Phase gradient-based ground deformation detection method, device, equipment and medium

Abstract

The present application discloses a surface deformation detection method, device, equipment and medium based on phase gradient, the method comprising: obtaining a time series interference pattern by synthetic aperture radar interferometry; calculating a deformation phase gradient according to the time series interference pattern; superimposing the calculated deformation phase gradient, and performing an unwrapping operation on the superimposed deformation phase gradient to obtain an unwrapped absolute phase gradient; inputting the unwrapped absolute phase gradient into a pre-trained local surface deformation recognition model to identify the deformation area. The surface deformation detection method based on phase gradient provided in the embodiment of the present application not only improves the accuracy and reliability of surface deformation detection, but also reduces the difficulty of identifying local deformation signals from the rate map obtained from the time series analysis, greatly improves the detection efficiency of local deformation signals in a large range, and at the same time the computational cost is also greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of ground deformation detection, and in particular to a ground deformation detection method, device, equipment and medium based on phase gradient. Background Art

[0002] The surface of the earth is affected by natural factors and human activities, which can produce different degrees of deformation, causing damage to surface buildings, threatening people's lives and safety, and causing economic losses. Rapidly and accurately detecting the location of surface deformation is crucial for the prevention and control of ground deformation disasters and the health and safety assessment of surface structures. With the accumulation of synthetic aperture radar data, it provides important observation support for the detection of surface deformation. However, how to efficiently and quickly process massive SAR image data and obtain local surface deformation in a wide area is still a challenge.

[0003] Synthetic Aperture Radar (InSAR) is a space geodetic and remote sensing technology developed in recent decades. It can obtain large-scale surface deformation monitoring results. It has the advantages of large monitoring range, high measurement accuracy, and all-day and all-weather. It has been widely used in urban subsidence, glacier movement, mining subsidence, volcanic eruption, landslide, mudslide and other surface deformation monitoring. After years of development, InSAR technology has gradually evolved from traditional differential InSAR technology to time series InSAR technology. However, when SAR images are affected by various decoherence factors, whether it is differential InSAR or multi-time series InSAR technology, in the case of long time and space baselines or vegetation coverage areas, low interference phase quality and poor interference effect will inevitably occur, which will lead to the problem of a small number of time series monitoring points, reduced monitoring accuracy, and difficulty in extracting surface deformation information in a wide area. It is impossible to fully reveal the law of surface deformation, which limits the application of InSAR technology in surface deformation monitoring. Summary of the invention

[0004] The embodiments of the present application provide a method, device, equipment and medium for detecting surface deformation based on phase gradient. In order to have a basic understanding of some aspects of the disclosed embodiments, a simple summary is given below. This summary is not a general review, nor is it intended to identify key / important components or describe the scope of protection of these embodiments. Its only purpose is to present some concepts in a simple form as a preface to the detailed description that follows.

[0005] In a first aspect, an embodiment of the present application provides a method for detecting ground deformation based on phase gradient, comprising:

[0006] Synthetic aperture radar interferometry is used to obtain time series interference patterns;

[0007] Calculate deformation phase gradient based on time series interference pattern;

[0008] The calculated deformation phase gradients are superimposed, and an unwrapping operation is performed on the superimposed deformation phase gradients to obtain an unwrapped absolute phase gradient;

[0009] The unwrapped absolute phase gradient is input into the pre-trained local surface deformation recognition model to obtain the identified deformation area.

[0010] In an optional embodiment, calculating the deformation phase gradient according to the time series interference pattern includes:

[0011] The deformation phase gradient in the azimuth direction and the deformation phase gradient in the range direction are calculated based on the time series interference pattern.

[0012] In an optional embodiment, the calculated deformation phase gradients are superimposed, and an unwrapping operation is performed on the superimposed deformation phase gradients to obtain an unwrapped absolute phase gradient, including:

[0013] The deformation phase gradient in the azimuth direction and the deformation phase gradient in the distance direction are respectively superimposed to obtain the superimposed deformation phase gradient in the azimuth direction and the deformation phase gradient in the distance direction;

[0014] The superimposed azimuth deformation phase gradient and range deformation phase gradient are unwrapped to obtain the unwrapped azimuth absolute phase gradient and range absolute phase gradient.

