Method and system for monitoring automatic switch of switch machine based on single-pixel cutting algorithm

Abstract

The invention relates to the technical field of data identification, and provides a method and a system for monitoring an automatic switch of a switch machine based on a single-pixel cutting algorithm, and the method comprises the steps: collecting dynamic and static contact images of the automatic switch of the switch machine through a camera; cutting the obtained image through a fast fuzzy clustering algorithm, carrying out edge extraction on the marked image through a single-pixel cutting algorithm, traversing a single-pixel edge image to carry out point switch state data measurement, recording the state data in a work log, and storing the work log in a database. According to the point switch automatic shutter monitoring method based on the single pixel cutting algorithm provided by the invention, the detection efficiency can be improved, the breakthrough progress of the measurement precision can be realized, the qualitative leap of the prediction capability and the environmental adaptability can be obviously enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of data recognition technology, and in particular to a method and system for monitoring an automatic switch door based on a single-pixel cutting algorithm. Background Art

[0002] Safety issues have always been a subject of close attention. With the continuous development of the times, people are faced with a variety of large and complex systems in their daily lives, and high standards and strict requirements are imposed on the reliability and safety of these systems. However, due to the complexity of the system itself, various failures often occur. In the future, with the continuous increase in traffic volume, the impact on the turnout equipment will be more frequent, wear and tear will increase, and the number of diseases will increase. The turnout will also deviate from the close contact and the change of the switch machine switch will be faster than before. These problems pose a great threat to the driving safety of trains. In particular, the automatic switch machine switch requires a higher workload than the previous methods of high-frequency inspection and metering, and will also bring the following problems:

[0003] (1) Environmental aspects. At night, the dim light usually brings great inconvenience to the inspection and maintenance work, and the weather also causes certain troubles for the maintenance of turnout machines.

[0004] (2) The switch machine itself. Since the indicating switch is set inside the switch machine and its position is relatively special, there are some difficulties in observing the indicating switch of the switch machine. The switch parameter setting standard makes it difficult to measure it accurately manually. At the same time, the oil, dust, etc. accumulated inside the switch machine for a long time will also increase the difficulty of observing and measuring the switch inside the switch machine.

[0005] (3) Personnel. Due to lack of experience, low professional knowledge and technical level, as well as carelessness, some mistakes may occur. It often requires multiple maintenance personnel to inspect the same switch machine multiple times to obtain the final result. Summary of the Invention

[0006] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of this application to avoid obscuring the purpose of this section, the abstract and the title of the invention, and such simplifications or omissions should not be used to limit the scope of the present invention.

[0007] In view of the above problems and / or the problems existing in the existing switch machine automatic switch monitoring method based on the single pixel cutting algorithm, the present invention is proposed.

[0008] Therefore, the problem to be solved by the present invention is:

[0009] Dynamically detect the status of the switch machine's automatic switch and display it in real time.

[0010] To solve the above technical problems, the present invention provides the following technical solution: a method for monitoring an automatic switch switch based on a single-pixel cutting algorithm, comprising:

[0011] Use the camera to collect the images of the dynamic and static contacts of the automatic switch;

[0012] The acquired image is segmented by a fast fuzzy clustering algorithm, and the segmented image is marked;

[0013] The edge of the marked image is extracted using a single-pixel cutting algorithm to obtain a single-pixel edge image of the marked image;

[0014] Traversing the edge image of a single pixel to measure the state data of the switch machine, the state data includes the driving depth data and the shaking amount data during the vehicle passing, the fastening state data of the dynamic and static contact seats, and the fastening state data of the dynamic contact crank shaft adjustment screw;

[0015] The status data is recorded in a work log, and the work log is stored in a database.

[0016] As a preferred solution of the switch machine automatic switch monitoring method based on the single pixel cutting algorithm of the present invention, wherein: the obtained image is cut by the fast fuzzy clustering algorithm to obtain the single pixel edge image of the marked image, which includes:

[0017] Spatial Constrained Fast Fuzzy Clustering Function Algorithm:

[0018]

[0019] Where N is the total number of pixels, C is the number of cluster centers, and u ij is the membership degree of pixel i to cluster j, m∈(1,∞) is the fuzzy weight index, x i is the feature vector of the i-th pixel, v j is the jth cluster center, ‖‖ is the Euclidean distance, α is the original weight coefficient, β is the spatial neighborhood penalty coefficient, Ω i is the neighborhood window of pixel i, w ik is the neighborhood similarity weight based on Gaussian kernel, J SC-FCM In it, FCM represents clustering function algorithm, and SC represents spatial constraint;

[0020] The obtaining of a single pixel edge image of the marked image comprises: calculating J SC-FCM After that, for u ij With v j Iterative update, u ij With v j Mark.

