HOG-SVM-based railway tunnel vibration signal detection method and system

Abstract

The invention discloses an HOG-SVM-based railway tunnel vibration signal detection method and system. The method comprises the following steps: acquiring tunnel portal foreign matter intrusion detection sensing data in real time by using a foreign matter sentry sensing terminal; the holographic guard sensing terminal is used for performing intelligent illumination in the tunnel and acquiring intelligent risk monitoring sensing data in real time; performing real-time analysis and calculation on the sensing data, detecting internal and external risks of the tunnel, and performing gain amplification and filtering noise reduction processing on vibration signal data; the HOG-SVM algorithm is adopted to extract vibration signal features and perform vibration signal classification detection so as to analyze the frequency spectrum of the vibration signals to invert the position of the train, and the abnormal vibration signals of the single sensor are utilized to judge the lamp looseness risk detection state; according to the invention, intelligent monitoring and analysis of tunnel risks and equipment and facility states can be realized according to monitoring and analysis of railway tunnel vibration signals, requirements for on-site inspection personnel are reduced, time required for maintenance and overhaul is shortened, and tunnel diseases and risks can be found as soon as possible.

Description

Technical Field

[0001] The present invention belongs to the technical field of railway tunnel detection, and particularly relates to a railway tunnel vibration signal detection method and system based on HOG-SVM. Background Art

[0002] The safety monitoring of railway tunnels is an important link to ensure railway transportation safety. Traditional railway tunnel detection methods mainly rely on manual inspections and some simple sensor technologies, and these methods have many limitations. For example, manual inspections are time-consuming and laborious, and cannot monitor the dynamic changes in the tunnel in real time. In addition, traditional sensor technologies also have deficiencies in detection accuracy and coverage.

[0003] With the development of technology, railway tunnel monitoring systems based on multi-source perception and intelligent analysis have been gradually developed. These systems can achieve real-time monitoring of various risk factors in the tunnel by integrating multiple sensors, such as vibration sensors, temperature sensors, light sensors, etc. For example, vibration signals are used to monitor the impact of passing trains on the tunnel structure, and environmental changes in the tunnel are monitored through temperature and light sensors.

[0004] However, there are still some problems in existing railway tunnel monitoring systems when dealing with complex vibration signals. For example, vibration signals often contain a large amount of noise, which will affect the extraction of signal features and the accuracy of classification. In addition, for the detection of the loose state of devices such as tunnel lights, traditional monitoring methods often require a large number of sensors and complex algorithm support. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a railway tunnel vibration signal detection method and system based on HOG-SVM.

[0006] To achieve the above object, the present invention is implemented by the following technical solutions:

[0007] In the first aspect, the present invention provides a railway tunnel vibration signal detection method based on HOG-SVM, including the following steps:

[0008] Utilize a number of foreign object sentry perception terminals installed on the side walls on both sides of the railway tunnel entrance to obtain foreign object intrusion detection perception data at the tunnel entrance in real time;

[0009] Utilize a number of holographic guard perception terminals installed at intervals and staggered on both sides of the railway tunnel to obtain intelligent risk monitoring perception data for intelligent lighting in the tunnel in real time;

[0010] The edge computing terminal performs real-time analysis, calculation, and detection of internal and external risks in the tunnel on the perception data transmitted by the holographic guard perception terminal and the foreign object sentry perception terminal. It utilizes the vibration signal generated when the train runs on the railway, and processes the vibration signal data through gain amplification and filter noise reduction;

[0011] The HOG-SVM algorithm is adopted to extract the vibration signal features and classify and detect the vibration signals, so as to analyze the spectrum of the vibration signal to invert the position of the train, and utilize the abnormal vibration signal of a single sensor to judge the detection status of the lamp loosening risk.

[0012] Furthermore, the holographic guard perception front end uses the configured sensors to collect the vibration signals in the railway tunnel and analyze the train and lamp loosening conditions. Utilizing the vibration signal generated when the train runs on the railway, after processing the vibration signal data through gain amplification and filter noise reduction, the HOG-SVM algorithm is adopted to extract the vibration signal features and classify and detect the vibration signals, so as to analyze the spectrum of the vibration signal to invert the position of the train, and utilize the abnormal vibration signal of a single sensor to judge the lamp loosening state.

