Underground target change detection method, device, equipment and medium based on ground penetrating radar

Abstract

The present invention provides a method, device, equipment and medium for detecting underground target changes based on ground penetrating radar. The method for detecting underground target changes based on ground penetrating radar includes: obtaining GPR Bscan image data collected by ground penetrating radar on a detection area, wherein the detection area includes a target area and a background area; obtaining GPR Bscan image data at different times, wherein the shape of the underground target area at each time is different, and the dielectric constant of the medium in the target area and the background area may be variable or constant; combining the GPR Bscan image data at two different times to obtain a GPR Bscan image data pair; and processing the GPR Bscan image data pair using a change detection network to obtain a change detection result of the detection target. The efficiency and accuracy of identifying underground target changes are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of computers and ground penetrating radar, and in particular to a method, device, equipment and medium for detecting underground target changes based on ground penetrating radar. Background Art

[0002] The study of the evolution of structural defects over time can provide guidance for the maintenance of buildings such as tunnel linings and roads. By comparing data collected at different times and judging the change process of structural defects, civil engineers can analyze the causes of structural aging and structural defects. Ground penetrating radar is a commonly used non-destructive testing equipment that detects internal defects of structures by emitting and receiving electromagnetic waves, thereby inferring the internal conditions of the structure. By collecting data at different times and seasons for the same detection scene, and then comparing and analyzing the obtained GPR scattered echo data, the evolution process of underground structural defects can be grasped. However, when collecting data at different times, due to seasonal changes in weather, the relative dielectric constant of the underground soil medium will change (for example, clay usually has a relative dielectric constant of 6, a relative dielectric constant of 2 in dry conditions, and a relative dielectric constant of 15 in wet conditions). Changes in the water content of underground structural defects will also cause changes in their relative dielectric constant. The GPR echo signal is affected by the relative dielectric constant of the background medium and the underground structural defects. When detecting the same underground structural defect, if the relative dielectric constant of the background medium is large, the target feature in the GPR Bscan image will be biased downward, and the opening of the scattering curve will be small. If the relative dielectric constant of the background medium is small, the target feature will be biased upward, and the opening of the scattering curve will be large. If the relative dielectric constant of the underground structural defect increases, the target feature will change, and the multiple waves will be more obvious. This brings difficulties to the detection of the evolution and development of underground structural defects over time.

[0003] When using GPR to detect underground structural defects, imaging and inversion are commonly used processing methods. Imaging algorithms mainly include back-projection imaging, offset imaging, and diffraction tomography. The quality of GPR inversion depends on the detailed modeling of the GPR electromagnetic scattering path. These processing algorithms rely heavily on the experience of data processing personnel and have high computational complexity.

[0004] Existing technologies such as GPRNet use convolutional neural networks to directly learn the intrinsic relationship between GPR Bscan image data and the propagation speed of electromagnetic waves. Based on the inversion characteristics of GPR for complex data, GPRInvNet proposes GPRInvNet composed of tracetotrace encoder and decoder, which solves the spatial alignment problem between time series Bscan data and spatial dielectric constant maps. For the inversion task under heterogeneous soil conditions, the DMRF-Unet two-stage deep neural network (DNN) is designed. First, the convolutional network is used to remove the clutter caused by the heterogeneity of the background medium, and then the dielectric constant distribution of underground objects of GPR BScan under heterogeneous soil conditions is reconstructed. 3DinvNet realizes the reconstruction of 3D dielectric constant maps from Cscan.

[0005] The above-mentioned prior art addresses the problem of target change detection in underground detection scenarios. It is necessary to perform imaging inversion on the GPR Bscan image data at two moments, and then compare the two imaging inversion results to obtain the target change area. The quality of imaging inversion determines the performance of target change detection. The quality of imaging inversion is affected by factors such as environmental changes, unknown dielectric constants of background media and targets, resulting in reduced accuracy of target change detection. Summary of the invention

[0006] The main purpose of the embodiments of the present invention is to propose a method, device, equipment and medium for detecting underground target changes based on ground penetrating radar, which improves the efficiency and accuracy of identifying underground target changes.