[0021] As a preferred solution of the switch machine automatic switch monitoring method based on the single pixel cutting algorithm of the present invention, wherein: using the single pixel cutting algorithm to extract the edge of the marked image to obtain the single pixel edge image of the marked image includes:

[0022]

[0023] Where x is the input image matrix, k is the slope coefficient, and λ s is the multi-scale weight, Z is the normalization factor, ReLU(x) is the linear rectification function, T s，high With T s，low They are the high threshold and low threshold under the Gaussian smoothing kernel scale s, s∈{1, 2, 4} represents the scale parameter of the Gaussian smoothing kernel, where {1, 2, 4} represents the three scale parameters of 1×1, 2×2, and 4×4, G s (x) represents the image gradient amplitude;

[0024] The amplitude variation range of the image gradient amplitude is:

[0025] When G s (x)>T s，high When is directly marked as a strong edge, ReLU outputs a positive value;

[0026] When T s，high >G s (x)>T s，low When weighted by the sigmoid function, it is determined whether it is connected to the strong edge;

[0027] When G s (x)<T s，low When cutting is suppressed, single-pixel edge image acquisition is cancelled;

[0028] The sigmoid function is a curve function.

[0029] As a preferred solution of the switch machine automatic switch monitoring method based on the single pixel cutting algorithm of the present invention, wherein: the switch machine state data measurement by traversing the single pixel switch edge image includes:

[0030] Using machine vision methods to detect the penetration depth of the switch machine's dynamic and static contacts requires locating the dynamic and static contact areas of the switch machine and obtaining the switch machine status data through secondary positioning.

[0031] As a preferred solution of the switch machine automatic switch monitoring method based on the single pixel cutting algorithm of the present invention, the secondary positioning method includes:

[0032] First, perform coarse positioning to obtain the mask and prediction box of the moving contact area;

[0033] Secondly, fine positioning is performed, and the image preprocessing of the moving contact area obtained by rough positioning is performed to obtain the coordinates and diameter of the moving contact circle center;

[0034] Finally, the subgraph is intercepted according to the coordinates of the center of the moving contact circle, and then the static contact area in the subgraph is re-positioned.

[0035] In view of the above problems and / or the problems existing in the existing switch machine automatic switch monitoring method based on the single pixel cutting algorithm, the present invention is proposed.

[0036] Therefore, the problem to be solved by the present invention is:

[0037] Dynamically detect the status of the switch machine's automatic switch and display it in real time.

[0038] To solve the above technical problems, the present invention provides the following technical solutions: a switch automatic switch monitoring system based on a single pixel cutting algorithm, comprising:

[0039] Network extensions, data concentrators and communication hosts;

[0040] The network extension includes an image acquisition module, an image cutting module, an image edge extraction module and a data analysis module;

[0041] The image acquisition module uses an image acquisition sensor installed inside the switch machine to acquire an image of the switch machine's automatic opener and closer, and transmits the acquired image to the image cutting module;

[0042] The image segmentation module is equipped with a single-pixel segmentation algorithm that can segment the collected image information, mark the segmented image, and pass the marked image to the image edge extraction module;

[0043] The image edge extraction module uses a single-pixel cutting algorithm to extract the edge of the marked image, obtains a single-pixel edge image of the marked image, and transmits the single-pixel edge image to the data analysis module;

[0044] The data analysis module traverses the edge image of a single pixel to measure the state data of the switch machine and transmits the measured data to the data concentrator; the image acquisition module, image cutting module, image edge extraction module and data analysis module are sequentially connected in the network extension;

[0045] The data concentrator includes a central unit that collects the measured data of each network extension and transmits the signal to the communication host; the communication host sends the data to the switch monitoring station through Ethernet and CAN bus;

[0046] The communication host sends the data to the monitoring station through Ethernet and CAN bus, and displays the data through the display;

[0047] The CAN bus is a bus method that uses twisted-pair cables to transmit signals.