[0013] Furthermore, the steps for extracting the vibration signal features by using the HOG-SVM algorithm include:

[0014] Utilize the HOG histogram of oriented gradients to extract the gradient information of the time-frequency spectrum diagram of the vibration signal. The gradient at the edge of the time-frequency spectrum diagram is relatively strong, and the extracted time-frequency spectrum diagram of the vibration signal is grayscale and normalized. The formula is as follows:

[0015]

[0016] In the formula, I(x, y) represents the grayscale pixel of the time-frequency spectrum diagram, and G(·) represents the grayscale processing function;

[0017] Perform convolution operation processing using the first-order template gradient operator. The formula for the pixel gradient and direction in the frequency spectrum image is:

[0018]

[0019] In the formula, θ(x, y) and m(x, y) respectively represent the gradient direction and amplitude of the pixel point;

[0020] Divide the vibration time-frequency spectrum diagram of the railway tunnel into pixel blocks block of size M×N, and divide each block into square cells of size a×a according to the block. Combine the calculated gradient directions above to perform weighted projection on the pixels within the cell. After the projection processing, normalize the gradient of the time-frequency spectrum diagram in units of block to reduce the influence of interference factors on the HOG feature extraction. The formula is:

[0021]

[0022] where x n represents the pixel block vector, and η represents a constant;

[0023] After the regularization process is completed, connecting all the pixel blocks completes the HOG features of the railway tunnel vibration signal.

[0024] Furthermore, based on the HOG feature extraction results obtained by using the HOG-SVM algorithm, a support vector machine (SVM) is used to construct a vibration signal classification hyperplane to complete the classification monitoring of railway tunnel vibration signals, including the steps:

[0025] Construct the objective function for SVM classification, which is expressed by the formula:

[0026]

[0027] where w represents the hyperplane weight, C represents the penalty parameter, and ξ i represents the slack variable;

[0028] According to the classification mechanism of SVM, transform the objective function for SVM classification constructed into a convex optimization problem, which is expressed by the formula:

[0029]

[0030] And obtain the dual expression of the convex optimization function as:

[0031]

[0032] Solve the quadratic optimization problem of the dual expression of the convex optimization function to obtain the optimal solution:

[0033]

[0034] According to the optimal solution result, construct the weight w of the classification hyperplane, which is expressed by the formula:

[0035]

[0036] When the optimal solution component satisfies Then the bias of the SVM classification hyperplane can be calculated:

[0037]

[0038] According to the calculation result of the bias of the SVM classification hyperplane, construct the optimal SVM classification hyperplane:

[0039] w * ·x + b *= 0;

[0040] Under the support of the optimal classification hyperplane, a decision function for classifying and monitoring the vibration signals of rotating machinery is constructed:

[0041] f(x) = sign(w * ·x + b * ).

[0042] Furthermore, the foreign object sentry sensing terminal monitors the intrusion of foreign objects at the tunnel entrance based on 3D point clouds and intelligent images, including the steps of:

[0043] By obtaining the 3D point cloud data of the monitoring area in real time, a point cloud background of the monitoring area is established. The monitoring area includes the tunnel entrance, the track surface, and the catenary;

[0044] Analyze and calculate the newly scanned point cloud at the current moment and the background point cloud at the previous moment to determine whether there are foreign objects;

[0045] And link the camera of the foreign object sentry sensing terminal to capture and analyze, and use the point cloud and image to jointly calibrate the coordinates of the foreign object in the point cloud and the coordinates of the foreign object in the image and mark the foreign object in the image;

[0046] At the same time, use the fusion perception technology to filter the normal targets of trains and personnel, and finally upload the foreign object alarm pictures to the edge computing terminal.

[0047] In a second aspect, the present invention provides a railway tunnel vibration signal detection system based on HOG-SVM, including the following modules:

[0048] A foreign object sentry module, which is used to use a number of foreign object sentry sensing terminals installed on the side walls on both sides of the railway tunnel entrance to obtain the perception data of foreign object intrusion detection at the tunnel entrance in real time;

[0049] A holographic guard module, which is used to use a number of holographic guard sensing terminals installed at intervals and staggered on both sides of the railway tunnel to obtain the intelligent lighting in the tunnel and the perception data of intelligent risk monitoring in real time;

[0050] A data processing module, which is used for the edge computing terminal to analyze, calculate, and detect the internal and external risks in the tunnel in real time for the perception data transmitted by the holographic guard sensing terminal and the foreign object sentry sensing terminal, and use the vibration signal of the train running on the railway to perform gain amplification and filter noise reduction processing on the vibration signal data;

[0051] A signal judgment module, which is used to extract the vibration signal features and perform vibration signal classification detection by using the HOG-SVM algorithm to analyze the spectrum of the vibration signal to invert the position of the train, and use the abnormal vibration signal of a single sensor to judge the detection status of the lamp loosening risk.

