A low-power 4G seamless line rail visual displacement detection method and device

Abstract

The embodiment of the present invention discloses a low-power 4G seamless line rail visual displacement detection method and device, which obtain image texture features, perform feature screening through feature element spacing, obtain image identification line edge features and corner point features, perform feature fusion on the above-screened texture features, identification line edge features and corner point features, and compare the fused features corresponding to the initial image when no rail displacement occurs to output rail displacement information; image processing based on the above features reduces the amount of feature data that the processor needs to process, and can also fully reflect the position information of the rail identification in the image, has lower energy consumption, realizes low-power rail displacement detection, and the processing speed is also significantly improved; and the present invention proposes an extraction algorithm for the above features, and provides a method for screening texture features, so that the feature extraction accuracy is high, thereby making the rail displacement detection have a higher accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of seamless line rail displacement measurement, and more specifically, to a low-power 4G seamless line rail visual displacement detection method and device. Background Art

[0002] Rail creep on seamless lines is caused by the longitudinal displacement of the rails due to the starting and braking of the train and changes in external temperature. In seamless lines, the difference between the real-time temperature and the locked rail temperature will cause temperature stress inside the rails. This temperature stress increases with the increase in the amplitude of temperature change. Since the rails are constrained by fasteners, their free expansion and contraction in the longitudinal and transverse directions are restricted, and temperature forces will continue to accumulate inside the rails. With the continuous increase in railway operating mileage and service life, if it is not effectively handled for a long time, when the track creep exceeds the allowable set range, it will cause rail breakage, sleeper deflection and other problems, threatening the safe operation of the train. Therefore, accurately detecting the displacement state changes of the rails is an important task in railway operation safety management.

[0003] The existing measurement methods mainly include displacement observation piles, ruler method, laser measurement method, magnetic sensor method, drone measurement method, optical fiber measurement method and traditional image processing method; in the prior art, in order to detect the displacement of seamless rail lines, the traditional image processing method uses a camera to obtain images of track marks and roadside reference marks, and detects track displacement through canny edge detection and Hough transform. This detection is based on the image obtained by the camera for processing and analysis to obtain the detection of rail displacement, and requires an overall analysis of the entire image. The processing process involves a large amount of data processing, which has high requirements on the processor's own processing capabilities. The processing of a large amount of data leads to high power consumption of the equipment, which makes the measurement process have high energy consumption, so that the power of the measuring device cannot meet the needs of long-term outdoor work, and the processor has to process a large amount of data, so the response speed is limited; some researchers have also proposed to extract features from the image and detect track displacement based on the extracted features, but the characteristics of the image containing the track mark are not considered in the feature extraction process, so that although the amount of data is reduced to a certain extent, the accuracy of the displacement determined based on the extracted features will be affected, resulting in poor measurement results. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a low-power 4G seamless line rail visual displacement detection method and device, wherein a track mark is pasted at the displacement measurement point of the seamless line rail, and a rail image containing the track mark is obtained by a monocular camera module fixed at the measurement position, and the collected image is transmitted to a processing module based on a 4G data transmission channel. The processing module obtains the texture features of the image through a texture feature extraction algorithm, and performs feature screening through the feature element spacing to obtain the screened texture features, and obtains the image's identification line edge features and corner point feature extraction through an identification line edge feature extraction algorithm and a corner point feature extraction algorithm, respectively, and performs feature fusion on the above-mentioned screened texture features, identification line edge features and corner point features to obtain the fusion features of the current image, and compares the fusion features of the current image with the fusion features corresponding to the initial image when no rail displacement occurs, thereby outputting the displacement information of the rail; in the process of detecting track displacement, the present invention determines the three features of the image to be extracted according to the characteristics of the image containing the track mark The three features respectively represent the overall features of the image, the marking line features of the rail marking in the image, and the inflection point features of the rail marking in the image. While reducing the amount of data that the processor needs to process, it can also fully reflect the most important information of the image containing the track marking, and can completely represent the position information of the rail marking in the image based on the fusion results of the above three features. The texture features representing the overall features of the image are further screened, so that the amount of feature data that the processor needs to process is further reduced, so that the processor has a smaller energy consumption, and low-power rail displacement detection is achieved. The processing capability requirements of the processor are greatly reduced, and the processing speed of the processor is significantly improved; and in order to accurately extract the texture features, marking line edge features and corner point features of the above image, the present invention proposes an extraction algorithm for the above three features, and provides a method for screening texture features. The feature extraction algorithm proposed based on the present invention can improve the accuracy of feature extraction, so that the displacement detection of the rail has a higher accuracy.