[0015] In an optional embodiment, before inputting the unwrapped absolute phase gradient into a pre-trained local surface deformation recognition model to obtain the identified deformation area, the method further includes:

[0016] Construct a training data set and divide it into a training set, a validation set, and a test set;

[0017] Improve the YOLOv3 neural network to obtain an improved YOLOv3 neural network;

[0018] The local surface deformation recognition model is trained based on the training set, validation set, test set and the improved YOLOv3 neural network.

[0019] In an optional embodiment, the YOLOv3 neural network is improved to obtain an improved YOLOv3 neural network, including:

[0020] Add the attention mechanism module, shortcut module and Drop-Block module to the original YOLOv3 neural network to obtain the improved YOLOv3 neural network.

[0021] In an optional embodiment, an attention mechanism module, a shortcut module, and a Drop-Block module are added to the original YOLOv3 neural network to obtain an improved YOLOv3 neural network, including:

[0022] Embed the attention mechanism module behind each residual block in the YOLOv3 neural network;

[0023] Connect the adjacent network layers before and after the attention mechanism module through the shortcut module;

[0024] The Drop-Block module is embedded into the three detection layers in the YOLOv3 neural network to obtain the improved YOLOv3 neural network.

[0025] In an optional embodiment, it also includes:

[0026] The model is trained using the CIoU loss function.

[0027] In a second aspect, the embodiment of the present application further provides a surface deformation detection device based on phase gradient, comprising:

[0028] An interference pattern generation module, used for obtaining a time series interference pattern by synthetic aperture radar interferometry;

[0029] A phase gradient calculation module, used to calculate the deformation phase gradient according to the time series interference pattern;

[0030] A phase gradient superposition module is used to superimpose the calculated deformation phase gradient and perform an unwrapping operation on the superimposed deformation phase gradient to obtain the unwrapped absolute phase gradient;

[0031] The model recognition module is used to input the unwrapped absolute phase gradient into the pre-trained local surface deformation recognition model to obtain the recognized deformation area.

[0032] In a third aspect, an embodiment of the present application provides a surface deformation detection device based on phase gradient, comprising a processor and a memory storing program instructions, wherein the processor is configured to execute the surface deformation detection method based on phase gradient provided in the above embodiment when executing the program instructions.

[0033] In a fourth aspect, an embodiment of the present application provides a computer-readable medium having computer-readable instructions stored thereon, and the computer-readable instructions are executed by a processor to implement a surface deformation detection method based on phase gradient provided in the above embodiment.

[0034] The technical solution provided by the embodiments of the present application may have the following beneficial effects:

[0035] According to the surface deformation detection method based on phase gradient provided in the embodiment of the present application, the unwrapping of a large number of SAR interferograms is avoided through phase gradient stacking processing, which greatly improves the data processing efficiency and greatly reduces the computing resource cost. The influence of atmospheric effects and unwrapping errors in InSAR technology on the data processing results is avoided. The deformation signal in the gradient stacking result is clear and easy to distinguish, which is significantly different from the weak background information, reducing the difficulty of identifying the deformation signal. Furthermore, the target detection algorithm in the deep learning network is applied to the detection of local deformation information in InSAR, which can quickly detect local deformation information over a large range, has high reusability, and the processing results are not affected by the professional and technical knowledge of the data processing personnel. It not only improves the accuracy and reliability of surface deformation monitoring, but also improves the detection efficiency of local deformation signals over a large range.