[0052] Further, the signal judgment module is also used to collect vibration signals in the railway tunnel by using configured sensors and analyze the detection of the looseness of trains and lamps. Using the vibration signals generated when the train travels on the railway, after performing gain amplification and filter noise reduction processing on the vibration signal data, the HOG-SVM algorithm is used to extract the vibration signal features and classify and detect the vibration signals, so as to analyze the spectrum of the vibration signal to invert the position of the train, and use the abnormal vibration signals of a single sensor to judge the looseness state of the lamp.

[0053] Further, the signal judgment module is also used to extract the vibration signal features by using the HOG-SVM algorithm, including the steps of:

[0054] When using the HOG histogram of oriented gradients to extract the gradient information of the time-frequency spectrum of the vibration signal, the gradient at the edge of the time-frequency spectrum is relatively strong, and the extracted time-frequency spectrum of the vibration signal is grayscale and normalized. The formula is as follows:

[0055]

[0056] In the formula, I(x, y) represents the grayscale pixel of the time-frequency spectrum, and G(·) represents the grayscale processing function;

[0057] Perform convolution operation processing using the first-order template gradient operator. The formula for the pixel gradient and direction in the frequency spectrum image is:

[0058]

[0059] In the formula, θ(x, y ) , m(x, y) respectively represent the gradient direction and amplitude of the pixel point;

[0060] Divide the time-frequency spectrum map of the railway tunnel vibration into pixel blocks block of size M×N, and divide the pixel blocks block into square cells of size a×a according to the pixel blocks block. Combine the calculated gradient directions above to perform weighted projection on the pixels within the cell. After the projection processing is completed, normalize the gradient of the time-frequency spectrum map in units of pixel blocks block to reduce the influence of interference factors on the HOG feature extraction. The formula is:

[0061]

[0062] In the formula, x n represents the pixel block vector, and η represents a constant;

[0063] After the regularization processing is completed, connect all the pixel blocks block to complete the HOG features of the railway tunnel vibration signal.

[0064] Further, the signal judgment module is further configured to use the HOG-SVM algorithm to construct a vibration signal classification hyperplane based on the HOG feature extraction result extracted by the support vector machine SVM to complete the classification monitoring of the railway tunnel vibration signal, including the steps of:

[0065] Construct the objective function of SVM classification, which is expressed by the formula:

[0066]

[0067] In the formula, w represents the hyperplane weight, C represents the penalty parameter, and ξ i represents the slack variable;

[0068] According to the classification mechanism of SVM, transform the constructed objective function of SVM classification into a convex optimization problem, and the formula is expressed as:

[0069]

[0070] And obtain the dual expression of the convex optimization function as:

[0071]

[0072] Solve the quadratic optimization problem of the dual expression of the convex optimization function to obtain the optimal solution:

[0073]

[0074] According to the optimal solution result, construct the weight w of the classification hyperplane, and the formula is expressed as:

[0075]

[0076] When the optimal solution component satisfies Then the offset of the SVM classification hyperplane can be calculated:

[0077]

[0078] According to the calculation result of the offset of the SVM classification hyperplane, construct the optimal SVM classification hyperplane:

[0079] w * ·x + b * = 0;

[0080] Under the support of the optimal classification hyperplane, construct the decision function for the classification monitoring of the rotating machinery vibration signal:

[0081] f(x) = sign(w * ·x + b * ).

[0082] In a third aspect, the present invention provides an electronic system, comprising:

[0083] a processor;

[0084] a memory for storing executable instructions of the processor;

[0085] wherein the processor is configured to execute the instructions to implement the HOG-SVM-based railway tunnel vibration signal detection method according to any one of the first aspects.

