Ground penetrating radar image matching method and device based on vector symmetry

Through vector symmetry feature extraction and HOVPS algorithm, the problem of poor feature extraction effect in ground penetrating radar image matching is solved, and high accuracy and robust image matching is achieved, which is suitable for a variety of image types.

CN114723975BActive Publication Date: 2024-11-22NAT UNIV OF DEFENSE TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210248363.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-14
Publication Date
2024-11-22
Estimated Expiration
2042-03-14

AI Technical Summary

Technical Problem

The existing ground-penetrating radar image matching method has poor feature extraction effect in the case of low resolution and high noise interference, resulting in low matching accuracy.

Method used

Using the ground penetrating radar image matching method based on vector symmetry, the image symmetry model is constructed through the log-gabor filter and the target image convolution processing, the vector symmetry features are extracted, and the feature extraction and matching is used to achieve sliding Accurate matching of windows.

Benefits of technology

It greatly improves the feature extraction quality and matching accuracy of ground penetrating radar image, and can achieve accurate pixel-level matching in different optical and infrared images, enhancing robustness and noise resistance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114723975B_ABST
    Figure CN114723975B_ABST
Patent Text Reader

Abstract

The present invention relates to a ground penetrating radar image matching method, device, computer equipment and storage medium based on vector symmetry. The method comprises: performing convolution processing on a log-gabor filter and a target image to obtain the energy amplitude of the target image; calculating the noise estimation of the target image according to the distribution characteristics of the energy amplitude of the target image to construct an image symmetry model; performing direction calculation on the image symmetry model to obtain a vector symmetry model; performing feature extraction on the target image according to the vector symmetry model and the HOVPS algorithm to obtain a template window; using the template window to set the size of the sliding window, then performing feature extraction on the target image and the image to be matched, matching the HOVPS features in the sliding window of the target image with the HOVPS features of the image to be registered according to the normalized cross-correlation matching criterion, and obtaining the best matching position of the image to be matched in the target image. The use of this method can improve the accuracy of radar image matching.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a ground penetrating radar image matching method, device, computer equipment and storage medium based on vector symmetry. Background Art

[0002] Ground Penetrating Radar (GPR) is a non-destructive detection and all-weather detection system, which is widely used in road maintenance, underground target detection, GPR positioning and navigation, etc. Among them, GPR image matching is the main technical basis of current GPR positioning and navigation, and is the main hot research direction. The realization of target detection is mainly through classifiers to classify the extracted features, so robust features are the key to the success of target detection.

[0003] At present, the performance of template matching in GPR images mainly depends on the quality of image feature extraction, which can be divided into two categories: one is the algorithm based on the spatial domain features of the image, such as gradient and grayscale features. The other is the algorithm based on the frequency domain features of the image, such as phase information and filter features. In current research, the above methods have their own advantages and limitations: (1) The method based on the spatial domain features of the image is simple to implement and easy to understand, and is easy to extract from GPR images with less information. However, in GPR images with limited resolution and fewer features, the spatial domain features of the image are prone to feature distortion and inaccurate positioning. The current spatial domain features have poor matching effect in these cases. (2) The advantage of the method based on the frequency domain features of the image over the spatial domain features is that it is insensitive to changes in the light intensity of the image and can perceive features with weak amplitudes. However, the physical resolution of GPR is low and it is easily interfered by clutter and noise. Simple frequency domain features will detect unnecessary interference features and have low matching accuracy. Summary of the invention

[0004] Based on this, it is necessary to provide a ground penetrating radar image matching method, device, computer equipment and storage medium based on vector symmetry that can improve the accuracy of ground penetrating radar image matching in order to address the above technical problems.