[0007] One aspect of the present invention provides a method for detecting underground target changes based on ground penetrating radar, comprising:

[0008] Acquire GPR Bscan image data collected by ground penetrating radar on a detection area, wherein the detection area includes a target area and a background area;

[0009] Acquire GPR Bscan image data at different times, combine the GPR Bscan image data at two different times, and obtain a GPR Bscan image data pair;

[0010] The GPR Bscan image data is subjected to inversion detection processing using a change detection network to obtain a change detection result of the detection target.

[0011] According to the underground target change detection method based on ground penetrating radar, the GPR Bscan image data collected by the ground penetrating radar on the detection area includes:

[0012] According to the scene conditions and detection requirements of the detection area, determine the one-dimensional measurement line of the detection area and the spacing between each measurement point;

[0013] Acquire the echo signal scattered in the detection area according to the antenna configuration of the ground penetrating radar and the data acquisition parameters of the ground penetrating radar;

[0014] The echo signals of each measuring point are combined to obtain GPR Bscan image data.

[0015] According to the underground target change detection method based on ground penetrating radar, the GPRBscan image data at different times are obtained, including:

[0016] At least one of the shape, dielectric constant and background parameters of the detection target is changed, and GPR Bscan image data at different times are collected while keeping the spatial sampling parameters and the temporal sampling parameters unchanged.

[0017] According to the underground target change detection method based on ground penetrating radar, GPR Bscan image data at two different times are combined to obtain a GPR Bscan image data pair, including:

[0018] The detection area is spatially discretized in the horizontal line dimension and the depth dimension to obtain discrete grids of the target area and the background area;

[0019] Generate a target matrix according to the discrete grid of the target area and the discrete grid of the background area;

[0020] When the contour shape of the underground target changes with time, GPR Bscan image data at two different times are collected to obtain a GPR Bscan image data pair.

[0021] According to the underground target change detection method based on ground penetrating radar, the change detection network is obtained by the following steps:

[0022] Calculate the target matrix of each GPR Bscan image data in the GPR Bscan image data pair, subtract the two obtained target matrices, and obtain the change area label of the GPR Bscan image data pair;

[0023] The GPR Bscan image data pairs and the change area labels are combined into a data set, and the data set is divided into a training set and a test set. The change detection network is trained with the training set to obtain weight parameters, and the performance of the change detection network is tested with the test set.

[0024] Among them, the change detection network includes encoding and decoding. The encoding is used to extract features of the data set through two encoders with different weights to obtain image features of GPR Bscan image data at two different times; the GPR Bscan image features output by the MSMDconvBlock at the corresponding positions of the two encoders are subtracted to obtain differential images of different scales and depths, wherein the differential image is the change feature of the GPR Bscan image; the decoding is used to upsample the image change features through the decoder, use point convolution to reduce the dimension and normalize the upsampled image change features, and output the change detection results.

[0025] According to the underground target change detection method based on ground penetrating radar, the method further includes:

[0026] The encoder and the decoder both further include MSMDconvBlock, wherein the calculation formula of MSMDconvBlock is:

[0027]

[0028] represents convolution operation, Group represents group convolution, Concat represents concatenation, and F s It is the feature extracted by the grouped multi-scale convolution module. is a cyclic grouped multi-hole convolution, F out To output the change detection result, Conv represents a 1×1 convolution, where the input of MSMDconvBlock is a GPR Bscan feature of size H×W×C, H and W are the length and width of the GPR Bscan feature, and C is the number of channels.

[0029] According to the underground target change detection method based on ground penetrating radar, the method further includes:

[0030] An FTM feature block is also connected between the encoder and the decoder. The FTM feature block uses the change feature GPR Bscan feature of the GPRBscan image as input and calculates to obtain the dielectric constant model feature, and fuses the dielectric constant model feature with the output feature of the decoder to obtain the shape and position of the change area of ​​the detection target, wherein the FTM feature block adopts a fully convolutional network.

[0031] Another aspect of an embodiment of the present invention provides an underground target change detection device based on a ground penetrating radar, comprising:

[0032] The first module is used to obtain GPR Bscan image data collected by ground penetrating radar on the detection area, wherein the detection area includes a target area and a background area;

[0033] The second module is used to obtain GPR Bscan image data at different times, and combine the GPR Bscan image data at two different times to obtain a GPR Bscan image data pair;

[0034] The third module is used to process the GPR Bscan image data using a change detection network to obtain a change detection result of the detection target.