[0048] As a preferred solution of the switch automatic switch monitoring system based on the single pixel cutting algorithm of the present invention, the data concentrator determines the monitoring result of the switch automatic switch through the measurement data of the data analysis module, and transmits the signal to the communication host using the ADSL data transmission method;

[0049] The ADSL refers to Asymmetric Digital Subscriber Line.

[0050] As a preferred solution of the switch machine automatic switch monitoring system based on the single-pixel cutting algorithm described in the present invention, the communication host sends data to the switch monitoring station through Ethernet and CAN bus, and finally realizes real-time monitoring and display of the status of each switch machine Zidong switch in the indoor host computer.

[0051] The present invention provides the following technical solution: an electronic device comprising:

[0052] one or more processors;

[0053] a storage device having one or more programs stored thereon;

[0054] When the one or more programs are executed by the one or more processors, the one or more processors implement a method for monitoring an automatic switch switch based on a single-pixel cutting algorithm.

[0055] The present invention provides the following technical solution: an electronic device comprising:

[0056] A computer-readable storage medium stores executable instructions, which, when executed by a processor, enable the processor to implement a method for monitoring an automatic switch opener and closer based on a single-pixel cutting algorithm.

[0057] The present invention proposes a switch machine automatic switch monitoring method and system based on a single-pixel cutting algorithm, which combines a fast fuzzy clustering algorithm with a single-pixel cutting algorithm to achieve the switch machine automatic switch monitoring effect, thereby significantly improving the accuracy, robustness and monitoring efficiency of switch machine switch detection when the data volume is insufficient and the features are complex. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0059] in:

[0060] Figure 1 This is an operational flow chart of the switch machine automatic switch monitoring method based on the single-pixel cutting algorithm in Example 1.

[0061] Figure 2 This is an operation flow chart of the switch automatic switch monitoring system based on the single-pixel cutting algorithm in Example 2.

[0062] Figure 3 The results of using the Canny operator in the edge cutting algorithm of the switch automatic switch monitoring system based on the single-pixel cutting algorithm in Example 3 are compared with images of other traditional methods. DETAILED DESCRIPTION

[0063] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0064] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0065] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0066] Example 1

[0067] Reference Figure 1 , which is the first embodiment of the present invention, provides a method for monitoring an automatic switch switch based on a single-pixel cutting algorithm, comprising:

[0068] S1. Data collection and annotation

[0069] Cameras capture images of the static and dynamic contacts of the switch's automatic switch. Image sensors capture images or videos of the switches, reflecting changes in the switches' movement. One image sensor is installed in both the fixed and reverse positions. The core component of the image sensor is a miniature camera with megapixels or higher. Compared to photoelectric, inductive, and mechanical displacement sensors, image sensors offer greater stability. The image sensor is illuminated by white LEDs and is installed inside the switch.

[0070] S2. Data cutting and processing

[0071] The acquired image is cut by a fast fuzzy clustering algorithm, the cut image is marked, and the edge of the marked image is extracted by a single-pixel cutting algorithm to obtain a single-pixel edge image of the marked image;

[0072] S201, segmenting the acquired image using a fast fuzzy clustering algorithm to obtain a single-pixel edge image of the marked image includes:

[0073] Spatial Constrained Fast Fuzzy Clustering Function Algorithm:

[0074]

[0075] Where N is the total number of pixels, C is the number of cluster centers, and u ij is the membership degree of pixel i to cluster j, m∈(1,∞) is the fuzzy weight index, x i is the feature vector of the i-th pixel, v j is the jth cluster center, ‖‖ is the Euclidean distance, α is the original weight coefficient, β is the spatial neighborhood penalty coefficient, Ω i is the neighborhood window of pixel i, w ik is the neighborhood similarity weight based on Gaussian kernel, J SC-FCM In it, FCM represents clustering function algorithm, and SC represents spatial constraint;

[0076] Obtaining a single-pixel edge image of a labeled image includes, in calculating J SC-FCM After that, for u ij With v j Iterative update, u ij With v j Mark. Objective function J SC-FCM The smaller the value, the closer the spatial-feature distribution of the cluster center and the pixels is.