[0086] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0087] The HOG-SVM-based railway tunnel vibration signal detection method and system provided by the present invention utilize a plurality of foreign object sentry sensing terminals installed on the side walls on both sides of the railway tunnel entrance to obtain real-time foreign object intrusion detection sensing data at the tunnel entrance; utilize a plurality of holographic guard sensing terminals installed at intervals and staggered on both sides of the railway tunnel to obtain real-time intelligent lighting in the tunnel and intelligent risk monitoring sensing data; the edge computing terminal performs real-time analysis, calculation, and detection of the external and internal risks of the tunnel on the sensing data transmitted by the holographic guard sensing terminal and the foreign object sentry sensing terminal, utilizes the vibration signal generated when the train travels on the railway, and performs gain amplification and filtering and noise reduction processing on the vibration signal data; adopts the HOG-SVM algorithm to extract the vibration signal features and classify and detect the vibration signal to analyze the spectrum of the vibration signal to invert the position of the train, and utilize the abnormal vibration signal of a single sensor to judge the detection status of the lamp loosening risk; the present invention can realize intelligent monitoring and analysis of tunnel risks and the status of equipment and facilities according to the monitoring and analysis of railway tunnel vibration signals, reduce the requirements for on-site inspection personnel, shorten the time required for maintenance and repair, and detect tunnel diseases and risks at an early stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] Figure 1 is a flowchart of the HOG-SVM-based railway tunnel vibration signal detection method. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0089] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be used to limit the protection scope of the present invention.

[0090] Embodiment

[0091] As Figure 1 shown, an HOG-SVM-based railway tunnel vibration signal detection method provided in an embodiment of the present invention includes the following steps:

[0092] Utilize a plurality of foreign object sentry sensing terminals installed on the side walls on both sides of the railway tunnel entrance to obtain real-time foreign object intrusion detection sensing data at the tunnel entrance;

[0093] Utilize a number of holographic guard sensing terminals installed alternately at intervals on both sides of the railway tunnel to real-time sense the intelligent lighting in the tunnel and obtain intelligent risk monitoring sensing data;

[0094] The edge computing terminal performs real-time analysis, calculation, and detection of internal and external risks in the tunnel on the sensing data transmitted by the holographic guard sensing terminal and the foreign object sentry sensing terminal. Utilize the vibration signal generated when the train travels on the railway, and perform gain amplification and filter noise reduction processing on the vibration signal data;

[0095] Adopt the HOG-SVM algorithm to extract the vibration signal features and classify and detect the vibration signal, so as to analyze the spectrum of the vibration signal to invert the position of the train, and utilize the abnormal vibration signal of a single sensor to judge the detection status of the risk of lamp loosening.

[0096] In the embodiment of the present invention, the holographic guard sensing front end uses the configured sensor to collect the vibration signal in the railway tunnel and analyze the situation of the train and lamp loosening detection. Utilize the vibration signal generated when the train travels on the railway. After performing gain amplification and filter noise reduction processing on the vibration signal data, adopt the HOG-SVM algorithm to extract the vibration signal features and classify and detect the vibration signal, so as to analyze the spectrum of the vibration signal to invert the position of the train, and utilize the abnormal vibration signal of a single sensor to judge the lamp loosening state.

[0097] In the embodiment of the present invention, the steps of adopting the HOG-SVM algorithm to extract the vibration signal features include:

[0098] Utilize the HOG histogram of oriented gradients to extract the gradient information of the time-frequency spectrum diagram of the vibration signal. The gradient at the edge of the time-frequency spectrum diagram is relatively strong, and perform grayscale and normalization processing on the extracted time-frequency spectrum diagram of the vibration signal. The formula is as follows:

[0099]

[0100] In the formula, I(x, y) represents the grayscale pixel of the time-frequency spectrum diagram, and G(·) represents the grayscale processing function;

[0101] Perform convolution operation processing using the first-order template gradient operator. The formula for the pixel gradient and direction in the spectral image, and the gradient and direction is:

[0102]

[0103] In the formula, θ(x, y) and m(x, y) respectively represent the gradient direction and amplitude of the pixel point;

[0104] Divide the time-frequency spectrum of the railway tunnel vibration into pixel blocks "block" of size M×N, and divide it into square cells "cell" of a×a according to the pixel blocks "block". Combine the calculated gradient direction above to perform weighted projection on the pixels within the cell "cell". After the projection process, regularize the gradient of the time-frequency spectrum in units of pixel blocks "block" to reduce the influence of interference factors on the extraction of HOG features. The formula is:

[0105]

[0106] In the formula, x n represents the pixel block "block" vector, and η represents a constant;

[0107] After the regularization process is completed, connect all the pixel blocks "block" to complete the HOG features of the railway tunnel vibration signal.