[0005] A ground penetrating radar image matching method based on vector symmetry, the method comprising:

[0006] Get the ground penetrating radar dataset and the image to be matched;

[0007] Performing filtering on the ground penetrating radar data set to obtain an initial data set; the initial data set includes a plurality of different target images;

[0008] Convolve the log-gabor filter with the target image to obtain the energy amplitude of the target image;

[0009] Calculating a noise estimate of the target image according to a distribution characteristic of the energy amplitude of the target image, and constructing an image symmetry model using the energy amplitude and the noise estimate;

[0010] Direction calculation is performed on the symmetry features in the image symmetry model to obtain a vector symmetry model;

[0011] According to the vector symmetry model and HOVPS algorithm, the target image is feature extracted to obtain a template window;

[0012] The size of the sliding window is set using the template window, and the features of the target image and the image to be matched are extracted according to the sliding window with the set size, so as to obtain the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be matched;

[0013] According to the normalized cross-correlation matching criterion, the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be registered are matched to obtain the best matching position of the image to be matched in the target image.

[0014] In one embodiment, convolving the log-gabor filter with the target image to obtain the energy amplitude of the target image includes:

[0015]

[0016]

[0017]

[0018] Where I(x) represents the target image, represents the even-symmetric component of the log Gabor filter, e n (x) represents the even symmetric component of the image, represents the odd-symmetric component of the log Gabor filter, o n (x) represents the odd symmetric component of the image, A n (x) represents the energy amplitude of the target image at different frequency scales n.

[0019] In one embodiment, calculating a noise estimate of a target image according to a distribution characteristic of an energy amplitude of the target image, and constructing an image symmetry model using the energy amplitude and the noise estimate, comprises:

[0020] According to the energy amplitude response of the target image, the noise estimate of the target image is calculated as T = μ R +k·σ R , where μ R represents the mean of the Rayleigh distribution, σR represents the variance of the Rayleigh distribution, and k represents a custom multiplier factor.

[0021] In one embodiment, the image symmetry model is constructed using energy amplitude and noise estimation:

[0022]

[0023] Among them, η represents a constant to prevent the denominator from being 0, (x, y) represents the position coordinates of the image, and φ n Represents the phase of the image signal at scale n, A n,o Represents the energy amplitude of the image signal at scale n and direction o.

[0024] In one embodiment, the direction of the symmetry features in the image symmetry model is calculated to obtain a vector symmetry model, including:

[0025] Directionally calculate the symmetry features in the image symmetry model to obtain the projections of the phase symmetry in the horizontal and vertical directions;

[0026] The projection of the phase symmetry in the horizontal and vertical directions is calculated by the inverse tangent function to obtain the directional dimension amplitude of the phase symmetry;

[0027] A vector symmetry model is constructed according to the image symmetry amplitude and direction dimension amplitude in the image symmetry model.

[0028] In one embodiment, feature extraction is performed on the target image according to the vector symmetry model and the HOVPS algorithm to obtain a template window, including:

[0029] The target image is divided into multiple sub-image blocks; each sub-image block contains multiple sub-cell units;

[0030] According to the vector symmetry model, the vector symmetry amplitude and direction of each pixel in each sub-cell unit are collected;

[0031] Accumulate the vector symmetry directions of all pixels in the sub-cell units of each sub-block to obtain multiple local histograms;

[0032] The local histogram is corrected, and the corrected local histogram is normalized to obtain a template window.

[0033] In one embodiment, the normalized cross-correlation matching criterion is

[0034]

[0035] Among them, A(x,y) represents the HOVPS feature vector of the image to be matched, represents the average value of the HOVPS feature vector of the image to be matched, B(u,v) represents the HOVPS feature vector in the sliding window of the target image, represents the average value of the HOVPS feature vector in the target image sliding window, W1 represents the sliding window of the image to be matched, and W2 represents the sliding window of the target image.

[0036] A ground penetrating radar image matching device based on vector symmetry, the device comprising:

[0037] The data preprocessing module is used to obtain the ground penetrating radar data set and the image to be matched; filter the ground penetrating radar data set to obtain an initial data set; the initial data set includes multiple different target images;

[0038] Construct an image symmetry model module, which is used to convolve the log-gabor filter with the target image to obtain the energy amplitude of the target image; calculate the noise estimation of the target image according to the distribution characteristics of the energy amplitude of the target image, and construct the image symmetry model using the energy amplitude and noise estimation; calculate the direction of the symmetry features in the image symmetry model to obtain a vector symmetry model;

[0039] The feature extraction module is used to extract features of the target image according to the vector symmetry model and the HOVPS algorithm to obtain a template window; the size of the sliding window is set using the template window, and features of the target image and the image to be matched are extracted according to the sliding window with the set size to obtain the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be matched;

[0040] The matching module is used to match the HOVPS features in the sliding window of the target image with the HOVPS features of the image to be registered according to the normalized cross-correlation matching criterion to obtain the best matching position of the image to be matched in the target image.