[0035] Another aspect of an embodiment of the present invention provides an electronic device, including a processor and a memory;

[0036] The memory is used to store programs;

[0037] The processor executes the program to implement the method described above.

[0038] The embodiment of the present invention also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the method described above.

[0039] The beneficial effects of the present invention are as follows: by detecting changes in any one of the shape, dielectric constant and background parameters of the target, and by collecting and constructing GPR Bscan image data pairs for change detection processing using a change detection network while keeping the spatial sampling parameters and the temporal sampling parameters unchanged, direct detection and output of the shape and position of the target change area from the GPR Bscan data is achieved, without the need for the traditional method of separately imaging inverting the GPR Bscan data at two moments and then comparing the two imaging inversion results to obtain the processing process of the target change area, thereby reducing the system resource consumption of the target change area detection and improving the efficiency and accuracy of underground target change recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0041] Figure 1 It is a schematic flow chart of an underground target change detection method based on ground penetrating radar according to an embodiment of the present invention.

[0042] Figure 2 It is a schematic diagram of the GPR Bscan image data acquisition process according to an embodiment of the present invention.

[0043] Figure 31 is a schematic diagram of a GPR Bscan image data pair generation process according to an embodiment of the present invention.

[0044] Figure 4 1 is a schematic diagram of a change detection network training process according to an embodiment of the present invention.

[0045] Figure 5 It is a schematic diagram of the structure of a network for underground target change detection based on ground penetrating radar according to an embodiment of the present invention.

[0046] Figure 6 It is a schematic diagram of the structure of the MSMDconv module of an embodiment of the present invention.

[0047] Figure 7 It is a schematic diagram of the FTM module structure of an embodiment of the present invention.

[0048] Figure 8 Schematic diagram of a GPR Bscan data pair of a hole data set according to an embodiment of the present invention.

[0049] Fig. 9 Schematic diagram of a dielectric constant model of a void data set according to an embodiment of the present invention.

[0050] Fig.10 A schematic diagram of detection results of changed area labels and hole shape contour changes in a hole data set according to an embodiment of the present invention.

[0051] Fig.11 It is a schematic diagram of a GPR Bscan data pair of a crack data set, a dielectric constant model, a change region label, and a detection result of a crack shape contour change using a network and an algorithm of the present invention in an embodiment of the present invention.

[0052] Fig.12 It is a schematic diagram of an underground target change detection device based on ground penetrating radar according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. In the subsequent description, the use of suffixes such as "module", "component" or "unit" used to represent elements is only for the purpose of facilitating the description of the present invention, and has no specific meaning in itself. Therefore, "module", "component" or "unit" can be used in a mixed manner. "First", "second" and the like are only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features or implicitly indicating the order of the indicated technical features. In this subsequent description, the continuous numbering of the method steps is for the convenience of review and understanding. In combination with the overall technical solution of the present invention and the logical relationship between the various steps, adjusting the implementation order between the steps will not affect the technical effect achieved by the technical solution of the present invention. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0054] refer to Figure 1 , Figure 1 1 is a flow chart of a method for detecting underground target changes based on ground penetrating radar according to an embodiment of the present invention, which includes but is not limited to steps S100 to S300:

[0055] S100, obtaining GPR Bscan image data collected by a ground penetrating radar on a detection area, wherein the detection area includes a target area and a background area.

[0056] In some embodiments, reference Figure 2 The GPR Bscan image data acquisition process diagram shown in FIG. 1 includes but is not limited to steps S110 to S130:

[0057] S110, determining a one-dimensional measurement line of the detection area and a spacing between each measurement point according to the scene conditions and detection requirements of the detection area;

[0058] S120, acquiring an echo signal scattered from the detection area according to the antenna configuration of the ground penetrating radar and the data acquisition parameters of the ground penetrating radar;

[0059] S130, combining the echo signals of each measuring point to obtain GPR Bscan image data.

[0060] In some embodiments, at least one of the shape, dielectric constant and background parameters of the detection target is changed, and GPR Bscan image data at different times are collected while keeping the spatial sampling parameters and the temporal sampling parameters unchanged.