[0108] In the embodiment of the present invention, the HOG-SVM algorithm is used. Based on the extracted HOG feature extraction results, a support vector machine SVM is used to construct a vibration signal classification hyperplane to complete the classification monitoring of the railway tunnel vibration signal, including the steps:

[0109] Construct the objective function for SVM classification, and the formula is expressed as:

[0110]

[0111] In the formula, w represents the hyperplane weight, C represents the penalty parameter, and ξ i represents the slack variable;

[0112] According to the classification mechanism of SVM, transform the constructed objective function for SVM classification into a convex optimization problem, and the formula expression is:

[0113]

[0114] And obtain the dual expression of the convex optimization function as:

[0115]

[0116] Solve the quadratic optimization problem of the dual expression of the convex optimization function to obtain the optimal solution:

[0117]

[0118] According to the optimal solution result, construct the weight w of the classification hyperplane, and the formula is expressed as:

[0119]

[0120] When the optimal solution component satisfies Then, the offset of the SVM classification hyperplane can be calculated:

[0121]

[0122] According to the calculation result of the offset of the SVM classification hyperplane, the optimal SVM classification hyperplane is constructed:

[0123] w * ·x + b * = 0;

[0124] Under the support of the optimal classification hyperplane, a decision function for classifying and monitoring the vibration signals of rotating machinery is constructed:

[0125] f(x) = sign(w * ·x + b * ).

[0126] In the embodiment of the present invention, the foreign object sentry sensing terminal monitors the intrusion of foreign objects at the tunnel entrance based on 3D point clouds and intelligent images, including the steps of:

[0127] By obtaining the 3D point cloud data of the monitoring area in real time, a point cloud background of the monitoring area is established. The monitoring area includes the tunnel entrance, the track surface, and the catenary;

[0128] The point cloud newly scanned at the current moment is analyzed and calculated with the background point cloud at the previous moment to determine whether there are foreign objects;

[0129] And the camera of the foreign object sentry sensing terminal is linked for capture and analysis, and the coordinates of the foreign object in the point cloud and the coordinates of the foreign object in the image are jointly calibrated using the point cloud and the image, and the foreign object is marked in the image;

[0130] At the same time, the normal targets of trains and personnel are filtered using the fusion perception technology, and finally, a foreign object alarm picture is uploaded to the edge computing terminal.

[0131] In a second aspect, the embodiment of the present invention provides a railway tunnel vibration signal detection system based on HOG-SVM, including the following modules:

[0132] The foreign object sentry module is used to utilize a plurality of foreign object sentry sensing terminals installed on the side walls on both sides of the railway tunnel entrance to obtain the perception data of foreign object intrusion detection at the tunnel entrance in real time;

[0133] The holographic guard module is used to utilize a plurality of holographic guard sensing terminals installed at intervals and staggered on both sides of the railway tunnel to obtain the intelligent lighting in the tunnel and the perception data of intelligent risk monitoring in real time;

[0134] A data processing module is used for the edge computing terminal to perform real-time analysis, calculation, and detection of internal and external risks in the tunnel on the perception data transmitted by the holographic guard perception terminal and the foreign object sentry perception terminal. Utilize the vibration signal generated when the train travels on the railway, and perform gain amplification and filter noise reduction processing on the vibration signal data.

[0135] A signal judgment module is used to extract vibration signal features and classify and detect vibration signals using the HOG-SVM algorithm, to analyze the spectrum of the vibration signal to invert the position of the train, and to judge the detection status of the risk of lamp loosening using the abnormal vibration signal of a single sensor.

[0136] In an embodiment of the present invention, the signal judgment module is further used to collect vibration signals in the railway tunnel using a configured sensor and analyze the situation of the train and lamp loosening detection. Utilize the vibration signal generated when the train travels on the railway, and after performing gain amplification and filter noise reduction processing on the vibration signal data, extract vibration signal features and classify and detect vibration signals using the HOG-SVM algorithm, to analyze the spectrum of the vibration signal to invert the position of the train, and to judge the lamp loosening state using the abnormal vibration signal of a single sensor.