[0041] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0042] Get the ground penetrating radar dataset and the image to be matched;

[0043] Performing filtering on the ground penetrating radar data set to obtain an initial data set; the initial data set includes a plurality of different target images;

[0044] Convolve the log-gabor filter with the target image to obtain the energy amplitude of the target image;

[0045] Calculating a noise estimate of the target image according to a distribution characteristic of the energy amplitude of the target image, and constructing an image symmetry model using the energy amplitude and the noise estimate;

[0046] Direction calculation is performed on the symmetry features in the image symmetry model to obtain a vector symmetry model;

[0047] According to the vector symmetry model and HOVPS algorithm, the target image is feature extracted to obtain a template window;

[0048] The size of the sliding window is set using the template window, and the features of the target image and the image to be matched are extracted according to the sliding window with the set size, so as to obtain the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be matched;

[0049] According to the normalized cross-correlation matching criterion, the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be registered are matched to obtain the best matching position of the image to be matched in the target image.

[0050] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0051] Get the ground penetrating radar dataset and the image to be matched;

[0052] Performing filtering on the ground penetrating radar data set to obtain an initial data set; the initial data set includes a plurality of different target images;

[0053] Convolve the log-gabor filter with the target image to obtain the energy amplitude of the target image;

[0054] Calculating a noise estimate of the target image according to a distribution characteristic of the energy amplitude of the target image, and constructing an image symmetry model using the energy amplitude and the noise estimate;

[0055] Direction calculation is performed on the symmetry features in the image symmetry model to obtain a vector symmetry model;

[0056] According to the vector symmetry model and HOVPS algorithm, the target image is feature extracted to obtain a template window;

[0057] The size of the sliding window is set using the template window, and the features of the target image and the image to be matched are extracted according to the sliding window with the set size, so as to obtain the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be matched;

[0058] According to the normalized cross-correlation matching criterion, the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be registered are matched to obtain the best matching position of the image to be matched in the target image.

[0059] The above-mentioned ground penetrating radar image matching method, device, computer equipment and storage medium based on vector symmetry first use the phase symmetry designed by the filter characteristics of the log-gabor filter for ground penetrating radar image processing to construct an image symmetry model, then improve the symmetry scalar feature into a vector feature model for computer vision tasks, and propose a new template feature matching algorithm HOVPS based on vector symmetry to extract image features, which greatly improves the quality of ground penetrating radar image feature extraction. Finally, the template window extracted by the algorithm HOVPS is used to set a sliding window to search for the best matching position in the two images to achieve pixel-level accurate matching. The present invention does not depend on the specific image type, and can also achieve pixel-level registration in optical, remote sensing, infrared and other images. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic diagram of a flow chart of a ground penetrating radar image matching method based on vector symmetry in one embodiment;

[0061] Figure 2 The figure is a comparison diagram of similarity measurements between the present invention and the existing algorithm in one embodiment.

[0062] Figure 3 A comparison diagram of matching results between the present invention and the existing algorithm in one embodiment;

[0063] Figure 4 is a structural block diagram of a ground penetrating radar image matching device based on vector symmetry in one embodiment;

[0064] Figure 5 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0066] In one embodiment, Figure 1 As shown, a ground penetrating radar image matching method based on vector symmetry is provided, comprising the following steps:

[0067] Step 102, obtaining a ground penetrating radar data set and an image to be matched; filtering the ground penetrating radar data set to obtain an initial data set; the initial data set includes a plurality of different target images.