[0036] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0038] Figure 1 is a flow chart of a method for detecting ground deformation based on phase gradient according to an exemplary embodiment;

[0039] Figure 2 is a schematic diagram of a method for detecting ground deformation based on phase gradient according to an exemplary embodiment;

[0040] Figure 3 is a schematic diagram showing a superimposed deformation phase gradient according to an exemplary embodiment;

[0041] Figure 4 is a schematic diagram of an improved YOLOv3 neural network according to an exemplary embodiment;

[0042] Figure 5 is a schematic structural diagram of a surface deformation detection device based on phase gradient according to an exemplary embodiment;

[0043] Figure 6 is a schematic structural diagram of a surface deformation detection device based on phase gradient according to an exemplary embodiment;

[0044] Figure 7 It is a schematic diagram of a computer storage medium according to an exemplary embodiment. DETAILED DESCRIPTION

[0045] The following description and the drawings sufficiently illustrate specific embodiments of the invention to enable those skilled in the art to practice them.

[0046] It should be clear that the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are only examples of systems and methods consistent with some aspects of the present invention as detailed in the attached claims.

[0048] With the accumulation of synthetic aperture radar data, it provides important observation support for the detection of surface deformation. However, how to efficiently and quickly process massive SAR image data and obtain local surface deformation in a wide area is still a challenge. In view of the data processing cost and unwrapping error caused by the need to unwrap a large number of interference patterns in the existing InSAR methods, as well as the atmospheric effects that are difficult to avoid in data processing, which leads to the difficulty in extracting surface deformation information in a wide area, this application proposes a method of using superimposed deformation phase gradients, calculating the deformation phase gradient in the interference pattern time series, and then superimposing the gradient signal in the time domain to achieve the purpose of enhancing the characteristics of the surface deformation signal and separating it from the background noise information. Since the unwrapping error caused by the unwrapping of a large number of interference patterns is avoided, and the random noise signal is greatly suppressed while the deformation signal is enhanced, the processing efficiency of SAR data is greatly improved, and the errors in data processing and the required computing resources are greatly reduced, so that this method can be effectively applied in the identification of deformation areas within a million square kilometers. For deformation information with obvious features and easy to distinguish, this method improves the classic target detection algorithm YOLOv3 to be suitable for efficient automatic detection of deformation signal targets obtained by phase gradient superposition, forming a complete technical process for extracting and automatically detecting local surface deformation information. For example, it can effectively identify active landslide clusters that are undergoing deformation over a large area, thereby further providing effective target areas for dynamic monitoring and prevention of landslide disasters.

[0049] The following is a detailed description of the ground deformation monitoring method based on neural network provided by the embodiment of the present application in conjunction with the accompanying drawings. Figure 1 , the method specifically includes the following steps.

[0050] S101 uses synthetic aperture radar interferometry to obtain time series interference patterns.

[0051] InSAR technology has outstanding advantages in large-area terrain mapping, such as all-day, all-weather, high precision, and high efficiency, and has important application value in rapid terrain mapping and updating. In one embodiment, InSAR technology is used to obtain a time series interference map of the area to be measured.

[0052] S102 calculates the deformation phase gradient according to the time series interference pattern.

[0053] In an optional embodiment, the deformation phase gradient in the azimuth direction and the deformation phase gradient in the range direction are calculated based on the time series interference pattern.

[0054] Specifically, the deformation phase gradient in the interference pattern is calculated by using the continuity of the deformation signal in time and space. After the interference pattern time series is obtained by interferometric measurement, the phase gradients in the azimuth and range directions are calculated by calculating the phase difference between adjacent pixels in the intertwined interference pattern. The azimuth direction represents the flight direction of the satellite, which is nearly north-south, and the range direction represents the direction perpendicular to the azimuth direction, which is nearly east-west.

[0055] Optionally, in addition to the adjacent pixel difference method, the phase gradient can also be calculated by using the difference between two pixels separated by multiple pixels, or by calculating the gradient in directions other than the azimuth and range directions, and then the same deformation signal enhancement effect can be achieved by superimposing the phase gradients of multiple interference patterns.

[0056] S103 superimposes the calculated deformation phase gradients, and performs an unwrapping operation on the superimposed deformation phase gradients to obtain an unwrapped absolute phase gradient.

[0057] In an optional embodiment, the deformation phase gradient in the azimuth direction and the deformation phase gradient in the range direction are superimposed respectively to obtain the superimposed deformation phase gradient in the azimuth direction and the deformation phase gradient in the range direction.