[0137] In an embodiment of the present invention, the signal judgment module is further used to extract vibration signal features using the HOG-SVM algorithm, including the steps of:

[0138] Utilize the HOG (Histogram of Oriented Gradients) to extract the gradient information of the time-frequency spectrum diagram of the vibration signal. The gradient at the edge of the time-frequency spectrum diagram is relatively strong, and perform grayscale conversion and normalization processing on the extracted time-frequency spectrum diagram of the vibration signal. The formula is as follows:

[0139]

[0140] In the formula, I(x, y) represents the grayscale pixel of the time-frequency spectrum diagram, and G(·) represents the grayscale conversion function;

[0141] Perform convolution operation processing using a first-order template gradient operator. The formula for the pixel gradient and direction in the frequency spectrum image, and the gradient and direction is:

[0142]

[0143] In the formula, θ(x, y) and m(x, y) respectively represent the gradient direction and amplitude of the pixel point;

[0144] Divide the time-frequency spectrum of the railway tunnel vibration into pixel blocks block of size M×N, divide it into square cells cell of size a×a according to the pixel blocks block, and perform weighted projection on the pixels in the cell in combination with the calculated gradient direction above; after the projection processing is completed, regularize the gradient of the time-frequency spectrum in units of pixel blocks block to reduce the influence of interference factors on the extraction of HOG features. The formula is:

[0145]

[0146] In the formula, x n represents the pixel block vector, and η represents a constant;

[0147] After the regularization process is completed, connecting all the pixel blocks block completes the HOG features of the railway tunnel vibration signal.

[0148] In the embodiment of the present invention, the signal judgment module is further configured to use the HOG-SVM algorithm to construct a vibration signal classification hyperplane based on the extracted HOG feature extraction result by using the support vector machine SVM, and complete the classification monitoring of the railway tunnel vibration signal, including the steps:

[0149] Construct the objective function of SVM classification, and the formula is expressed as:

[0150]

[0151] In the formula, w represents the hyperplane weight, C represents the penalty parameter, and ξ i represents the slack variable;

[0152] According to the classification mechanism of SVM, transform the constructed objective function of SVM classification into a convex optimization problem, and the formula is expressed as:

[0153]

[0154] And obtain the dual expression of the convex optimization function as:

[0155]

[0156] Solve the quadratic optimization problem of the dual expression of the convex optimization function to obtain the optimal solution:

[0157]

[0158] According to the optimal solution result, construct the weight w of the classification hyperplane, and the formula is expressed as:

[0159]

[0160] When the optimal solution component satisfies Then, the offset of the SVM classification hyperplane can be calculated:

[0161]

[0162] According to the calculation result of the offset of the SVM classification hyperplane, construct the optimal SVM classification hyperplane:

[0163] w * ·x + b * = 0;

[0164] With the support of the optimal classification hyperplane, construct the decision function for the classification and monitoring of rotating machinery vibration signals:

[0165] f(x) = sign(w * ·x + b * ).

[0166] In a third aspect, an embodiment of the present invention provides an electronic system, including:

[0167] A processor;

[0168] A memory for storing executable instructions of the processor;

[0169] Wherein, the processor is configured to execute the instructions to implement the above-mentioned method for detecting railway tunnel vibration signals based on HOG-SVM.

[0170] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A railway tunnel vibration signal detection method based on HOG-SVM, characterized in that: The steps include: Using several foreign object sentinel sensing terminals installed on the side walls of both sides of the railway tunnel entrance, the foreign object intrusion detection sensing data of the tunnel entrance is obtained in real time; Using several holographic guard sensing terminals installed at intervals on both sides of the railway tunnel, the intelligent lighting in the tunnel and the intelligent risk monitoring sensing data are acquired in real time; The edge computing terminal analyzes, calculates, and detects risks inside and outside the tunnel in real time based on the perception data transmitted by the holographic guard perception terminal and the foreign object sentinel perception terminal. It uses the vibration signal of the train running on the railway to perform gain amplification and filtering noise reduction on the vibration signal data. The HOG-SVM algorithm is used to extract vibration signal features and classify vibration signals for detection, to parse the frequency spectrum of the vibration signal to invert the position of the train, and to use the abnormal vibration signal of a single sensor to determine the risk detection status of lamp looseness.