[0077] S202, using a single-pixel cutting algorithm to extract edges from the marked image, to obtain a single-pixel edge image of the marked image, including:

[0078]

[0079] Where x is the input image matrix, k is the slope coefficient, and λ s is the multi-scale weight, Z is the normalization factor, ReLU(x) is the linear rectification function, T s，high With T s，low They are the high threshold and low threshold under the Gaussian smoothing kernel scale s, s∈{1, 2, 4} represents the scale parameter of the Gaussian smoothing kernel, where {1, 2, 4} represents the three scale parameters of 1×1, 2×2, and 4×4, G s (x) represents the image gradient amplitude;

[0080] The calculation method of the internal function is:

[0081]

[0082] where μ s and σ s are the mean and standard deviation of the step amplitude under scale s, S x With S y They are horizontal and vertical convolution kernels respectively, * represents convolution operation, W s is a Gaussian smoothing kernel of scale s, a single-pixel edge image E(x)∈[0,1], and the closer the value is to 1, the more significant the edge is.

[0083] The range of image gradient amplitude is:

[0084] When G s (x)>T s，high When is directly marked as a strong edge, ReLU outputs a positive value;

[0085] When T s，high >G s (x)>T s，low When weighted by the sigmoid function, it is determined whether it is connected to the strong edge;

[0086] When G s (x)<T s，low When cutting is suppressed, single-pixel edge image acquisition is cancelled;

[0087] The sigmoid function is a curve function.

[0088] Using machine vision methods to detect the penetration depth of the switch machine's dynamic and static contacts requires locating the dynamic and static contact areas of the switch machine and obtaining the switch machine status data through secondary positioning.

[0089] Secondary positioning methods include:

[0090] First, perform coarse positioning to obtain the mask and prediction box of the moving contact area;

[0091] Secondly, fine positioning is performed, and the image preprocessing of the moving contact area obtained by rough positioning is performed to obtain the coordinates and diameter of the moving contact circle center;

[0092] Finally, the subgraph is intercepted according to the coordinates of the center of the moving contact circle, and then the static contact area in the subgraph is re-positioned.

[0093] Example 2

[0094] Reference Figure 2 , which is a second embodiment of the present invention, a switch automatic switch monitoring system based on a single pixel cutting algorithm includes a network extension 100, a data concentrator 200 and a communication host 300;

[0095] The network extension 100 includes an image acquisition module 101, an image segmentation module 102, an image edge extraction module 103, and a data analysis module 104. Installed at each switch switch monitoring point, the network extension determines the operating current of the switch motor and records the switch actuation video. Comprising a network bridge, a power supply, and an extension box, the network extension compresses and encodes the video image through a current transformer. It then automatically captures a static image of the switch, obtaining the switch's position and offset. The network extension incorporates internal memory, which allows it to upload image information, temperature and humidity information, and vibration acceleration data to the communication host for storage at the switch monitoring station.

[0096] The image acquisition module 101 uses an image acquisition sensor installed inside the switch machine to acquire an image of the switch machine's automatic switch and transmit the acquired image to the image cutting module 102. The image acquisition sensor is used to capture images or videos of the switch and reflect the changes in the switch. An image acquisition sensor is installed in the fixed and reverse positions. The core part of the image acquisition sensor is a miniature camera with millions of pixels or more. Compared with photoelectric sensors, inductive sensors, displacement mechanical sensors, etc., the image acquisition sensor has higher stability. The image acquisition sensor uses white LED lighting and is installed inside the switch machine.

[0097] The image cutting module 102 is equipped with a single pixel cutting algorithm that can cut and process the collected image information, mark the cut image, and pass the marked image to the image edge extraction module 103;

[0098] The image edge extraction module 103 uses a single-pixel cutting algorithm to extract the edge of the marked image, obtains a single-pixel edge image of the marked image, and transmits the single-pixel edge image to the data analysis module 104;

[0099] The data analysis module 104 traverses the edge image of a single pixel to measure the state data of the switch machine and transmits the measured data to the data concentrator 200; the image acquisition module 101, the image segmentation module 102, the image edge extraction module 103 and the data analysis module 104 are sequentially connected in the network extension 100;

[0100] The data concentrator 200 includes a device that collects the measured data of each network extension and transmits the signal to the communication host 300; the communication host sends the data to the switch monitoring station through Ethernet and CAN bus;

[0101] The communication host 300 sends the data to the monitoring station through Ethernet and CAN bus, and displays the data through the display;

[0102] The CAN bus is a bus method that uses twisted pair cables to transmit signals.