[0005] To achieve the above object, the present invention proposes the following technical solutions:

[0006] In a first aspect, the present invention provides a low-power 4G seamless line rail visual displacement detection method, comprising:

[0007] S1. Collecting an image of a measurement point of a seamless railway rail at a fixed measurement position, wherein a bar track mark of fixed width and length is affixed to the measurement point of the seamless railway rail;

[0008] Preferably, when no displacement of the rail occurs, the line connecting the measuring position and the center point of the track mark is perpendicular to the seamless line rail;

[0009] S2, removing noise and deblurring the image collected in step S1;

[0010] S3, extracting texture features from the image output in step S2;

[0011] S4, extracting the edge features of the track marking line from the image output in step S2;

[0012] S5, extracting corner features of track identification from the image output in step S2;

[0013] S6, screening the texture features outputted in step S3 to obtain screened texture features;

[0014] S7, fusing the marking line edge features outputted from step S4, the corner point features outputted from step S5, and the filtered texture features outputted from step S6 to obtain a fused feature corresponding to the image;

[0015] S8. Compare the fusion features output in step S7 with the fusion features corresponding to the initial image, and output the displacement information of the seamless line rail, wherein the initial image is the image at the measuring point when no rail displacement occurs, and the corresponding fusion features are the fusion features output after the initial image is processed by steps S2-S7.

[0016] In a second aspect, the present invention provides a low-power 4G seamless line rail visual displacement detection device, comprising:

[0017] An image acquisition module, used for acquiring images of measurement points of seamless railway rails at fixed measurement positions, wherein the measurement points of seamless railway rails are affixed with bar-shaped track markings of fixed width and length;

[0018] Preferably, when no displacement of the rail occurs, the line connecting the measuring position and the center point of the track mark is perpendicular to the seamless line rail;

[0019] Preferably, the image acquisition device is a monocular camera;

[0020] An image processing module removes noise and deblurs the image collected by the image acquisition device;

[0021] A texture feature extraction module is used to extract texture features from the image after noise removal and deblurring;

[0022] The marking line edge feature extraction module is used to extract the marking line edge features of the track marking from the image after noise removal and deblurring;

[0023] A corner feature extraction module is used to extract corner feature of track identification from the image after noise removal and deblurring; a feature screening module is used to screen the texture feature extracted by the texture feature extraction module to obtain the screened texture feature;

[0024] A feature fusion module, used for fusing the marking line edge feature, the corner point feature and the screened texture feature to obtain a fusion feature corresponding to the image;

[0025] The feature comparison module is used to compare the fusion features output by the feature fusion module with the fusion features corresponding to the initial image, and output the displacement information of the seamless line rail, wherein the initial image is the image at the measuring point when no rail displacement occurs, and the corresponding fusion features are the fusion features output after the initial image is processed by the above-mentioned image processing module, texture feature extraction module, marking line edge feature extraction module, corner point feature extraction module, feature screening module and feature fusion module.

[0026] In a third aspect, the present invention provides an electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is executed by the processor to implement a low-power 4G seamless line rail visual displacement detection method provided in the first aspect of the present invention;