2. The railway tunnel vibration signal detection method based on HOG-SVM according to claim 1 is characterized in that: The holographic guard perception front end uses the configured sensors to collect vibration signals in the railway tunnel and analyze the looseness of the train and lamps. It uses the vibration signal of the train running on the railway, amplifies the gain and performs filtering and noise reduction processing on the vibration signal data, and then uses the HOG-SVM algorithm to extract the vibration signal characteristics and classify the vibration signal to analyze the frequency spectrum of the vibration signal to invert the position of the train, and uses the abnormal vibration signal of a single sensor to determine the loose state of the lamp.

3. The railway tunnel vibration signal detection method based on HOG-SVM according to claim 2 is characterized in that: The HOG-SVM algorithm is used to extract vibration signal features, including the following steps: The HOG directional gradient histogram is used to extract the gradient information of the vibration signal time spectrum, where the gradient at the edge of the time spectrum is stronger, and the extracted vibration signal time spectrum is grayed and normalized. The formula is as follows: Where I(x, y) represents the grayscale pixel of the time-frequency spectrum, and G(·) represents the grayscale processing function; The first-order template gradient operator is used for convolution operation. The formulas of pixel gradient and direction in the spectrum image are: In the formula, θ(x, y) and m(x, y) represent the gradient direction and amplitude of the pixel respectively; The vibration time-frequency spectrum of the railway tunnel is divided into M×N pixel blocks, which are then divided into a×a square cells according to the pixel blocks. The pixels in the cells are weighted projected in combination with the gradient direction calculated above. After the projection process is completed, the gradient of the time-frequency spectrum is regularized in units of pixel blocks to reduce the influence of interference factors on HOG feature extraction. The formula is: In the formula, x n represents the pixel block vector, η represents a constant; After the regularization process is completed, all pixel blocks are connected to complete the HOG features of the railway tunnel vibration signal.

4. The railway tunnel vibration signal detection method based on HOG-SVM according to claim 3 is characterized in that: The HOG-SVM algorithm is used based on the extracted HOG feature extraction results, and the support vector machine SVM is used to construct the vibration signal classification hyperplane to complete the railway tunnel vibration signal classification monitoring, including the following steps: Construct the objective function of SVM classification, the formula is expressed as: In the formula, w represents the hyperplane weight, C represents the penalty parameter, ξ i represents the slack variable; According to the classification mechanism of SVM, the objective function of the constructed SVM classification is transformed into a convex optimization problem, which is expressed as follows: s.t.y i (w·x i +b)≥1,i=1,2,...,N; And the dual expression of the convex optimization function is obtained as: Solve the quadratic optimization problem of the dual expression of the convex optimization function and obtain the optimal solution: According to the optimal solution, the weight w of the classification hyperplane is constructed, and the formula is expressed as: When the optimal solution satisfy Then the bias of the SVM classification hyperplane can be calculated: According to the calculation results of the offset of the SVM classification hyperplane, the optimal hyperplane for SVM classification is constructed: w * ·x+b * =0; With the support of the optimal classification hyperplane, the decision function for classification monitoring of rotating machinery vibration signals is constructed: f(x)=sign(w * ·x+b * )。 5. The railway tunnel vibration signal detection method based on HOG-SVM according to claim 1 is characterized in that: The foreign body sentinel sensing terminal performs foreign body intrusion monitoring at the tunnel entrance based on three-dimensional point cloud and intelligent image, including the following steps: By acquiring the 3D point cloud data of the monitoring area in real time, the point cloud background of the monitoring area is established. The monitoring area includes the tunnel entrance, track surface, and contact network. Analyze and calculate the newly scanned point cloud at the current moment and the background point cloud at the previous moment to determine whether there is a foreign object; The camera of the foreign object sentinel sensing terminal is linked to capture and analyze, and the coordinates of the foreign object in the point cloud and the image are calibrated together to correspond to the coordinates of the foreign object in the image and mark the foreign object in the image; At the same time, fusion sensing technology is used to filter normal targets of trains and personnel, and finally foreign object warning pictures are uploaded to the edge computing terminal.

