Weak change detection method based on multi-temporal synthetic aperture radar image

Abstract

The invention discloses a weak change detection method based on a multi-temporal synthetic aperture radar image, which comprises the following steps of: S1, acquiring two multi-temporal SAR images in different periods in the same region, and respectively recording the two multi-temporal SAR images as # imgabs0 # and # imgabs1 # S2, performing multi-scale decomposition, noise removal and contrast enhancement on the SAR image acquired in the previous step, s3, for the preprocessed images, taking one image with higher image quality as a reference image, and taking the other image as an image to be registered to carry out accurate alignment, so as to obtain a high-quality preprocessed image # imgabs2 and a high-quality preprocessed image # imgabs3 # S3, for the preprocessed images, taking the image with higher image quality as a reference image, and taking the other image as an image to be registered to carry out accurate alignment; s4, according to the registered image, a difference graph is obtained through coherent statistics estimation; s5, performing change area classification on the difference image by using a statistical threshold to obtain a change area; and S6, obtaining a detection result and performing optimization in combination with context information. According to the method, the multi-temporal SAR image is preprocessed and enhanced, and the key point extraction algorithm is adopted for accurate registration, so that false alarm and leak detection are effectively reduced, and the weak change detection performance is improved.

Description

Technical Field

[0001] The invention relates to the field of radar remote sensing technology, and more specifically, to a method for detecting weak changes based on multi-phase synthetic aperture radar images. Background Art

[0002] Synthetic Aperture Radar (SAR) is an active microwave sensing system that increases the azimuth resolution through long-distance synthetic aperture in the azimuth dimension, and obtains high resolution in the distance dimension by emitting a large-bandwidth linear frequency modulation signal and using the principle of pulse compression, and finally obtains a two-dimensional high-resolution image. Different from optical and infrared images, SAR images are complex images. The amplitude part reflects the back reflection ability of the ground object and the target structure characteristics, while the phase part reflects the electromagnetic scattering characteristics of the ground object. Therefore, SAR images contain a lot of spatial information such as the surface characteristics of the ground object.

[0003] As people gradually improve the accuracy of data acquisition, the development and application of weak change detection technology has been promoted. The detection of footprints, ruts, and tracks left by vehicles caused by human activities can effectively reflect the activity trajectory, so that abnormal people and vehicles can be dynamically and continuously monitored. These activity trajectory data contain rich information, which can be of guiding significance for the intention of human activities and the search and rescue of trapped people in disaster areas.

[0004] Coherence change detection (CCD) uses both the amplitude and phase information of SAR images to obtain a difference map by calculating coherence statistics, and then detects weak changes in the scene, thereby realizing the detection of weak traces, especially in scenes where low signal intensity targets need to be identified or low contrast changes need to be detected. Traditional SAR image processing methods are difficult to accurately detect weak target signals, mainly because weak signals are often masked by high background noise. In order to solve this problem, it is necessary to propose a new detection method and system that can efficiently and accurately extract weak trace information. Summary of the invention

[0005] The present invention aims to provide a method for improving the accuracy of weak change detection in synthetic aperture radar images, so as to solve the problem that the results of traditional SAR image target detection algorithms contain a large number of false alarms and missed detections, which seriously affects the weak target detection performance of SAR images. Therefore, by preprocessing and enhancing multi-phase SAR images, using key point extraction algorithms for accurate registration, false alarms and missed detections are effectively reduced, and weak change detection performance is improved.

[0006] To solve the above technical problems, the object of the present invention is achieved as follows: a method for detecting weak changes based on multi-temporal synthetic aperture radar images, comprising the following steps:

[0007] S1. Collect two multi-temporal SAR images of the same area at different times, respectively and

[0008] S2. Perform multi-scale decomposition, noise removal and contrast enhancement on the SAR image acquired in the previous step to obtain a high-quality pre-processed image and

[0009] S3. For the preprocessed images, one with higher image quality is used as the reference image, and the other is used as the image to be registered for precise alignment;

[0010] S4. according to the registered images, obtain a difference map by using coherence statistics estimation;

[0011] S5. classifying the change regions of the difference map using a statistical threshold to obtain the change regions;

[0012] S6. Obtain the detection results and optimize them in combination with context information.

[0013] The present invention is further configured as follows: the specific method of step S1 is: collecting two multi-phase SAR images of the same region at different times, respectively recorded as and Respectively represent the image data collected at time t1 and t2; in order to facilitate subsequent processing, the collected images should maintain consistent geometric perspective and sensor parameter settings.