[0061] In some embodiments, the shape of the detection target is changed, or when the shape of the detection target is changed, at least one of the dielectric constant of the detection target and the dielectric constant of the background medium is changed.

[0062] It can be understood that the embodiments of the present invention simulate the changes in the GPR detection scene over time by changing the shape, dielectric constant and dielectric constant of the underground target and the background medium. The above three factors may be changes in the target shape only, or may include changes in the target dielectric constant or the background medium dielectric constant.

[0063] In some embodiments, the change of the GPR detection scene over time is simulated by changing the shape, dielectric constant and dielectric constant of the underground target. The above three factors may be only the change of the target shape, or may include the change of the target dielectric constant or the dielectric constant of the background medium.

[0064] In some embodiments, the detection area is detected by using GPR, data is collected along a one-dimensional survey line, and the spatial and temporal sampling parameters are kept unchanged to obtain GPR Bscan image data of the detection scene at different times. These data collected at different times are combined in pairs to form GPR Bscan data pairs. Each GPR Bscan data pair includes two GPR Bscan images at different times.

[0065] S200, acquiring GPR Bscan image data at different times, and combining GPR Bscan image data at two different times to obtain a GPR Bscan image data pair.

[0066] In some embodiments, reference Figure 3 The schematic diagram of the GPR Bscan image data pair generation process shown includes but is not limited to steps S210 to S230:

[0067] S210, spatially discretizing the detection area in the horizontal measurement line dimension and the depth dimension to obtain a discrete grid of the target area and a discrete grid of the background area;

[0068] S220, generating a target matrix according to the discrete grid of the target area and the discrete grid of the background area;

[0069] S230, when the contour shape of the underground target changes with time, GPR Bscan image data at two different times are collected to obtain a GPR Bscan image data pair.

[0070] In some embodiments, the detection area is spatially discretized in the lateral survey line dimension and the depth dimension, the discrete grid of the target area is set to 1, and the discrete grid of the background area is set to 0 to generate a target matrix; when the contour shape of the underground target changes over time, GPR Bscan data at two different times are collected to form a data pair.

[0071] S300, processing the GPR Bscan image data pair using a change detection network to obtain a change detection result of the detection target.

[0072] In some embodiments, change detection of a detection target includes a training step of a change detection network and a detection step of the change detection network.

[0073] In some embodiments, reference Figure 4 The change detection network training process diagram shown in FIG. 1 includes but is not limited to steps S310 to S320:

[0074] S310, calculating a target matrix of each GPR Bscan image data in the GPR Bscan image data pair, subtracting the obtained two target matrices, and obtaining a change region label of the GPR Bscan image data pair;

[0075] S320, GPR Bscan image data pairs and change area labels are combined into a data set, and the data set is divided into a training set and a test set, a change detection network is trained by the training set to obtain weight parameters, and the performance of the change detection network is tested by the test set.

[0076] The change detection network includes encoding and decoding. The encoding is used to extract features from the data set through two encoders with different weights to obtain the image features of the GPR Bscan image data at two different times. In the two encoders, the GPR Bscan image features output by the corresponding position MSMDconvBlock are subtracted to obtain differential images of different scales and depths. The differential image is the change feature of the GPR Bscan image. The decoding is used to upsample the image change features through the encoder, reduce the dimension and normalize the upsampled image change features by point convolution, and output the change detection results.

[0077] In some embodiments, the change detection network follows an encoder-decoder structure such as Figure 5As shown in the figure, the encoder is composed of a dual-branch non-weight-sharing pseudo-Siamese architecture to extract the change features of the input GPR Bscan data pair. The encoder consists of a ConvBlock and 4 MSMDconvBlocks, where ConvBlock consists of convolution, BN layer and Relu. Each MSMDconvBlock consists of MSMDconv followed by a Maxpooling downsampling module, so the network is downsampled 16 times as a whole. MSMDconvBlock uses MSMDconv to extract multi-scale features and multi-hole convolution features. After each Block of the encoder extracts features, the GPR Bscan change features are obtained by subtracting the outputs of the two encoder branches and connecting them to FTM. FTM transforms the GPR Bscan features into dielectric constant model features to obtain richer feature information of the changing object, and then connects the transformed features to the corresponding positions of the decoder. The decoder also consists of 4 MSMDconvBlocks, each of which uses ConvTranspose for upsampling. Finally, point convolution is used for dimensionality reduction and normalization operations to output the final change detection results. The number of channels for extracting GPRBscan features of the four encoders MSMDconvBlock is 64, 128, 256, and 512 respectively. The number of channels for extracting GPRBscan features of the four decoders MSMDconvBlock is 512, 256, 128, and 64 respectively.