[0068] Multiple two-dimensional slices of underground targets of ground-penetrating radar and several two-dimensional slices of clutter interference are obtained from radar echo data, and these slices are scaled to obtain data sets of the same size. The data set includes target types that are difficult to detect, such as targets of different sizes and uneven reflection intensities. The data set is used as a training test sample for training classifiers. Mean filtering is performed on the acquired ground-penetrating radar data set to filter out the surface reflection wave and filter high-frequency noise to obtain a filtered initial data set. The initial data set includes multiple different target images, and the target image is matched with the image to be matched using the method of the present invention.

[0069] Step 104, convolving the log-gabor filter with the target image to obtain the energy amplitude of the target image; calculating the noise estimation of the target image according to the distribution characteristics of the energy amplitude of the target image, and constructing the image symmetry model using the energy amplitude and the noise estimation; performing direction calculation on the symmetry features in the image symmetry model to obtain a vector symmetry model.

[0070] The present invention improves and expands the high-level visual feature of image symmetry by introducing it, and endows it with vector characteristics to obtain vector symmetry, which is conducive to extracting stable image features. First, the log-gabor wavelet is used as an orthogonal basis for signal frequency decomposition to detect the phase symmetry feature, that is, the log-gabor filter is convolved with the target image at multiple frequency scales and multiple angles to obtain the energy amplitudes at different scales and angles on the template image. The frequency response of the image is also very sensitive to noise. The present invention estimates the noise intensity through the noise statistical characteristics to perform image denoising, and constructs an image symmetry model using energy amplitude and noise estimation. Image symmetry as a scalar feature cannot be applied to visual tasks. The present invention combines the convolution response of the odd function log-gabor filter with the image to calculate the response in the horizontal and vertical directions, obtains the projection of the phase symmetry in the horizontal and vertical directions, and then obtains the direction information of the phase symmetry at each pixel through the inverse tangent function. Finally, the symmetry amplitude and the directional dimension amplitude are combined to obtain the vector symmetry model.

[0071] Step 106, extracting features of the target image according to the vector symmetry model and the HOVPS algorithm to obtain a template window; using the template window to set the size of the sliding window, extracting features of the target image and the image to be matched according to the sliding window with the set size, and obtaining the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be matched.

[0072] The size of the sliding window is set using the template window. The sliding window is selected in the target image. The size of the sliding window is the same as the size of the image to be registered. The corresponding HOVPS features are extracted in both the sliding window and the image to be registered. After extracting the features, the best matching position (x, y) of the image to be matched in the target image can be obtained by using the mutual correlation coefficient as the matching criterion.

[0073] Step 108, matching the HOVPS features in the sliding window of the target image with the HOVPS features of the image to be registered according to the normalized cross-correlation matching criterion, and obtaining the best matching position of the image to be matched in the target image.

[0074] The larger the value of the normalized cross-correlation coefficient NCC is, the higher the similarity between the two images is, that is, the position (x, y) with the largest NCC value is selected as the best matching position.

[0075] In the above-mentioned ground penetrating radar image matching method based on vector symmetry, the phase symmetry designed by the filter feature of the log-gabor filter is first used for ground penetrating radar image processing to construct an image symmetry model, and then the symmetry scalar feature is improved into a vector feature model for computer vision tasks, and a new template feature matching algorithm HOVPS is proposed based on vector symmetry to extract image features, which greatly improves the quality of ground penetrating radar image feature extraction. Finally, the template window extracted by the algorithm HOVPS is used to set a sliding window to search for the best matching position in the two images to achieve pixel-level accurate matching. The present invention does not depend on the specific image type, and can also achieve pixel-level registration in optical, remote sensing, infrared and other images.

[0076] In one embodiment, convolving the log-gabor filter with the target image to obtain the energy amplitude of the target image includes:

[0077]

[0078]

[0079]

[0080] Where I(x) represents the target image, represents the even-symmetric component of the log Gabor filter, e n (x) represents the even symmetric component of the image, represents the odd-symmetric component of the log Gabor filter, o n (x) represents the odd symmetric component of the image, A n (x) represents the energy amplitude of the target image at different frequency scales n.