[0014] The present invention is further configured as follows: the specific method of step S2 is:

[0015] SAR image preprocessing: the original SAR image and Perform preprocessing to increase the difference between signal and background; this step includes:

[0016] (a) Multiscale decomposition: Decompose the SAR image into different scales to better separate the target and background information; use methods such as wavelet transform or Laplace pyramid decomposition to obtain multiscale image representation;

[0017] The formula is:

[0018]

[0019] Among them, I s(x, y) is the image after multi-scale image decomposition, W i is the weight of each scale, s i is the scale parameter, f(I(x,y),s i ) is the scale decomposition function;

[0020] (b) Noise removal: Use a filter to remove noise, such as mean filtering, median filtering or adaptive filtering; the formula is:

[0021] I denoise (x,y)=g(I s (x,y)

[0022] Among them, I denoise (x, y) represents the denoised image, g(·) represents the filter function;

[0023] (c) Contrast enhancement: Use histogram equalization or contrast stretching technology to enhance the contrast of the image to highlight the area with slight changes; the formula is:

[0024] I enhance (x,y)=h(I denoise (x,y)

[0025] h(·) is the contrast enhancement operation.

[0026] The present invention is further configured as follows: the specific method of step S3 is: due to terrain fluctuations, satellite attitude changes, etc., it is necessary to collect pre-processed images at different times. and Perform high-precision registration; select the image with relatively better image quality among the two images as the reference image for registration; use the registration technology based on feature point matching to accurately superimpose SAR images at different times to ensure that the corresponding pixels represent the same surface point, which is convenient for subsequent weak trace detection and analysis;

[0027] Feature point matching can use algorithms such as scale-invariant feature transform (SIFT) and speeded up robust features (SURF); the formula is expressed as:

[0028]

[0029] Where T(·) is the registration transformation, Represents the aligned image.

[0030] The present invention is further configured as follows: the specific method of step S4 is: assuming that the image pair after registration The target area is separated from the background using a feature-based algorithm to generate a difference map. The basic method for generating a difference map is to obtain a difference map by calculating the coherence statistics of the registered image and extract the changed area. The calculation formula is as follows:

[0031]

[0032] It is a difference map, where each pixel represents the degree of change between two phases.

[0033] The present invention is further configured as follows: the specific method of step S5 is: change difference map analysis: by analyzing the difference map The target is identified by the change characteristics in the image; the statistical threshold is used to classify the change area; the threshold classification formula is as follows:

[0034]

[0035] Among them, M(x, y) is the change mask image, and θ is the coherence threshold of change detection. By setting this threshold, significant change areas can be identified.

[0036] The present invention is further configured as follows: the specific manner of the step S6 is: optimizing the detection result by using the context information, including an optimization method based on spatiotemporal consistency, the formula of which is as follows, to further improve the robustness and accuracy of the detection;

[0037] M opt (x,y)=φ(M(x,y),neigh(x,y))

[0038] Among them, M opt (x, y) represents the optimized change mask, neigh(x, y) is the neighborhood information around the pixel, and φ(·) is the spatial consistency optimization operation.

[0039] In summary, the present invention has the following beneficial effects: the method for detecting weak changes based on multi-phase synthetic aperture radar images involved in the present invention can accurately detect weak changes by analyzing the differences of SAR images in different phases through coherent statistics. Compared with traditional methods, this method uses phase information to effectively improve the sensitivity of change detection and is suitable for monitoring weak changes in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is the overall flow chart of the present invention;

[0041] Figure 2 is the SAR imaging result of a certain area at time T1;

[0042] Figure 3is the SAR imaging result of a certain area at time T2;

[0043] Figure 4 This is the result diagram of weak change detection. DETAILED DESCRIPTION

[0044] In order to enable those skilled in the art to better understand the technical solution of the present invention, the preferred embodiments of the present invention are described below in conjunction with specific embodiments, but it should be understood that these descriptions are only for further illustrating the features and advantages of the present invention, rather than limiting the patent requirements of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present invention.

[0045] The present invention is further described below in conjunction with the accompanying drawings and preferred embodiments.

[0046] Example 1

[0047] See also Figures 1 to 4 As shown, a method for detecting weak changes based on multi-temporal synthetic aperture radar images according to this embodiment includes the following steps:

[0048] S1. Collect two multi-temporal SAR images of the same area at different times, respectively and

[0049] S2. Perform multi-scale decomposition, noise removal and contrast enhancement on the SAR image acquired in the previous step to obtain a high-quality pre-processed image and

[0050] S3. For the preprocessed images, one with higher image quality is used as the reference image, and the other is used as the image to be registered for precise alignment;

[0051] S4. according to the registered images, obtain a difference map by using coherence statistics estimation;

[0052] S5. classifying the change regions of the difference map using a statistical threshold to obtain the change regions;

[0053] S6. Obtain the detection results and optimize them in combination with context information.