[0078] MSMDconv, as a feature extraction block, is the core module of the encoder and decoder. Through multi-scale and multi-hole convolution, it greatly increases the receptive field of the network, can obtain broader context information, and has low computing resource requirements. Among them, hole convolution is mainly used to process features with larger spatial scales, such as edges, textures, etc.; while multi-scale convolution is suitable for processing situations with multi-scale features, such as different sizes and resolutions of targets, so hole convolution can extract wider feature information and is suitable for capturing global context; while multi-scale convolution can extract feature information at different levels and scales, and can obtain more details and local features.

[0079] like Figure 6 As shown in Figure 2, the input of MSMDconv is feature data of size H×W×C, where H and W are the length and width of the feature and C is the number of channels. MSMDconv can be expressed by formula (3): represents convolution operation, Group represents group convolution, Concat represents concatenation, and F s It is the feature extracted by the grouped multi-scale convolution module. is a cyclic grouped multi-hole convolution, Fout is the final output result, Conv represents 1×1 convolution. Conv reduces the number of feature channels extracted by the grouped multi-scale convolution module to 1 / 2 of the original, and Concat concatenates the features processed by Conv with the features extracted by the cyclic grouped multi-hole convolution.

[0080]

[0081] There is a big difference between GPR Bscan images and natural images. The underground structural defect profile under the dielectric constant model is closer to the natural image, which makes it easier for engineers to extract the location and contour information of underground structural defects. The two are related. According to the propagation characteristics of electromagnetic waves in underground media, the conventional method is to invert the GPR Bscan image to reconstruct the dielectric constant profile of underground anomalies. In the embodiment of the present invention, the mapping between GPRBscan features and dielectric constant model features is realized by constructing an FTM feature block conversion module, such as Figure 7 As shown, FTM consists of a fully convolutional network, and the input is the GPR Bscan features extracted by the encoder. Output is the dielectric constant model feature The first layer consists of a convolution kernel of size 3×3, a dilation rate of 1, a padding size of 1, a BN layer, and a ReLU layer. The number of channels is reduced to the number of channels in the middle layer to reduce the amount of computation. The middle layer consists of 3×3 convolutions with different dilation rates (2 n ,2 n-1 …1,…2 n-1 ,2 n ), where n is the depth level, and in this embodiment n=3. Then a convolution kernel size of 3×3 and a dilation rate of 1 are connected, and a convolution with a padding size of 1 is adjusted to the number of output layer channels. Finally, four parallel 3×3 dilated convolution branches are connected to capture contextual information at different scales. The dilation rates of each branch are 1, 6, 12, and 18, respectively. The different dilation rates determine the size of the receptive field. Here, a wider dilation rate is used than that in the feature extraction stage, because the features at this stage are already relatively clear. Although a larger dilation rate will reduce the resolution of the features, more global information can be observed. The feature extraction stage needs to extract more refined features, but it also needs to ensure a certain size of receptive field to obtain wider contextual information.

[0082] The loss function of the network is a combination of Bce and Dice loss functions. Bce is as shown in formula (4), which can effectively measure and optimize the difference between the change result predicted by the model and the true label. Where n is the number of samples, here is the output size H×W, which is 256×256 in the embodiment of the present invention, y is the label, and y′ is the prediction result, which ranges between 0 and 1.

[0083]

[0084] In scenarios with imbalanced sample categories such as cracks, the number of pixels that change is far less than the number of pixels that do not change, which leads to the problem of imbalanced categories during model training. Dice Loss can balance the impact of different categories by calculating the overlap between the predicted results and the true labels, thereby better handling the problem of imbalanced categories. The expression of the dice coefficient is formula (5), where |X∩Y| is the intersection of X and Y, |X, |Y| is the number of calculated positive samples, multiplied by 2 so that its value range is between 0 and 1. The larger the value, the more similar |X| and |Y| are. In deep learning, using formula (6) L dice The loss function makes the network parameters gradient descent in the direction similar to X and Y.