[0103] The data concentrator 200 uses the measurement data of the data analysis module 104 to determine the monitoring results of the switch automatic switch, and uses the ADSL data transmission method to transmit the signal to the communication host 300;

[0104] ADSL stands for Asymmetric Digital Subscriber Line.

[0105] The communication host 300 transmits data to the switch monitoring station via Ethernet and the CAN bus. Finally, the indoor host computer monitors and displays the status of each switch's Zidong switch in real time. The system features large storage capacity and utilizes an industrial control computer, primarily consisting of a display, network host, UPS, switch, lightning protection devices, and cabinets. This allows for extended storage of real-time data. The industrial control computer connects to each host via network cables. Support for multiple transmission methods, including ADSL, cable, fiber optic, and power carrier, is required. Power is also provided to the front-end switch monitoring substations, as well as various sensors, such as image acquisition sensors and temperature and humidity sensors.

[0106] Example 3

[0107] refer to Figure 3 In a third embodiment of the present invention, a switch machine automatic switch monitoring system based on a single pixel cutting algorithm includes, in one embodiment, attributes of the collected data include:

[0108] To verify the innovativeness of the switch monitoring method of the present invention, a test platform covering railway field environment simulation was built. The test system includes the following components:

[0109] Hardware configuration:

[0110] Image acquisition module: 8 industrial-grade high-speed cameras (Basler acA2440-75um, resolution 2448×2048, frame rate 75fps), installed around the switch machine contact group, covering a viewing angle of 0° to 270°.

[0111] Processing terminal: server cluster equipped with NVIDIA A100 GPU and deep learning accelerator card.

[0112] Test samples: 120 sets of switch machine switches (including 60 sets in normal and abnormal states) collected from 6 railway hubs, covering complex working conditions such as dust, rain, fog, and night.

[0113] Implementation process:

[0114] Image acquisition: During a simulated train passage (vibration frequency 5-15Hz), multiple cameras were synchronously triggered to capture images of the moving and static joints, capturing 30 frames per motion cycle (exposure time 50μs). The raw images were preprocessed (including illumination compensation and motion blur correction) to generate a standard 2000×2000 pixel input.

[0115] Fast Fuzzy Clustering Cutting: This algorithm uses an improved FCM algorithm, sets the initial number of clusters K to 5, and optimizes the number of iterations using dynamic entropy. Experiments show that compared to traditional FCM, the number of iterations is reduced from an average of 18 to 9 (convergence threshold ε = 1e-5), and the cutting time is shortened from 45ms / frame to 22ms / frame.

[0116] Edge extraction optimization: The anti-interference Canny operator is applied to the marked sub-image, combined with morphological closing operations to fill broken edges. An adaptive dual-threshold mechanism is introduced: when the image signal-to-noise ratio (SNR) is less than 15dB, the high threshold is automatically increased to 0.8 times the grayscale range.

[0117] State parameter measurement:

[0118] Penetration depth: Using sub-pixel edge positioning technology, the overlapping distance between the moving contact and the static contact notch is measured with an accuracy of 0.02mm.

[0119] Sway analysis: In the vibration environment of a passing vehicle, the displacement trajectory of the characteristic points of the contact group is tracked and the standard deviation is calculated as the sway index.

[0120] Tightening status detection: Determine whether loosening occurs based on the Fourier descriptor matching degree of the bolt profile (threshold > 0.92).

[0121] Data recording and early warning: The measurement results are compared with the preset safety thresholds (insertion depth ≥4.5mm, shaking amount ≤0.15mm, tightening matching degree ≥0.95), and abnormal data is pushed to the operation and maintenance terminal in real time.

[0122] Comparison plan:

[0123] Traditional manual inspection: Physical measurements are performed using tools such as feeler gauges and torque wrenches, taking 25-40 minutes per device.

[0124] Existing machine vision solutions: Switch machine automatic switch monitoring method based on single pixel cutting algorithm, such as Figure 3 As shown, (a) is the result of using the canny operator of the present invention in the lower edge cutting algorithm of the present invention, (b) is the result of using the prewitt operator in the lower edge cutting algorithm of the present invention, (c) is the result of using the Sobel operator in the lower edge cutting algorithm of the present invention, and (d) is the result of using the Roberts operator in the lower edge cutting algorithm of the present invention, wherein the canny operator, prewitt operator, Sobel operator, and Roberts operator are four different types of differential operators for image cutting and have different algorithm logics. Figure 3 The present invention uses the canny operator and the other three types of operators to compare the monitoring result images based on the visual solution of the present invention.