[0081] First, the log-gabor filter is convolved with the target image at multiple frequency scales and angles to obtain the energy amplitude A of the template image at different scales and angles. n response.

[0082] The log gabor filter expression is as follows:

[0083]

[0084] ω is the log Gabor filter frequency, θ is the log Gabor filter direction, k is a constant used to determine the filter scale, ω0 is the center frequency of the log Gabor filter, σ θ is the filter variance.

[0085] In one embodiment, calculating a noise estimate of a target image according to a distribution characteristic of an energy amplitude of the target image, and constructing an image symmetry model using the energy amplitude and the noise estimate, comprises:

[0086] According to the energy amplitude response of the target image, the noise estimate of the target image is calculated as T = μ R +k·σ R , where μ R represents the mean of the Rayleigh distribution, σ R represents the variance of the Rayleigh distribution, and k represents a custom multiplier factor.

[0087] The frequency response of the image is also very sensitive to noise. The present invention estimates the noise intensity through the noise statistical characteristics to perform image denoising. Assuming that the image contains additive Gaussian noise, the image energy amplitude response obeys the Rayleigh distribution:

[0088]

[0089] Therefore, the noise estimate is defined as the sum of the mean and variance of the Rayleigh distribution multiplied by a coefficient T = μ R +k·σ R .

[0090] In one embodiment, the image symmetry model is constructed using energy amplitude and noise estimation:

[0091]

[0092] Among them, η represents a constant to prevent the denominator from being 0, (x, y) represents the position coordinates of the image, and φ n Represents the phase of the image signal at scale n, A n,o Represents the energy amplitude of the image signal at scale n and direction o.

[0093] In one embodiment, the direction of the symmetry features in the image symmetry model is calculated to obtain a vector symmetry model, including:

[0094] Directionally calculate the symmetry features in the image symmetry model to obtain the projections of the phase symmetry in the horizontal and vertical directions;

[0095] The projection of the phase symmetry in the horizontal and vertical directions is calculated by the inverse tangent function to obtain the directional dimension amplitude of the phase symmetry;

[0096] A vector symmetry model is constructed according to the image symmetry amplitude and direction dimension amplitude in the image symmetry model.

[0097] Image symmetry detection as a scalar feature detection model has a robust feature representation in ground penetrating radar images. By calculating the horizontal projection component of the symmetry feature and the vertical projection component Calculate the directional dimension of the symmetry φ = arctan (y, x), and expand the symmetry feature into a vector feature model VPS = PS·e jφ , which enables symmetry features to be flexibly used in computer vision tasks, filling the gap in designing algorithms for ground penetrating radar image matching based on advanced visual features.

[0098] In one embodiment, feature extraction is performed on the target image according to the vector symmetry model and the HOVPS algorithm to obtain a template window, including:

[0099] The target image is divided into multiple sub-image blocks; each sub-image block contains multiple sub-cell units;

[0100] According to the vector symmetry model, the vector symmetry amplitude and direction of each pixel in each sub-cell unit are collected;

[0101] Accumulate the vector symmetry directions of all pixels in the sub-cell units of each sub-block to obtain multiple local histograms;

[0102] The local histogram is corrected, and the corrected local histogram is normalized to obtain a template window.

[0103] According to the vector symmetry model and HOVPS algorithm, feature extraction of the target image can be summarized into the following four steps:

[0104] S1 Image division: The target image slice is evenly divided into n×n sub-blocks image=[block1,block2,...,block n×n ], the obtained block is then divided into k sub-cell units The vector symmetry magnitude and direction of each pixel are collected in each sub-cell unit;

[0105] S2 generates a symmetry direction histogram: the local histogram of the vector symmetry direction accumulated over all pixels in the cell unit of each sub-block is divided into 8 direction histograms.

[0106] S3 cumulative histogram correction: When the variable range is fixedly divided into several regions for certain calculations, the boundary variables are also correlated with the adjacent regions. If only the current region is calculated and the relationship with the adjacent regions is completely ignored, an aliasing effect will occur, causing the extracted feature vector to mutate. The present invention uses trilinear interpolation to perform interpolation operations on the pixel coordinates (x, y) and the direction coordinate θ in three dimensions to correct the cumulative histogram.