[0085]

[0086]

[0087] The final loss function Loss in this paper is as shown in formula (7), where λ1 is a hyperparameter and is taken as 0.5 here.

[0088] Loss = λ1L bce +(1-λ1)L dice (7)

[0089] After training the network with the training set, the network weight parameters are obtained, and the change detection network is initialized with the parameters. The GPR Bscan data collected at two different time periods are obtained and input into the initialized change detection network to obtain the shape and position of the changed area. Figure 8 The prediction results output by the change detection network for the hollow GPR Bscan data pair are as follows: Fig.10 The second column shows, Figure 8 The dielectric constant model corresponding to the data pair is: Fig. 9 , Fig.10 The first column is the label of the change area in this scenario. In the crack target scenario, enter Fig.11 (a, b) The crack GPR Bscan data pair, the prediction results output by the change detection network are as follows Fig.11 (f) Fig.11 The dielectric constant model corresponding to the (a, b) data pair is: Fig.11 (c,d), Fig.11 (e) is the label of the changed area in this scenario.

[0090] The embodiment of the present invention is carried out on a workstation equipped with a 13th Gen Intel(R) Core(TM) i9-13900K CPU, 32gb RAM and NVIDIA GeForce RTX 4090 GPU. The proposed method is implemented based on Python and PyTorch framework. Adam is used as the weight optimization algorithm, and the learning rate is set to 0.0001. The steps are as follows:

[0091] Firstly, the finite-difference time-domain (FDTD) algorithm implemented by Gprmax is used to construct a simulation data set to simulate the change of GPR detection scene over time. The background dielectric constant model is set to include clay, dry clay, wet clay, the target is set to air and water, and its shape is set to void, crack, etc. Fig. 9 and 11 (c, d), a dielectric constant model with a length of 2.0m and a depth of 1.2m was established. The transmitter and receiver of the ground penetrating radar are placed at a position of 0.9m. The transmitter and receiver move continuously from left to right along the scanning trajectory to obtain Bscan, and the antenna transmits a Ricker wavelet with a center frequency of 500MHz. The distance between the transmitter and the receiver is 0.02m, the antenna step distance is 0.02m, the sampling time window is 26ns, and the size of the generated GPR Bscan data is (1104,90). The spatial width of the FDTD forward modeling grid is set to 0.005m, and 10 grid points are used in the top, bottom, left and right as ideal matching layer absorption boundary conditions. Each dielectric constant model consists of grid points of size (220,380). Among them Fig. 9 (a) The first column represents the scene at time T1, where the background dielectric constant is 6 and the target relative dielectric constant is 1. Fig. 9 (a) The second column represents the scene at time T2, with a background dielectric constant of 15 and a target relative dielectric constant of 80. Fig. 9 (b) The first column represents the scene at time T1, where the background dielectric constant is 6 and the target relative dielectric constant is 1. Fig. 9 (b) The second column represents the scene at time T2, where the background dielectric constant is 2 and the target relative dielectric constant is 1. Fig. 9 (c) The first column represents the scene at time T1, where the background dielectric constant is 6 and the target relative dielectric constant is 1. Fig. 9 (c) The second column represents the scene at time T2, where the background dielectric constant is 6 and the target relative dielectric constant is 80. Fig. 9 (d) The first column represents the scene at time T1, where the background dielectric constant is 6 and the target relative dielectric constant is 1. Fig. 9(d) The second column represents the scene in period T2, with a background dielectric constant of 6 and a target relative dielectric constant of 1. It represents four types of situations: the target shape, background and target dielectric constants all change due to the influence of environmental weather; the target shape and background dielectric constant change; the target shape and target dielectric constant change; and the target shape only changes. There are 400 pairs in each group, and the shape features can be divided into voids and cracks, so there are 3200 data pairs in total. Fig. 9 is the void dielectric constant model scenario, Fig. 9 In the scenario, the GPR Bscan echo data collected by the ground penetrating radar is used for Figure 8 shown. Fig.11 (c, d) are the crack dielectric constant models. Fig.11 In the (c, d) scenario, the GPR Bscan echo data collected by the ground penetrating radar is Fig.11 As shown in (a,b).