[0027] In a fourth aspect, the present invention provides a computer-readable storage medium, on which is stored a computer program executable by a processor, and the computer program is executed by the processor to implement a low-power 4G seamless line rail visual displacement detection method provided in the first aspect of the present invention.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The present invention proposes a low-power 4G seamless line rail visual displacement detection method and system, which obtains a rail image containing track markings through a monocular camera module fixed at a measuring position, obtains texture features of the image through a texture feature extraction algorithm, and performs feature screening through feature element spacing to obtain screened texture features, and obtains mark line edge features and corner point features of the image through a mark line edge feature extraction algorithm and a corner point feature extraction algorithm, respectively, and performs feature fusion on the above-mentioned screened texture features, mark line edge features and corner point features to obtain the fusion features of the current image, and compares the fusion features of the current image with the fusion features corresponding to the initial image when no rail displacement occurs, thereby outputting the displacement information of the rail; the image processing based on the above three features reduces the processing The amount of data that the processor needs to process can also fully reflect the position information of the rail mark in the image containing the track mark, and the texture features representing the overall features of the image are further screened, so that the amount of feature data that the processor needs to process is further reduced, so that the processor has a smaller energy consumption, and low-power rail displacement detection is achieved. The processing capability requirements of the processor are greatly reduced, and the processing speed of the processor is significantly improved. In order to accurately extract the texture features, marking line edge features and corner point features of the above-mentioned image, the present invention proposes an extraction algorithm for the above three features, and provides a method for screening texture features. The feature extraction algorithm proposed based on the present invention can improve the accuracy of feature extraction, so that the displacement detection of the rail has a higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The above and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0031] Figure 1 A flow chart of a low-power 4G seamless line rail visual displacement detection method according to an embodiment of the present invention; and

[0032] Figure 2 It is a structural diagram of a low-power 4G seamless line rail visual displacement detection device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The present application is described below based on embodiments, but the present application is not limited to these embodiments. In the detailed description of the present application below, some specific details are described in detail. It is possible for those skilled in the art to fully understand the present application without the description of these details. In order to avoid confusing the essence of the present application, known methods, processes, flows, components and circuits are not described in detail.

[0034] In addition, persons of ordinary skill in the art will appreciate that the drawings provided herein are for illustration purposes and are not necessarily drawn to scale.

[0035] Unless the context clearly requires otherwise, the words "include", "including" and similar words throughout the application should be interpreted as including rather than exclusive or exhaustive; that is, the meaning is "including but not limited to".

[0036] In the description of this application, it should be understood that the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" is two or more.

[0037] Embodiment 1:

[0038] The present invention provides a low-power 4G seamless line rail visual displacement detection method, the flow chart of which is as follows: Figure 1 As shown, including:

[0039] S1. Collecting an image of a measurement point of a seamless railway rail at a fixed measurement position, wherein a bar track mark of fixed width and length is affixed to the measurement point of the seamless railway rail;

[0040] Preferably, when no displacement of the rail occurs, the line connecting the measuring position and the center point of the track mark is perpendicular to the seamless line rail;

[0041] In the displacement detection process of seamless line rails, the most critical thing is to determine the difference between the coordinates of the strip track marker pasted at the measurement point of the seamless line rail in the original image and the coordinates in the image where the displacement occurs. Therefore, the present invention integrates the edge features of the marking line and the corner point features to characterize the coordinate information of the strip track marker in the image, and combines the overall texture features of the image to accurately determine the edge line and corner point coordinate information of the strip track marker in the collected image. The displacement information of the seamless line rail can be determined by comparing with the edge line and corner point coordinate information of the strip track marker in the original image.

[0042] S2, removing noise and deblurring the image collected in step S1;

[0043] Generally, the noise can be filtered out by Gaussian filtering the image, and the image deblurring can adopt the inverse filtering image deblurring method in the prior art, the deblurring method based on the adversarial network, etc. The specific image deblurring method is not specifically limited in the present invention;

[0044] S3, extracting texture features from the image output in step S2;

[0045] Step S3 specifically includes the following steps:

[0046] S31: dividing the image outputted in step S2 into N sub-images of the same size from top to bottom according to the coordinate sequence, so that the N sub-images can be spliced ​​in the coordinate sequence to completely restore the image outputted in step S2;

[0047] S32: extracting texture features from each sub-image to obtain N d-dimensional feature vectors, where d is the dimension of the feature vector of the sub-image;

[0048] Pair of images Perform texture feature extraction, including:

[0049] S321: The characteristic coefficient under scale R and direction D is expressed as: ,in is the image grayscale value of pixel i, is the coordinate of pixel i, is the coordinate wavelet transform for pixel i;

[0050] S322: Calculation Get the statistics of The d statistical information of the sub-image constitutes N d-dimensional feature vectors , as the texture feature vector of sub-image n;

[0051] S33: N texture feature vectors of the N sub-images together constitute the texture features of the image;

[0052] In the present invention, texture features are N d-dimensional feature vectors calculated from an image, that is, N*d related statistical information calculated from an image, which is a quantification of the overall image features.