[0058] Specifically, the winding phase gradients in the azimuth and range directions are accumulated respectively, and are repeatedly wound to [-π,π], so that the accumulated phase gradient is in the interval [-π,π]. The superimposed azimuth deformation phase gradient and range deformation phase gradient can be calculated respectively according to the following formulas:

[0059]

[0060]

[0061] in, represents the superimposed phase gradient in azimuth, It represents the superimposed phase gradient in the distance upward, and Wrap represents the wrapping operation.

[0062] Furthermore, unwrapping operations are performed on the superimposed azimuth deformation phase gradient and range deformation phase gradient, respectively, to obtain the unwrapped azimuth absolute phase gradient and range absolute phase gradient.

[0063] In a possible implementation, an unwrapping algorithm in the prior art is used to perform unwrapping operations on the superimposed azimuth deformation phase gradient and the distance deformation phase gradient, respectively, to obtain the unwrapped azimuth absolute phase gradient and the distance absolute phase gradient. It is only necessary to perform an unwrapping operation on the superimposed phase once, without performing an unwrapping operation on each interference pattern, which greatly reduces the complexity of the calculation.

[0064] According to this step, in order to solve the problem of difficulty in extracting effective information of local deformation in the application of InSAR method in a wide area, a method of using superimposed deformation phase gradient is proposed. The deformation phase gradient is calculated in the interference pattern sequence, and then the gradient signal is superimposed in the time domain to enhance the active deformation signal and separate it from the background noise information.

[0065] In an exemplary scenario, since the gradient of the differential interferometric phase depends on the topography and deformation of the surface and the difference in the atmosphere during the two imagings, the phase gradient of the deformation shows the small-scale changes in the deformation gradient. Averaging or superimposing the phase gradient map is very useful for reducing the noise level of the data and eliminating atmospheric errors. The phase difference caused by the atmospheric state is very small between two adjacent pixels. In particular, after superimposing the phase gradient map, the temporally unrelated atmospheric signals are further offset, while the deformation gradient of the continuously occurring deformation signal is continuously enhanced. Therefore, the influence of atmospheric turbulence on the superimposed deformation gradient map can be ignored.

[0066] Assuming that the surface deformation is manifested as phase increase, the phase gradient gradually changes from positive (negative) to negative (positive) from the non-deformed (deformed) area to the deformed (non-deformed) area along the direction of the calculated phase gradient, which is visually manifested as a red-blue coupled gradient characteristic pattern. As the number of superimposed phase gradient maps increases, the gradient characteristics can be continuously strengthened.

[0067] like Figure 3 As shown, (a) shows the result of phase gradient superposition in the azimuth direction, and (b) shows the result of phase gradient superposition in the range direction. (c) shows the result of the prior art after superposition of the unwrapped interferogram, and (d) and (e) are enlarged images of the rectangular area in (a), showing the details of the superimposed phase gradient in the azimuth direction (left) and the range direction (middle), as well as the comparison of the result of the prior art unwrapped phase superposition (right).

[0068] Compared with the result obtained by superimposing the unwrapped phase, the local deformation signal (coupled red-blue features) in the phase gradient image can be clearly distinguished from the background noise, the features are more prominent and easier to identify.

[0069] Since the disentanglement of a large number of interference patterns is avoided and the deformation signal is enhanced while the noise signal is greatly suppressed, this method can be effectively applied to the identification of deformation areas within a million square kilometers. Then it can provide an effective target area for the analysis and judgment of deformation clustering areas, such as active landslide clustering areas.

[0070] S104 inputs the unwrapped absolute phase gradient into a pre-trained local surface deformation recognition model to obtain an identified deformation area.

[0071] In one possible implementation, the superimposed phase gradient makes the local deformation signal and the background noise significantly different. The rapid positioning of the deformation signal becomes a key issue that needs to be solved. The embodiment of the present application is based on the classic target detection algorithm YOLOv3, and has made corresponding improvements to the detection task of the deformation signal obtained by phase gradient superposition, so as to quickly and automatically detect the local deformation area with phase gradient characteristics.