[0092] Fig.12 1 is a diagram of an underground target change detection and analysis device based on ground penetrating radar according to an embodiment of the present invention. The device includes a first module 1210 , a second module 1220 and a third module 1230 .

[0093] Among them, the first module is used to obtain GPR Bscan image data collected by the ground penetrating radar on the detection area, wherein the detection area includes the target area and the background area; the second module is used to obtain GPR Bscan image data at different times, and combine the GPR Bscan image data at two different times to obtain a GPR Bscan image data pair; the third module is used to process the GPR Bscan image data pair using a change detection network to obtain a change detection result of the detection target.

[0094] Exemplarily, with the cooperation of the first module, the second module and the third module in the device, the embodiment device can implement any of the aforementioned underground target change detection methods based on ground penetrating radar, that is, obtaining GPR Bscan image data collected by the ground penetrating radar on the detection area, wherein the detection area includes the target area and the background area; obtaining GPR Bscan image data at different times, combining the GPR Bscan image data at two different times to obtain a GPR Bscan image data pair; processing the GPR Bscan image data pair using a change detection network to obtain a change detection result of the detection target. The beneficial effects of the present invention are as follows: by detecting changes in any one of the shape, dielectric constant and background parameters of the target, and by collecting and constructing GPR Bscan image data pairs for inversion detection and processing using a change detection network while keeping the spatial sampling parameters and the temporal sampling parameters unchanged, direct detection and output of the shape and position of the target change area from GPR Bscan data is achieved, without the need for the traditional method of separately imaging inverting the GPR Bscan data at two moments and then comparing the two imaging inversion results to obtain the processing process of the target change area, thereby reducing the system resource consumption of the target change area detection and improving the efficiency and accuracy of underground target change recognition.

[0095] An embodiment of the present invention further provides an electronic device, the electronic device comprising a processor and a memory;

[0096] The memory stores a program;

[0097] The processor executes the program to perform the aforementioned underground target change detection method based on ground penetrating radar; the electronic device has the function of carrying and running the software system for underground target change detection based on ground penetrating radar provided in an embodiment of the present invention, for example, a personal computer, a minicomputer, a main frame, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or communicating with a charged particle tool or other imaging device, etc.

[0098] An embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a program, and the program is executed by a processor to implement the underground target change detection method based on ground penetrating radar as described above.

[0099] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0100] The embodiment of the present invention also discloses a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the aforementioned underground target change detection method based on ground penetrating radar.

[0101] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features described may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions, and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0102] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0103] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0104] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0105] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0106] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0107] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0108] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the described embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A method for detecting underground target changes based on ground penetrating radar, characterized in that: include: Acquire GPR Bscan image data collected by ground penetrating radar on a detection area, wherein the detection area includes a target area and a background area; Acquire GPR Bscan image data at different times, combine the GPR Bscan image data at two different times, and obtain a GPR Bscan image data pair; Performing inversion detection processing on the GPR Bscan image data pair using a change detection network to obtain a change detection result of the detection target; The obtaining of GPR Bscan image data at different times includes: Changing at least one of the shape, dielectric constant and background parameters of the detection target, and collecting GPR Bscan image data at different times while keeping the spatial sampling parameters and the temporal sampling parameters unchanged; The combining of the GPR Bscan image data at two different times to obtain a GPR Bscan image data pair includes: The detection area is spatially discretized in the horizontal line dimension and the depth dimension to obtain a discrete grid of the target area and a discrete grid of the background area; Generate a target matrix according to the discrete grid of the target area and the discrete grid of the background area; When the contour shape of the underground target changes with time, GPR Bscan image data at two different times are collected to obtain a GPR Bscan image data pair; The change detection network is obtained by the following steps: Calculate the target matrix of each GPR Bscan image data in the GPR Bscan image data pair, subtract the two obtained target matrices, and obtain the change region label of the GPR Bscan image data pair; The GPR Bscan image data pairs and the change area labels are combined into a data set, and the data set is divided into a training set and a test set. The change detection network is trained with the training set to obtain weight parameters, and the performance of the change detection network is tested with the test set. Among them, the change detection network includes encoding and decoding. The encoding is used to extract features of the data set through two encoders with different weights to obtain image features of GPR Bscan image data at two different times; the GPR Bscan image features output by the MSMDconvBlock at the corresponding positions of the two encoders are subtracted to obtain differential images of different scales and depths, wherein the differential image is the change feature of the GPR Bscan image; the decoding is used to upsample the image change features through the decoder, use point convolution to reduce the dimension and normalize the upsampled image change features, and output the change detection results.

