of the input image data is as follows: ； in, and These are the width and height of the input image, respectively. The pixel value in the j-th row and i-th column of the input image.
4. The method for monitoring outdoor turnout equipment based on neural networks according to claim 3, characterized in that, An edge detector is used to extract component edges from images of key components of outdoor turnout equipment after removing illumination factors. Specifically, this includes: Apply Gaussian filtering to the image; Calculate the gradient value and gray-level direction of the filtered image; Based on the calculated gradient values, filter out non-maximum values; Edge detection is performed using upper and lower thresholds.
5. The method for monitoring outdoor turnout equipment based on neural networks according to claim 4, characterized in that, The edge detection using upper and lower thresholds specifically includes: Pixels larger than the upper threshold are identified as edges, and pixels smaller than the lower threshold are identified as non-edges. For pixels that are greater than or equal to the lower threshold and less than or equal to the upper threshold, if they are adjacent to pixels identified as edges, they are identified as edges; otherwise, they are non-edges.
6. A turnout outdoor equipment monitoring device based on a neural network, wherein the turnout outdoor equipment monitoring device is applicable to the monitoring method described in any one of claims 1-5, characterized in that, The outdoor equipment monitoring device for turnouts includes turnout switch rail monitoring equipment, monitoring sub-unit, and internal monitoring equipment for switch machines; The internal monitoring equipment of the switch machine is used to monitor the automatic switch, turnout gap, turnout operating current, turnout switching force and oil level inside the switch machine. The switch point rail monitoring equipment is used to monitor the switch point rail opening and the switch point rail creep. The monitoring unit is used to control the turnout switch rail monitoring equipment and the internal monitoring equipment of the switch machine.
7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-5.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-5.