[0107] S4 Histogram Normalization: In order to make the feature invariant to illumination and shadow, the direction histogram of each cell in each block is normalized so that each block generates a feature descriptor s i The feature descriptors of all blocks in a template window are connected in series as the feature descriptor of this template window f = {s1, s2, s3...s n}, n is the number of blocks in the template window.

[0108] The present invention constructs a feature descriptor with high-resolution features and amplitude invariance through vector symmetry, which greatly improves the quality and robustness of feature extraction in ground penetrating radar images and enhances the template matching accuracy of ground penetrating radar images.

[0109] In one embodiment, the normalized cross-correlation matching criterion is

[0110]

[0111] Among them, A(x,y) represents the HOVPS feature vector of the image to be matched, represents the average value of the HOVPS feature vector of the image to be matched, B(u,v) represents the HOVPS feature vector in the sliding window of the target image, represents the average value of the HOVPS feature vector in the target image sliding window, W1 represents the sliding window of the image to be matched, and W2 represents the sliding window of the target image.

[0112] In one embodiment, the root mean square error (RMSE) is used as the evaluation scheme, and the expression of RMSE is as follows:

[0113]

[0114] in, is the reference best matching position coordinate obtained by the present invention, is the true best matching position coordinate.

[0115] The present invention verifies the technical effect of the present invention by comparing it with other algorithms, such as Figure 2 As shown, Figure 2 The similarity peak graphs of ground penetrating radar images collected by different algorithms on different actual roads are shown in Figure 2. The similarity peak graph is an image describing the robustness of similarity drawn by offsetting 20 pixels around the true value matching area of ​​the image to be registered. The more concentrated the peak is in the center of the similarity peak graph, the more stable the algorithm performance is. As can be seen from the figure, the proposed method is most obvious and concentrated in the center of the similarity peak graph, with obvious sidelobe suppression, and can achieve higher matching accuracy.

[0116] Figure 3 The template matching performance of various algorithms on GPR images of different sizes is verified. The experiments are conducted on real data collected on different road sections. As can be seen from the figure, the detection performance of the proposed HOVPS is better than that of the traditional methods, especially when the image slices are small.

[0117] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0118] In one embodiment, Figure 4 As shown, a ground penetrating radar image matching device based on vector symmetry is provided, comprising: a data preprocessing module 402, an image symmetry model building module 404, a feature extraction module 406 and a matching module 408, wherein:

[0119] The data preprocessing module 402 is used to obtain a ground penetrating radar data set and an image to be matched; filter the ground penetrating radar data set to obtain an initial data set; the initial data set includes a plurality of different target images;

[0120] The image symmetry model building module 404 is used to perform convolution processing on the log-gabor filter and the target image to obtain the energy amplitude of the target image; calculate the noise estimation of the target image according to the distribution characteristics of the energy amplitude of the target image, and build the image symmetry model by using the energy amplitude and the noise estimation; calculate the direction of the symmetry feature in the image symmetry model to obtain the vector symmetry model;

[0121] The feature extraction module 406 is used to extract features of the target image according to the vector symmetry model and the HOVPS algorithm to obtain a template window; use the template window to set the size of the sliding window, and extract features of the target image and the image to be matched according to the sliding window with the set size to obtain the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be matched;

[0122] The matching module 408 is used to match the HOVPS features in the sliding window of the target image with the HOVPS features of the image to be registered according to the normalized cross-correlation matching criterion to obtain the best matching position of the image to be matched in the target image.

[0123] In one embodiment, the image symmetry model building module 404 is further used to perform convolution processing on the log-gabor filter and the target image to obtain the energy amplitude of the target image, including:

[0124]

[0125]

[0126]

[0127] Where I(x) represents the target image, represents the even-symmetric component of the log Gabor filter, e n (x) represents the even symmetric component of the image, represents the odd-symmetric component of the log Gabor filter, o n (x) represents the odd symmetric component of the image, A n (x) represents the energy amplitude of the target image at different frequency scales n.