2. The underground target change detection method based on ground penetrating radar according to claim 1 is characterized in that: The GPR Bscan image data collected by the ground penetrating radar on the detection area includes: According to the scene conditions and detection requirements of the detection area, determine the one-dimensional measurement line of the detection area and the spacing between each measurement point; Acquire the echo signal scattered in the detection area according to the antenna configuration of the ground penetrating radar and the data acquisition parameters of the ground penetrating radar; The echo signals of each measuring point are combined to obtain GPR Bscan image data.

3. The underground target change detection method based on ground penetrating radar according to claim 1 is characterized in that: The method further comprises: The encoder and the decoder both further include MSMDconvBlock, wherein the calculation formula of MSMDconvBlock is: represents convolution operation, Group represents group convolution, Concat represents concatenation, and F s It is the feature extracted by the grouped multi-scale convolution module. is a cyclic grouped multi-hole convolution, F out To output the change detection result, Conv represents a 1×1 convolution, where the input of MSMDconvBlock is a GPR Bscan feature of size H×W×C, H and W are the length and width of the GPR Bscan feature, and C is the number of channels.

4. The underground target change detection method based on ground penetrating radar according to claim 3 is characterized in that: The method further comprises: An FTM feature block is also connected between the encoder and the decoder. The FTM feature block uses the change feature GPR Bscan feature of the GPR Bscan image as input and calculates to obtain the dielectric constant model feature, and fuses the dielectric constant model feature with the output feature of the decoder to obtain the shape and position of the change area of ​​the detection target, wherein the FTM feature block adopts a fully convolutional network.

5. An underground target change detection device based on ground penetrating radar, characterized in that: include: The first module is used to obtain GPR Bscan image data collected by ground penetrating radar on the detection area, wherein the detection area includes a target area and a background area; The second module is used to obtain GPR Bscan image data at different times, and combine the GPR Bscan image data at two different times to obtain a GPR Bscan image data pair; The third module is used to process the GPR Bscan image data pair using a change detection network to obtain a change detection result of the detection target; The obtaining of GPR Bscan image data at different times includes: Changing at least one of the shape, dielectric constant and background parameters of the detection target, and collecting GPR Bscan image data at different times while keeping the spatial sampling parameters and the temporal sampling parameters unchanged; The combining of the GPR Bscan image data at two different times to obtain a GPR Bscan image data pair includes: The detection area is spatially discretized in the horizontal line dimension and the depth dimension to obtain a discrete grid of the target area and a discrete grid of the background area; Generate a target matrix according to the discrete grid of the target area and the discrete grid of the background area; When the contour shape of the underground target changes with time, GPR Bscan image data at two different times are collected to obtain a GPR Bscan image data pair; The change detection network is obtained by the following steps: Calculate the target matrix of each GPR Bscan image data in the GPR Bscan image data pair, subtract the two obtained target matrices, and obtain the change region label of the GPR Bscan image data pair; The GPR Bscan image data pairs and the change area labels are combined into a data set, and the data set is divided into a training set and a test set. The change detection network is trained with the training set to obtain weight parameters, and the performance of the change detection network is tested with the test set. Among them, the change detection network includes encoding and decoding. The encoding is used to extract features of the data set through two encoders with different weights to obtain image features of GPR Bscan image data at two different times; the GPR Bscan image features output by the MSMDconvBlock at the corresponding positions of the two encoders are subtracted to obtain differential images of different scales and depths, wherein the differential image is the change feature of the GPR Bscan image; the decoding is used to upsample the image change features through the decoder, use point convolution to reduce the dimension and normalize the upsampled image change features, and output the change detection results.

6. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the underground target change detection method based on ground penetrating radar as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that: The storage medium stores a program, and the program is executed by a processor to implement the underground target change detection method based on ground penetrating radar as described in any one of claims 1 to 4.

Images