[0125] It can be seen from this that the canny operator used in the present invention has relatively higher detection accuracy and is more consistent with the operator optimization result of the present invention.

[0126] Table 1: Comparison of detection efficiency (unit: seconds / unit)

[0127]

[0128] Table 2: Comparison of detection accuracy (unit: %)

[0129]

[0130] Table 3: Fault prediction capability (unit: times)

[0131]

[0132] Table 4: System stability test (unit: %)

[0133]

[0134] Experimental data show that the technical solution of the present invention is significantly superior to the existing technology in terms of multi-dimensional performance indicators:

[0135] Improved detection efficiency: Table 1 shows the efficiency improvement ratio for the detection effect comparison between the present invention and the existing visual solution. As shown in Table 1, the present invention shortens the comprehensive detection time from 326 seconds for manual detection to 18 seconds per unit (an increase of 94.5%), which is 76.9% higher than the existing visual solution. This is due to the optimization of the fast fuzzy clustering algorithm of the present invention - by adjusting the cluster center through dynamic entropy value, the image cutting time is reduced by 51.1% (from 45ms to 22ms). The multi-camera collaborative acquisition strategy increases the data density to 137 monitoring points / time, far exceeding the single-view detection of the existing solution (an average of 16 points / time).

[0136] Breakthrough Measurement Accuracy: Table 2 shows that the present invention achieves absolute accuracy improvements of 9.2% to 18.8% across key metrics. The accuracy rate for driving depth measurements with an error of ≤0.1mm reaches 98.7%, a 16.4 percentage point improvement over manual inspection. This is due to the innovative application of single-pixel edge extraction technology in the present invention: by combining an improved Canny operator with morphological closing operations, edge localization accuracy reaches 0.02 pixels (equivalent to a physical resolution of 0.0015mm).

[0137] A qualitative leap in predictive capabilities: As shown in Table 3, the present invention provides 4.5 times the number of early warnings for progressive bolt loosening compared to manual inspections, detecting potential problems an average of 120 hours in advance. This demonstrates the effectiveness of the present invention's sway data analysis method—combining a composite criterion based on Fourier descriptor matching and sway standard deviation—to boost the early fault detection rate to 97.4% (compared to 88.2% for existing solutions).

[0138] Significantly Enhanced Environmental Adaptability: Table 4 demonstrates that, even under rain and fog obstruction, the proposed method maintains 89.2% availability (compared to 58.7% for existing solutions). This is attributed to the innovative image preprocessing stage of the proposed method: the use of an illumination compensation algorithm (which extends the dynamic range to 120dB) and a motion blur correction model (which improves PSNR to 38.6dB) enables the system to achieve data integrity exceeding 97.3% under adverse conditions.

[0139] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for monitoring the automatic switch of a switch machine based on a single pixel cutting algorithm, characterized in that: The method comprises: Use the camera to collect the images of the dynamic and static contacts of the automatic switch; The acquired image is segmented by a fast fuzzy clustering algorithm, and the segmented image is marked; The edge of the marked image is extracted using a single-pixel cutting algorithm to obtain a single-pixel edge image of the marked image; Traversing the edge image of a single pixel to measure the state data of the switch machine, the state data includes the driving depth data and the shaking amount data during the vehicle passing, the fastening state data of the dynamic and static contact seats, and the fastening state data of the dynamic contact crank shaft adjustment screw; The status data is recorded in a work log, and the work log is stored in a database.

2. The method for monitoring the automatic switch of a switch machine based on a single pixel cutting algorithm according to claim 1, characterized in that: The step of segmenting the acquired image by a fast fuzzy clustering algorithm to obtain a single-pixel edge image of the marked image includes: Spatial Constrained Fast Fuzzy Clustering Function Algorithm: Where N is the total number of pixels, C is the number of cluster centers, and u ij is the membership degree of pixel i to cluster j, m∈(1,∞) is the fuzzy weight index, x i is the feature vector of the i-th pixel, v j is the jth cluster center, ‖‖ is the Euclidean distance, α is the original weight coefficient, β is the spatial neighborhood penalty coefficient, Ω i is the neighborhood window of pixel i, w ik is the neighborhood similarity weight based on Gaussian kernel, J SC-FCM In it, FCM represents clustering function algorithm, and SC represents spatial constraint; The obtaining of a single pixel edge image of the marked image comprises: calculating J SC-FCM After that, for u ij With v j Iterative update, u ij With v j Mark.