[0053] S4, extracting the edge features of the track marking line from the image output in step S2;

[0054] Step S4 specifically includes the following steps:

[0055] S41: Calculate the grayscale gradient of pixel i: ,in is the image grayscale value of pixel i, is the image grayscale value of the neighboring pixels of pixel i, where , where M is the number of neighboring pixels of pixel i;

[0056] S42: Determine the grayscale gradient of pixel i The size relationship between the grayscale gradient and the set gradient threshold is that the coordinates of the pixel i whose grayscale gradient is greater than the set gradient threshold are The corresponding gray value As a landmark edge feature;

[0057] In the present invention, the edge feature of the track mark can be understood as a set of pixels whose image characteristics have a step change due to the edge of the track mark.

[0058] S5, extracting corner features of track identification from the image output in step S2;

[0059] Step S5 specifically includes the following steps:

[0060] S51: Definition To move pixel i to pixel k, The grayscale change value, pixel k is different from pixel i; the coordinates of pixel i , the coordinates of pixel k , ,in Pixel The gray value of the image, is the image grayscale value of pixel k, is the Gaussian filter function, ,in for The variance of

[0061] S52: Definition The corresponding grayscale intermediate value and ,in is the image grayscale value of pixel i, ,in represents the Kronecker product;

[0062] S53: Calculate the correlation matrix of pixel i The eigenvalues ​​of

[0063]

[0064] ,in is the image grayscale value of pixel i, and for The corresponding grayscale intermediate amount;

[0065] S54: Compare the eigenvalue of the correlation matrix of pixel i with the set corner point threshold, and calculate the coordinates of the pixel i whose eigenvalue is greater than the set corner point threshold The corresponding gray value As corner features;

[0066] In the present invention, the corner point feature can be understood as a set of inflection point pixels in the image due to the existence of track markers in the image.

[0067] S6, screening the texture features outputted in step S3 to obtain screened texture features;

[0068] Step S6 specifically includes the following steps:

[0069] Calculate the feature element distance between the texture feature vectors of two adjacent sub-images, that is, ,in is the feature weight corresponding to the tth feature in the feature vector, ; Determine the distance of characteristic elements The relationship between the size of the set distance threshold and the distance between the feature elements When the texture feature vector corresponding to the nth sub-image is smaller than the set threshold, the texture feature vector corresponding to the nth sub-image is deleted from the image texture feature output in step S33;

[0070] S7, fusing the marking line edge features outputted from step S4, the corner point features outputted from step S5, and the filtered texture features outputted from step S6 to obtain a fused feature corresponding to the image;

[0071] S8. Compare the fusion features output in step S7 with the fusion features corresponding to the initial image, and output the displacement information of the seamless line rail, wherein the initial image is the image at the measuring point when no rail displacement occurs, and the corresponding fusion features are the fusion features output after the initial image is processed by steps S2-S7.

[0072] Embodiment 2:

[0073] The present invention proposes a low-power 4G seamless line rail visual displacement detection device 200, such as Figure 2 As shown, including:

[0074] An image acquisition module, used for acquiring images of measurement points of seamless railway rails at fixed measurement positions, wherein the measurement points of seamless railway rails are affixed with bar-shaped track markings of fixed width and length;

[0075] Preferably, when no displacement of the rail occurs, the line connecting the measuring position and the center point of the track mark is perpendicular to the seamless line rail;

[0076] Preferably, the image acquisition device is a monocular camera;

[0077] An image processing module removes noise and deblurs the image collected by the image acquisition device;