[0072] Specifically, the YOLOv3 neural network is first improved by adding an attention mechanism module, a shortcut module and a Drop-Block module to the original YOLOv3 neural network to obtain an improved YOLOv3 neural network.

[0073] In one possible implementation, a Convolutional Block Attention Module (CBAM) is introduced and embedded behind each residual block in the YOLOv3 neural network. At the same time, in order to prevent the gradient dispersion caused by the increase in the number of network layers, the shortcut module is used to connect the adjacent network layers before and after the attention mechanism module. Figure 4 As shown in (a), this application adds channel and spatial attention mechanisms to the backbone feature extraction network of YOLOv3, improving the network's ability to extract and filter key features of the target, hereinafter referred to as Attention-YOLOv3.

[0074] Furthermore, in order to avoid overfitting, this application applies the regularization method to Attention-YOLOV3. Specifically, the Drop-Block module is placed in the three detection layers of Attention-YOLOv3. Figure 4 As shown in (b), the Drop-Block module discards neurons in adjacent areas of the feature map together with a certain probability to reduce the fitting ability of the network.

[0075] Through the above improvements, the improved YOLOv3 neural network is obtained, as shown in Figure 4 As shown, (a) represents a residual block with a convolutional attention module. The white and dark units in (b) represent inactivated and activated neurons, respectively, and the black crosses represent randomly dropped neurons. (c) represents a CIoU graph for calculating network loss. In an embodiment of the present application, the CIoU loss is used to replace the MSE (mean square error) loss in YOLOv3. The CIoU loss takes into account the overlapping area between target detection frames, the center point distance, and the aspect ratio between the detection frames, thereby improving the accuracy of target detection.

[0076] Furthermore, a training data set is constructed and divided into a training set, a validation set and a test set. The local surface deformation recognition model is trained based on the training set, the validation set, the test set and the improved YOLOv3 neural network.

[0077] Specifically, for the dataset, by considering the specific terrain, the local deformation phase gradient anomaly signals with red-blue coupling were manually marked. A total of 712 azimuth and 581 range gradient images were selected, and the dataset was divided into training, validation, and test datasets in a ratio of 8:1:1. The label smoothing method was also applied to smooth the classification labels and shorten the distance between labels within the category to enhance the generalization ability of the network. The total training duration was 300 rounds. In the first 100 rounds of training, the Adam automatic optimizer was used to find the approximate convergence direction. In the next 200 rounds, the SGDMomentum optimizer was used to carefully adjust the learning rate to achieve better training results. The k-means clustering method was used to divide the width and height of all labeled targets in the dataset into 9 categories of anchors to reduce the regression loss in training. Table 1 shows the clustering results:

[0078] Table.1 Anchor size settings during training (unit: pixel)

[0079]

[0080] After network training, the MAP is 0.75, the recall rate and the precision rate are 0.87 and 0.95 respectively. Compared with the original YOLOv3 network, the detection ability of the target is greatly improved.

[0081] Table 2. Comparison of deep learning network detection performance

[0082]

[0083] Through the above method, a trained local surface deformation recognition model is obtained, and the unwrapped absolute phase gradient is input into the trained local surface deformation recognition model. The model outputs the four vertex coordinates of the target detection frame of the identified deformation area to obtain the identified deformation area.

[0084] This application effectively combines the proposed new InSAR technology and deep learning technology to efficiently, automatically and quickly extract local surface information within a million square kilometers, such as early identification of landslide disasters across the country. Compared with traditional InSAR technology, it greatly reduces computing resources and labor costs, and has a wider deformation detection range.

[0085] In order to facilitate understanding of the surface deformation detection method based on phase gradient provided in the embodiment of the present application, the following is a Figure 2 For explanation. Figure 2 As shown, the method includes the following steps.

[0086] Firstly, the interferogram sequence is obtained by InSAR technology. After obtaining the time series interferogram, the deformation phase gradients are calculated in azimuth and range directions, respectively, and then superimposed to obtain the stacked azimuth winding gradient and range winding gradient. Then, the superimposed phase gradients are unwrapped to obtain the unwrapped azimuth absolute phase gradient and range absolute phase gradient.