[0128] In one embodiment, the image symmetry model building module 404 is further used to calculate a noise estimate of the target image according to a distribution characteristic of the energy amplitude of the target image, and to build an image symmetry model using the energy amplitude and the noise estimate, including:

[0129] According to the energy amplitude response of the target image, the noise estimate of the target image is calculated as T = μ R +k·σ R, where μ R represents the mean of the Rayleigh distribution, σ R represents the variance of the Rayleigh distribution, and k represents a custom multiplier factor.

[0130] In one embodiment, the image symmetry model is constructed using energy amplitude and noise estimation:

[0131]

[0132] Among them, η represents a constant to prevent the denominator from being 0, (x, y) represents the position coordinates of the image, and φ n Represents the phase of the image signal at scale n, A n,o Represents the energy amplitude of the image signal at scale n and direction o.

[0133] In one embodiment, the image symmetry model building module 404 is further used to calculate the direction of the symmetry features in the image symmetry model to obtain a vector symmetry model, including:

[0134] Directionally calculate the symmetry features in the image symmetry model to obtain the projections of the phase symmetry in the horizontal and vertical directions;

[0135] The projection of the phase symmetry in the horizontal and vertical directions is calculated by the inverse tangent function to obtain the directional dimension amplitude of the phase symmetry;

[0136] A vector symmetry model is constructed according to the image symmetry amplitude and direction dimension amplitude in the image symmetry model.

[0137] In one embodiment, the feature extraction module 406 is further used to extract features from the target image according to the vector symmetry model and the HOVPS algorithm to obtain a template window, including:

[0138] The target image is divided into multiple sub-image blocks; each sub-image block contains multiple sub-cell units;

[0139] According to the vector symmetry model, the vector symmetry amplitude and direction of each pixel in each sub-cell unit are collected;

[0140] Accumulate the vector symmetry directions of all pixels in the sub-cell units of each sub-block to obtain multiple local histograms;

[0141] The local histogram is corrected, and the corrected local histogram is normalized to obtain a template window.

[0142] In one embodiment, the normalized cross-correlation matching criterion is

[0143]

[0144] Among them, A(x,y) represents the HOVPS feature vector of the image to be matched, represents the average value of the HOVPS feature vector of the image to be matched, B(u,v) represents the HOVPS feature vector in the sliding window of the target image, represents the average value of the HOVPS feature vector in the target image sliding window, W1 represents the sliding window of the image to be matched, and W2 represents the sliding window of the target image.

[0145] The specific definition of a ground penetrating radar image matching device based on vector symmetry can be found in the above definition of a ground penetrating radar image matching method based on vector symmetry, which will not be repeated here. Each module in the above-mentioned ground penetrating radar image matching device based on vector symmetry can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0146] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a ground penetrating radar image matching method based on vector symmetry is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0147] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0148] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method in the above embodiment when executing the computer program.

[0149] In one embodiment, a computer storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method in the above embodiment are implemented.

[0150] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0151] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0152] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be based on the attached claims.

Claims

1. A ground penetrating radar image matching method based on vector symmetry, characterized in that: The method comprises: Get the ground penetrating radar dataset and the image to be matched; Performing filtering processing on the ground penetrating radar data set to obtain an initial data set; the initial data set includes a plurality of different target images; Convolution processing is performed on the target image with the log-gabor filter to obtain the energy amplitude of the target image; Calculating a noise estimate of the target image according to a distribution characteristic of the energy amplitude of the target image, and constructing an image symmetry model using the energy amplitude and the noise estimate; Performing direction calculation on the symmetry features in the image symmetry model to obtain a vector symmetry model; Perform feature extraction on the target image according to the vector symmetry model and the HOVPS algorithm to obtain a template window; The template window is used to set the size of the sliding window, and the target image and the image to be matched are subjected to feature extraction according to the sliding window after the size is set, so as to obtain the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be matched; According to the normalized cross-correlation matching criterion, the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be registered are matched to obtain the best matching position of the image to be matched in the target image.