6. Railway tunnel vibration signal detection system based on HOG-SVM, characterized in that: Includes the following modules: The foreign body sentinel module is used to obtain the foreign body intrusion detection perception data of the tunnel entrance in real time by using a number of foreign body sentinel sensing terminals installed on the side walls on both sides of the railway tunnel entrance; The holographic guard module is used to use a number of holographic guard sensing terminals installed at intervals on both sides of the railway tunnel to monitor the intelligent lighting in the tunnel and obtain intelligent risk monitoring sensing data in real time; The data processing module is used for the edge computing terminal to analyze, calculate, and detect risks inside and outside the tunnel in real time on the perception data transmitted by the holographic guard perception terminal and the foreign object sentinel perception terminal, and to use the vibration signal of the train running on the railway to perform gain amplification and filtering noise reduction on the vibration signal data; The signal judgment module is used to extract vibration signal features and classify vibration signals using the HOG-SVM algorithm to analyze the frequency spectrum of the vibration signal to invert the position of the train, and to use the abnormal vibration signal of a single sensor to determine the risk detection status of the lamp loosening.

7. The railway tunnel vibration signal detection system based on HOG-SVM according to claim 6 is characterized in that: The signal judgment module is also used to collect vibration signals in railway tunnels using configured sensors and analyze the looseness of trains and lamps. It uses the vibration signals of trains running on railways, performs gain amplification and filtering and noise reduction on the vibration signal data, and then uses the HOG-SVM algorithm to extract vibration signal features and classify and detect vibration signals, so as to analyze the frequency spectrum of the vibration signal to invert the position of the train, and use the abnormal vibration signal of a single sensor to judge the looseness of the lamp.

8. The railway tunnel vibration signal detection system based on HOG-SVM according to claim 7 is characterized in that: The signal judgment module is also used to extract vibration signal features using the HOG-SVM algorithm, including the following steps: The HOG directional gradient histogram is used to extract the gradient information of the vibration signal time spectrum, where the gradient at the edge of the time spectrum is stronger, and the extracted vibration signal time spectrum is grayed and normalized. The formula is as follows: Where I(x, y) represents the grayscale pixel of the time-frequency spectrum, and G(·) represents the grayscale processing function; The first-order template gradient operator is used for convolution operation. The formulas of pixel gradient and direction in the spectrum image are: In the formula, θ(x, y) and m(x, y) represent the gradient direction and amplitude of the pixel respectively; The vibration time-frequency spectrum of the railway tunnel is divided into M×N pixel blocks, which are then divided into a×a square cells according to the pixel blocks. The pixels in the cells are weighted projected in combination with the gradient direction calculated above. After the projection process is completed, the gradient of the time-frequency spectrum is regularized in units of pixel blocks to reduce the influence of interference factors on HOG feature extraction. The formula is: In the formula, x n represents the pixel block vector, η represents a constant; After the regularization process is completed, all pixel blocks are connected to complete the HOG features of the railway tunnel vibration signal.

9. The railway tunnel vibration signal detection system based on HOG-SVM according to claim 8 is characterized in that: The signal judgment module is also used to use the HOG-SVM algorithm based on the extracted HOG feature extraction results, and use the support vector machine SVM to construct a vibration signal classification hyperplane to complete the railway tunnel vibration signal classification monitoring, including the steps of: Construct the objective function of SVM classification, the formula is expressed as: In the formula, w represents the hyperplane weight, C represents the penalty parameter, ξ i represents the slack variable; According to the classification mechanism of SVM, the objective function of the constructed SVM classification is transformed into a convex optimization problem, which is expressed as follows: s.t.y i (w·x i +b)≥1,i=1,2,...,N; And the dual expression of the convex optimization function is obtained as: Solve the quadratic optimization problem of the dual expression of the convex optimization function and obtain the optimal solution: According to the optimal solution, the weight w of the classification hyperplane is constructed, and the formula is expressed as: When the optimal solution satisfy Then the bias of the SVM classification hyperplane can be calculated: According to the calculation results of the offset of the SVM classification hyperplane, the optimal hyperplane for SVM classification is constructed: w * ·x+b * =0; With the support of the optimal classification hyperplane, the decision function for classification monitoring of rotating machinery vibration signals is constructed: f(x)=sign(w * ·x+b * )。 10. An electronic system comprising: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instruction to implement the railway tunnel vibration signal detection method based on HOG-SVM as described in any one of claims 1 to 5.

Images