[0087] Furthermore, the unwrapped absolute phase gradient is input into the trained neural network model, and the local deformation gradient anomaly detection is performed using the deep learning network to output the rectangular coordinate information of the deformation area.

[0088] According to the surface deformation detection method based on phase gradient provided in the embodiment of the present application, the unwrapping of a large number of SAR interferograms is avoided through phase gradient stacking processing, which greatly improves the data processing efficiency and greatly reduces the computing resource cost. The influence of atmospheric effects and unwrapping errors in InSAR technology on the data processing results is avoided. The deformation signal in the gradient stacking result is clear and easy to distinguish, which is significantly different from the weak background information, reducing the difficulty of identifying the deformation signal. Furthermore, the target detection algorithm in the deep learning network is applied to the detection of local deformation information in InSAR, which can quickly detect local deformation information over a large range, has high reusability, and the processing results are not affected by the professional and technical knowledge of the data processing personnel. It not only improves the accuracy and reliability of surface deformation monitoring, but also improves the detection efficiency of local deformation signals over a large range.

[0089] The present application also provides a surface deformation detection device based on phase gradient, which is used to perform the surface deformation detection method based on phase gradient of the above embodiment, such as Figure 5As shown, the device comprises:

[0090] An interference pattern generation module 501 is used to obtain a time series interference pattern by synthetic aperture radar interferometry;

[0091] A phase gradient calculation module 502, used to calculate the deformation phase gradient according to the time series interference pattern;

[0092] The phase gradient superposition module 503 is used to superimpose the calculated deformation phase gradient and perform an unwrapping operation on the superimposed deformation phase gradient to obtain an unwrapped absolute phase gradient;

[0093] The model recognition module 504 is used to input the unwrapped absolute phase gradient into a pre-trained local surface deformation recognition model to obtain an identified deformation area.

[0094] It should be noted that the surface deformation detection device based on phase gradient provided in the above embodiment only uses the division of the above functional modules as an example when executing the surface deformation detection method based on phase gradient. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the surface deformation detection device based on phase gradient provided in the above embodiment and the surface deformation detection method based on phase gradient belong to the same concept. The implementation process is detailed in the method embodiment and will not be repeated here.

[0095] The embodiment of the present application also provides an electronic device corresponding to the surface deformation detection method based on phase gradient provided in the above embodiment, so as to execute the above surface deformation detection method based on phase gradient.

[0096] Please refer to Figure 6 , which shows a schematic diagram of an electronic device provided by some embodiments of the present application. Figure 6 As shown, the electronic device includes: a processor 600, a memory 601, a bus 602 and a communication interface 603, and the processor 600, the communication interface 603 and the memory 601 are connected via the bus 602; the memory 601 stores a computer program that can be run on the processor 600, and when the processor 600 runs the computer program, it executes the phase gradient-based surface deformation detection method provided in any of the aforementioned embodiments of the present application.

[0097] The memory 601 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 603 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0098] The bus 602 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory 601 is used to store programs, and the processor 600 executes the programs after receiving the execution instructions. The surface deformation detection method based on phase gradient disclosed in any implementation of the above-mentioned embodiment of the present application may be applied to the processor 600, or implemented by the processor 600.

[0099] The processor 600 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 600. The above processor 600 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor to be executed, or the hardware and software modules in the decoding processor can be executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 601, and the processor 600 reads the information in the memory 601 and completes the steps of the above method in combination with its hardware.

[0100] The electronic device provided in the embodiment of the present application and the surface deformation detection method based on phase gradient provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, operated or implemented therein.

[0101] The present application also provides a computer-readable storage medium corresponding to the surface deformation detection method based on phase gradient provided in the above-mentioned embodiment. Figure 7 The computer-readable storage medium shown is a CD 700 on which a computer program (ie, a program product) is stored. When the computer program is run by a processor, it executes the surface deformation detection method based on phase gradient provided in any of the aforementioned embodiments.