[0078] Generally, the noise can be filtered out by Gaussian filtering the image, and the image deblurring can adopt the inverse filtering image deblurring method in the prior art, the deblurring method based on the adversarial network, etc. The specific image deblurring method is not specifically limited in the present invention;

[0079] In the displacement detection process of seamless line rails, the most critical thing is to determine the difference between the coordinates of the strip track marker pasted at the measurement point of the seamless line rail in the original image and the coordinates in the image where the displacement occurs. Therefore, the present invention integrates the edge features of the marking line and the corner point features to characterize the coordinate information of the strip track marker in the image, and combines the overall texture features of the image to accurately determine the edge line and corner point coordinate information of the strip track marker in the collected image. The displacement information of the seamless line rail can be determined by comparing with the edge line and corner point coordinate information of the strip track marker in the original image.

[0080] A texture feature extraction module is used to extract texture features from the image after noise removal and deblurring;

[0081] The texture feature extraction module specifically performs the following steps:

[0082] S31: dividing the image after noise removal and deblurring from top to bottom into N sub-images of the same size according to the coordinate order, so that the N sub-images can be spliced ​​in the coordinate order to completely restore the image after noise removal and deblurring;

[0083] S32: extracting texture features from each sub-image to obtain N d-dimensional feature vectors, where d is the dimension of the feature vector of the sub-image;

[0084] Pair of images Perform texture feature extraction, including:

[0085] S321: The characteristic coefficient under scale R and direction D is expressed as: ,in is the image grayscale value of pixel i, is the coordinate of pixel i, is the coordinate wavelet transform for pixel i;

[0086] S322: Calculation Get the statistics of The d statistical information of the sub-image constitutes N d-dimensional feature vectors , as the texture feature vector of sub-image n;

[0087] S33: N texture feature vectors of the N sub-images together constitute the texture features of the image;

[0088] The marking line edge feature extraction module is used to extract the marking line edge features of the track marking from the image after noise removal and deblurring;

[0089] The marker line edge feature extraction module specifically performs the following steps:

[0090] S41: Calculate the grayscale gradient of pixel i: ,in is the image grayscale value of pixel i, is the image grayscale value of the neighboring pixels of pixel i, where , where M is the number of neighboring pixels of pixel i;

[0091] S42: Determine the grayscale gradient of pixel i The size relationship between the grayscale gradient and the set gradient threshold is that the coordinates of the pixel i whose grayscale gradient is greater than the set gradient threshold are The corresponding gray value As a landmark edge feature;

[0092] Corner feature extraction module, used to extract corner features of track identification from the image after noise removal and deblurring;

[0093] The corner feature extraction module specifically performs the following steps:

[0094] S51: Definition To move pixel i to pixel k, The grayscale change value, pixel k is different from pixel i; the coordinates of pixel i , the coordinates of pixel k , ,in Pixel The gray value of the image, is the image grayscale value of pixel k, is the Gaussian filter function, ,in for The variance of

[0095] S52: Definition The corresponding grayscale intermediate value and ,in is the image grayscale value of pixel i, ,in represents the Kronecker product;

[0096] S53: Calculate the correlation matrix of pixel i The eigenvalues ​​of

[0097] , ,in is the image grayscale value of pixel i, and for The corresponding grayscale intermediate amount;

[0098] S54: Compare the eigenvalue of the correlation matrix of pixel i with the set corner point threshold, and calculate the coordinates of the pixel i whose eigenvalue is greater than the set corner point threshold The corresponding gray value As corner features;

[0099] A feature screening module, used for screening the texture features extracted by the texture feature extraction module to obtain screened texture features;

[0100] The feature screening module specifically performs the following steps:

[0101] Calculate the feature element distance between the texture feature vectors of two adjacent sub-images, that is, ,in is the feature weight corresponding to the tth feature in the feature vector, ; Determine the distance of characteristic elements The relationship between the size of the set distance threshold and the distance between the feature elements When the texture feature vector corresponding to the nth sub-image is smaller than the set threshold, the texture feature vector corresponding to the nth sub-image is deleted from the image texture feature output in step S33;

[0102] A feature fusion module, used for fusing the marking line edge feature, the corner point feature and the screened texture feature to obtain a fusion feature corresponding to the image;

[0103] The feature comparison module is used to compare the fusion features output by the feature fusion module with the fusion features corresponding to the initial image, and output the displacement information of the seamless line rail, wherein the initial image is the image at the measuring point when no rail displacement occurs, and the corresponding fusion features are the fusion features output after the initial image is processed by the above-mentioned image processing module, texture feature extraction module, marking line edge feature extraction module, corner point feature extraction module, feature screening module and feature fusion module.