3. The method for monitoring the automatic switch of a switch machine based on a single pixel cutting algorithm according to claim 1, characterized in that: The method of extracting the edge of the marked image by using a single-pixel cutting algorithm to obtain a single-pixel edge image of the marked image includes: Where x is the input image matrix, k is the slope coefficient, and λ s is the multi-scale weight, Z is the normalization factor, ReLU(x) is the linear rectification function, T s，high With T s，low They are the high threshold and low threshold under the Gaussian smoothing kernel scale s, s represents the scale parameter of the Gaussian smoothing kernel, where {1, 2, 4} represents the three scale parameters of 1×1, 2×2, and 4×4, G s (x) represents the image gradient amplitude; The amplitude variation range of the image gradient amplitude is: When G s (x)>T s，high When is directly marked as a strong edge, ReLU outputs a positive value; When T s，high >G s (x)>T s，low When weighted by the sigmoid function, it is determined whether it is connected to the strong edge; When G s (x)<T s，low When cutting is suppressed, single-pixel edge image acquisition is cancelled; The sigmoid function is a curve function.

4. The method for monitoring the automatic switch of a switch machine based on a single pixel cutting algorithm according to claim 1, wherein: The method of traversing the edge image of a single pixel to measure the state data of the switch machine includes: Using machine vision methods to detect the penetration depth of the switch machine's dynamic and static contacts requires locating the dynamic and static contact areas of the switch machine and obtaining the switch machine status data through secondary positioning.

5. The method for monitoring the automatic switch of a switch machine based on a single pixel cutting algorithm according to claim 4, characterized in that: The secondary positioning method includes: First, perform coarse positioning to obtain the mask and prediction box of the moving contact area; Secondly, fine positioning is performed, and the image preprocessing of the moving contact area obtained by rough positioning is performed to obtain the coordinates and diameter of the moving contact circle center; Finally, the subgraph is intercepted according to the coordinates of the center of the moving contact circle, and then the static contact area in the subgraph is re-positioned.

6. The switch machine automatic switch monitoring system based on single pixel cutting algorithm is characterized by: The system includes a network extension (100), a data concentrator (200) and a communication host (300); The network extension (100) includes an image acquisition module (101), an image segmentation module (102), an image edge extraction module (103) and a data analysis module (104); The image acquisition module (101) uses an image acquisition sensor installed inside the switch machine to acquire an image of the switch machine's automatic switch and transmits the acquired image to the image cutting module (102); The image cutting module (102) is equipped with a single-pixel cutting algorithm capable of cutting and processing the collected image information, marking the cut image, and transmitting the marked image to the image edge extraction module (103); The image edge extraction module (103) uses a single-pixel cutting algorithm to extract the edge of the marked image, obtains a single-pixel edge image of the marked image, and transmits the single-pixel edge image to the data analysis module (104); The data analysis module (104) traverses the edge image of a single pixel to measure the state data of the switch machine and transmits the measured data to the data concentrator (200); the image acquisition module (101), the image cutting module (102), the image edge extraction module (103) and the data analysis module (104) are sequentially connected in the network extension (100); The data concentrator (200) includes a device for concentrating the measurement data of each network extension and transmitting the signal to the communication host (300); The communication host sends data to the switch monitoring station through Ethernet and CAN bus; The communication host (300) sends data to the monitoring station through Ethernet and CAN bus, and displays the data through a display; The CAN bus is a bus method that uses twisted-pair cables to transmit signals.

7. The switch automatic switch monitoring system based on the single pixel cutting algorithm according to claim 6, characterized in that: The data concentrator (200) determines the monitoring result of the switch automatic switch through the measurement data of the data analysis module (104), and transmits the signal to the communication host (300) using the ADSL data transmission method; The ADSL refers to Asymmetric Digital Subscriber Line.

8. The switch automatic switch monitoring system based on the single pixel cutting algorithm according to claim 6, characterized in that: The communication host (300) sends data to the switch monitoring station through Ethernet and CAN bus, and finally realizes real-time monitoring and display of the status of each switch machine Zidong switch in the indoor host computer.

9. An electronic device comprising: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 5.

10. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to implement the method according to any one of claims 1 to 5.

Images