[0102] It should be noted that examples of computer-readable storage media may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical or magnetic storage media, which are not listed here one by one.

[0103] The computer-readable storage medium provided in the above-mentioned embodiments of the present application and the surface deformation detection method based on phase gradient provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0104] The technical features of the above embodiments may be combined arbitrarily. 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.

[0105] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the attached claims.

Claims

1. A method for detecting ground deformation based on phase gradient, characterized in that ,include: Synthetic aperture radar interferometry is used to obtain time series interference patterns; Calculating deformation phase gradient according to the time series interference pattern; superimposing the calculated deformation phase gradients, and performing an unwrapping operation on the superimposed deformation phase gradients to obtain an unwrapped absolute phase gradient; Construct a training data set, divide the training data set into a training set, a validation set and a test set; improve the YOLOv3 neural network to obtain an improved YOLOv3 neural network; including: adding an attention mechanism module, a shortcut module and a Drop-Block module to the original YOLOv3 neural network, embedding the attention mechanism module behind each residual block in the YOLOv3 neural network; connecting the adjacent network layers before and after the attention mechanism module through the shortcut module; embedding the Drop-Block module into the three detection layers in the YOLOv3 neural network to obtain the improved YOLOv3 neural network; A local surface deformation recognition model is trained according to the training set, the validation set, the test set and the improved YOLOv3 neural network; the unwrapped absolute phase gradient is input into the pre-trained local surface deformation recognition model to obtain an identified deformation area.

2. The method according to claim 1, characterized in that Calculating the deformation phase gradient according to the time series interference pattern includes: The deformation phase gradient in the azimuth direction and the deformation phase gradient in the range direction are calculated according to the time series interference pattern.

3. The method according to claim 2, characterized in that The calculated deformation phase gradient is superimposed, and an unwrapping operation is performed on the superimposed deformation phase gradient to obtain an unwrapped absolute phase gradient, including: The deformation phase gradient in the azimuth direction and the deformation phase gradient in the distance direction are respectively superimposed to obtain the superimposed deformation phase gradient in the azimuth direction and the deformation phase gradient in the distance direction; The superimposed azimuth deformation phase gradient and range deformation phase gradient are unwrapped to obtain the unwrapped azimuth absolute phase gradient and range absolute phase gradient.

4. The method according to claim 1, characterized in that: Also includes: The model is trained using the CIoU loss function.

5. A surface deformation detection device based on phase gradient, characterized in that ,include: An interference pattern generation module, used for obtaining a time series interference pattern by synthetic aperture radar interferometry; A phase gradient calculation module, used for calculating the deformation phase gradient according to the time series interference pattern; A phase gradient superposition module, used for superimposing the calculated deformation phase gradient and performing an unwrapping operation on the superimposed deformation phase gradient to obtain an absolute phase gradient after unwrapping; The model recognition module is used to construct a training data set, divide the training data set into a training set, a validation set and a test set; improve the YOLOv3 neural network to obtain an improved YOLOv3 neural network; including: adding an attention mechanism module, a shortcut module and a Drop-Block module to the original YOLOv3 neural network, embedding the attention mechanism module behind each residual block in the YOLOv3 neural network; connecting the adjacent network layers before and after the attention mechanism module through the shortcut module; embedding the Drop-Block module into the three detection layers in the YOLOv3 neural network to obtain the improved YOLOv3 neural network; training a local surface deformation recognition model according to the training set, the validation set, the test set and the improved YOLOv3 neural network; inputting the unwrapped absolute phase gradient into the pre-trained local surface deformation recognition model to obtain an identified deformation area.

6. A surface deformation detection device based on phase gradient, characterized in that: It comprises a processor and a memory storing program instructions, wherein the processor is configured to execute the surface deformation detection method based on phase gradient as claimed in any one of claims 1 to 4 when executing the program instructions.

7. A computer-readable medium, characterized in that Computer-readable instructions are stored thereon, and the computer-readable instructions are executed by a processor to implement a surface deformation detection method based on phase gradient as described in any one of claims 1 to 4.

Images