[0104] Preferably, the above-mentioned image processing module, texture feature extraction module, marker line edge feature extraction module, corner feature extraction module, feature screening module and feature fusion module are all arranged in a processor, and the monocular camera and the processor transmit the collected image through a 4G channel;

[0105] Preferably, a near-infrared light source is installed below the front lens of the monocular camera, and the light source switch is triggered by a light source sensing resistor. For a dim light environment, the near-infrared light source built into the monocular camera is used for fill light;

[0106] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The solutions in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and interpreted scripting language JavaScript, etc.

[0107] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0108] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0109] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0110] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A low-power 4G seamless line rail visual displacement detection method, characterized in that: The steps include: S1. Collecting an image of a measurement point of a seamless railway rail at a fixed measurement position, wherein a bar track mark of fixed width and length is affixed to the measurement point of the seamless railway rail; S2, removing noise and deblurring the image collected in step S1; S3, extracting texture features from the image output in step S2; S4, extracting the edge features of the track marking line from the image output in step S2; S5, extracting corner features of track identification from the image output in step S2; S6, screening the texture features outputted in step S3 to obtain screened texture features; S7, fusing the marking line edge features outputted from step S4, the corner point features outputted from step S5, and the filtered texture features outputted from step S6 to obtain a fused feature corresponding to the image; S8, comparing the fusion feature outputted in step S7 with the fusion feature corresponding to the initial image, and outputting the displacement information of the seamless line rail; The initial image is the image at the measuring point when no rail displacement occurs, and the corresponding fusion feature is the fusion feature output after the initial image is processed in steps S2-S7; The step S3 specifically includes the following steps: S31: dividing the image outputted in step S2 into N sub-images of the same size from top to bottom according to the coordinate sequence, so that the N sub-images can be spliced ​​in the coordinate sequence to completely restore the image outputted in step S2; S32: extracting texture features from each sub-image to obtain N d-dimensional feature vectors, where d is the dimension of the feature vector of the sub-image; Pair of images Perform texture feature extraction, including: S321: The characteristic coefficient under scale R and direction D is expressed as: ,in Pixel The gray value of the image, Pixel The coordinates of For pixels Coordinate wavelet transform of ; S322: Calculation Get the statistics of The d statistical information of the sub-image constitutes N d-dimensional feature vectors , as a sub-image Texture feature vector of S33: The N texture feature vectors of the N sub-images together constitute the texture features of the image.

2. The low-power 4G seamless line rail visual displacement detection method according to claim 1 is characterized in that: The step S4 specifically includes the following steps: S41: Counting pixels Grayscale gradient: ,in Pixel The gray value of the image, Pixel The image grayscale values ​​of the adjacent pixels are ,in Pixel The number of neighboring pixels; S42: Determine pixel Grayscale gradient The size relationship between the grayscale gradient and the set gradient threshold is that the pixels with grayscale gradient greater than the set gradient threshold are Coordinates The corresponding gray value As a landmark edge feature.

3. The low-power 4G seamless line rail visual displacement detection method according to claim 1 is characterized in that: The step S5 specifically includes the following steps: S51: Definition Pixel Move to pixel Grayscale change value, pixel point Different from pixels pixel Coordinates , pixels Coordinates , ,in Pixel The gray value of the image, Pixel The gray value of the image, is the Gaussian filter function, ,in for The variance of S52: Definition The corresponding grayscale intermediate value and ,in Pixel The gray value of the image, , ,in represents the Kronecker product; S53: Counting pixels The correlation matrix Q = The eigenvalues ​​of ， ， ,in Pixel The gray value of the image, and for The corresponding grayscale intermediate amount; S54: Pixel The eigenvalue of the correlation matrix is ​​compared with the set corner threshold, and the pixels with eigenvalues ​​greater than the set corner threshold are selected. Coordinates The corresponding gray value as corner features.