2. The method according to claim 1, characterized in that The log-gabor filter is convolved with the target image to obtain the energy amplitude of the target image, including: Where I(x) represents the target image, represents the even-symmetric component of the log Gabor filter, e n (x) represents the even symmetric component of the image, represents the odd-symmetric component of the log Gabor filter, o n (x) represents the odd symmetric component of the image, A n (x) represents the energy amplitude of the target image at different frequency scales n.

3. The method according to claim 2, characterized in that Calculating a noise estimate of the target image according to a distribution characteristic of the energy amplitude of the target image, and constructing an image symmetry model using the energy amplitude and the noise estimate, including: The noise estimate of the target image is calculated based on the energy amplitude response of the target image following the Rayleigh distribution as T = μ R +k·σ R , where μ R represents the mean of the Rayleigh distribution, σ R represents the variance of the Rayleigh distribution, and k represents a custom multiplier factor.

4. The method according to claim 3, characterized in that The method further comprises: The image symmetry model is constructed using the energy amplitude and the noise estimation: Among them, η represents a constant to prevent the denominator from being 0, (x, y) represents the position coordinates of the image, and φ n Represents the phase of the image signal at scale n, A n,o Represents the energy amplitude of the image signal at scale n and direction o.

5. The method according to claim 1, characterized in that Direction calculation is performed on the symmetry features in the image symmetry model to obtain a vector symmetry model, including: Performing direction calculation on the symmetry features in the image symmetry model to obtain projections of the phase symmetry in the horizontal and vertical directions; The projection of the phase symmetry in the horizontal and vertical directions is calculated by the inverse tangent function to obtain the directional dimension amplitude of the phase symmetry; A vector symmetry model is constructed according to the image symmetry amplitude and the directional dimension amplitude in the image symmetry model.

6. The method according to claim 5, characterized in that The target image is subjected to feature extraction according to the vector symmetry model and the HOVPS algorithm to obtain a template window, including: Performing image division on the target image to obtain a plurality of sub-image blocks; the sub-image blocks contain a plurality of sub-cell units; According to the vector symmetry model, the vector symmetry amplitude and direction of each pixel in each sub-cell unit are collected; Accumulate the vector symmetry directions of all pixels in the sub-cell units of each sub-block to obtain multiple local histograms; The local histogram is corrected, and the corrected local histogram is normalized to obtain a template window.

7. The method according to claim 1, characterized in that The normalized cross-correlation matching criterion is: Among them, A(x,y) represents the HOVPS feature vector of the image to be matched, represents the average value of the HOVPS feature vector of the image to be matched, B(u,v) represents the HOVPS feature vector in the sliding window of the target image, represents the average value of the HOVPS feature vector in the target image sliding window, W1 represents the sliding window of the image to be matched, and W2 represents the sliding window of the target image.

8. A ground penetrating radar image matching device based on vector symmetry, characterized in that: The device comprises: A data preprocessing module is used to obtain a ground penetrating radar data set and an image to be matched; filter the ground penetrating radar data set to obtain an initial data set; the initial data set includes a plurality of different target images; Constructing an image symmetry model module, for convolving the log-gabor filter with the target image to obtain the energy amplitude of the target image; calculating the noise estimate of the target image according to the distribution characteristics of the energy amplitude of the target image, and constructing an image symmetry model using the energy amplitude and the noise estimate; performing direction calculation on the symmetry features in the image symmetry model to obtain a vector symmetry model; A feature extraction module is used to extract features of the target image according to the vector symmetry model and the HOVPS algorithm to obtain a template window; the size of the sliding window is set using the template window, and the target image and the image to be matched are extracted according to the sliding window with the set size to obtain the HOVPS features in the sliding window of the target image and the HOVPS features of the image to be matched; The matching module is used to match the HOVPS features in the sliding window of the target image with the HOVPS features of the image to be registered according to the normalized cross-correlation matching criterion to obtain the best matching position of the image to be matched in the target image.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.