4. The low-power 4G seamless line rail visual displacement detection method according to claim 1 is characterized in that: Step S6 specifically includes the following steps: Calculate the feature element distance between the texture feature vectors of two adjacent sub-images, that is, ,in is the feature weight corresponding to the tth feature in the feature vector, ; Determine the distance between characteristic elements The relationship between the size of the set distance threshold and the distance between the feature elements When the image texture feature output in step S33 is smaller than the set threshold, the first The texture feature vector corresponding to the sub-image.

5. A low-power 4G seamless line rail visual displacement detection device, operating as any one of claims 1-4 low-power 4G seamless line rail visual displacement detection methods, characterized in that: The device comprises: An image acquisition module, used for acquiring images of measurement points of seamless railway rails at fixed measurement positions, wherein the measurement points of seamless railway rails are affixed with bar-shaped track markings of fixed width and length; An image processing module removes noise and deblurs the image collected by the image acquisition device; A texture feature extraction module is used to extract texture features from the image after noise removal and deblurring; The marking line edge feature extraction module is used to extract the marking line edge features of the track marking from the image after noise removal and deblurring; A corner feature extraction module is used to extract corner feature of track identification from the image after noise removal and deblurring; a feature screening module is used to screen the texture feature extracted by the texture feature extraction module to obtain the screened texture feature; A feature fusion module, used for fusing the marking line edge feature, the corner point feature and the screened texture feature to obtain a fusion feature corresponding to the image; A feature comparison module is used to compare the fusion features output by the feature fusion module with the fusion features corresponding to the initial image, and output displacement information of the seamless line rail, wherein the initial image is the image at the measuring point when no rail displacement occurs; The corresponding fusion feature is the fusion feature output after the initial image is processed by the above-mentioned image processing module, texture feature extraction module, marker line edge feature extraction module, corner point feature extraction module, feature screening module and feature fusion module.

6. The low-power 4G seamless line rail visual displacement detection device according to claim 5 is characterized in that: The marker line edge feature extraction module specifically performs the following steps: S41: Counting pixels Grayscale gradient: ,in Pixel The gray value of the image, Pixel The image grayscale values ​​of the adjacent pixels are ,in Pixel The number of neighboring pixels; S42: Determine pixel Grayscale gradient The size relationship between the grayscale gradient and the set gradient threshold is that the pixels with grayscale gradient greater than the set gradient threshold are Coordinates The corresponding gray value As a landmark edge feature.

7. The low-power 4G seamless line rail visual displacement detection device according to claim 5 is characterized in that: The corner feature extraction module specifically performs the following steps: S51: Definition Pixel Move to pixel Grayscale change value, pixel point Different from pixels pixel Coordinates , pixels Coordinates , ,in Pixel The gray value of the image, Pixel The gray value of the image, is the Gaussian filter function, ,in for The variance of S52: Definition The corresponding grayscale intermediate value and ,in Pixel The gray value of the image, , ,in represents the Kronecker product; S53: Counting pixels The correlation matrix Q = The eigenvalues ​​of ， ， ,in Pixel The gray value of the image, and for The corresponding grayscale intermediate amount; S54: Pixel The eigenvalue of the correlation matrix is ​​compared with the set corner threshold, and the pixels with eigenvalues ​​greater than the set corner threshold are selected. Coordinates The corresponding gray value as corner features.

8. The low-power 4G seamless line rail visual displacement detection device according to claim 5 is characterized in that: The feature screening module specifically performs the following steps: Calculate the feature element distance between the texture feature vectors of two adjacent sub-images, that is, ,in is the feature weight corresponding to the tth feature in the feature vector, ; Determine the distance between characteristic elements The relationship between the size of the set distance threshold and the distance between the feature elements When the image texture feature output in step S33 is smaller than the set threshold, the first The texture feature vector corresponding to the sub-image.

Images