[0054] The present invention is further configured as follows: the specific method of step S1 is: collecting two multi-phase SAR images of the same region at different times, respectively recorded as and Respectively represent the image data collected at time t1 and t2; in order to facilitate subsequent processing, the collected images should maintain consistent geometric perspective and sensor parameter settings.

[0055] The present invention is further configured as follows: the specific method of step S2 is:

[0056] SAR image preprocessing: the original SAR image and Perform preprocessing to increase the difference between signal and background; this step includes:

[0057] (d) Multiscale decomposition: Decompose the SAR image into different scales to better separate the target and background information; use methods such as wavelet transform or Laplace pyramid decomposition to obtain multiscale image representation;

[0058] The formula is:

[0059]

[0060] Among them, I s (x, y) is the image after multi-scale image decomposition, W i is the weight of each scale, s i is the scale parameter, f(I(x,y),s i ) is the scale decomposition function;

[0061] (e) Noise removal: Use a filter to remove noise, such as mean filtering, median filtering or adaptive filtering; the formula is:

[0062] I denoise (x,y)=g(I s (x,y)

[0063] Among them, I denoise (x, y) represents the denoised image, g(·) represents the filter function;

[0064] (f) Contrast enhancement: Use histogram equalization or contrast stretching technology to enhance the contrast of the image.

[0065] To highlight the area of ​​weak changes; the formula is:

[0066] I enhance (x,y)=h(I denoise (x,y)

[0067] h(·) is the contrast enhancement operation.

[0068] The present invention is further configured as follows: the specific method of step S3 is: due to terrain fluctuations, satellite attitude changes, etc., it is necessary to collect pre-processed images at different times. and Perform high-precision registration; select the image with relatively better image quality among the two images as the reference image for registration; use the registration technology based on feature point matching to accurately superimpose SAR images at different times to ensure that the corresponding pixels represent the same surface point, which is convenient for subsequent weak trace detection and analysis;

[0069] Feature point matching can use algorithms such as scale-invariant feature transform (SIFT) and speeded up robust features (SURF); the formula is expressed as:

[0070]

[0071] Where T(·) is the registration transformation, Represents the aligned image.

[0072] The present invention is further configured as follows: the specific method of step S4 is: assuming that the image pair after registration The target area is separated from the background using a feature-based algorithm to generate a difference map. The basic method for generating a difference map is to obtain a difference map by calculating the coherence statistics of the registered image and extract the changed area. The calculation formula is as follows:

[0073]

[0074] It is a difference map, where each pixel represents the degree of change between two phases.

[0075] The present invention is further configured as follows: the specific method of step S5 is: change difference map analysis: by analyzing the difference map The target is identified by the change characteristics in the image; the statistical threshold is used to classify the change area; the threshold classification formula is as follows:

[0076]

[0077] Among them, M(x, y) is the change mask image, and θ is the coherence threshold of change detection. By setting this threshold, significant change areas can be identified.

[0078] The present invention is further configured as follows: the specific manner of the step S6 is: optimizing the detection result by using the context information, including an optimization method based on spatiotemporal consistency, the formula of which is as follows, to further improve the robustness and accuracy of the detection;

[0079] M opt (x,y)=φ(M(x,y),neigh(x,y))

[0080] Among them, M opt(x, y) represents the optimized change mask, neigh(x, y) is the neighborhood information around the pixel, and φ(·) is the spatial consistency optimization operation.

[0081] The method for detecting weak changes based on multi-phase synthetic aperture radar images involved in the present invention can accurately detect weak changes by analyzing the differences of SAR images in different phases through coherent statistics. Compared with traditional methods, this method uses phase information to effectively improve the sensitivity of change detection, is suitable for monitoring weak changes in complex environments, has complete overall functions, and is highly practical.

[0082] Unless otherwise specified, in the present invention, the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicating orientation or positional relationship are based on the orientation or positional relationship actually shown, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the orientation or positional relationship in the present invention are only used for illustrative purposes and cannot be understood as limiting this patent. For ordinary technicians in this field, the specific meanings of the above terms can be understood in combination with the embodiments and according to specific circumstances.

[0083] Unless otherwise clearly specified and limited, in the present invention, the terms "disposed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0084] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for detecting weak changes based on multi-temporal synthetic aperture radar images, characterized in that: The following steps are involved: S1. Collect two multi-temporal SAR images of the same area at different times, respectively and S2. Perform multi-scale decomposition, noise removal and contrast enhancement on the SAR image acquired in the previous step to obtain a high-quality pre-processed image and S3. For the preprocessed images, one with higher image quality is used as the reference image, and the other is used as the image to be registered for precise alignment; S4. according to the registered images, obtain a difference map by using coherence statistics estimation; S5. classifying the change regions of the difference map using a statistical threshold to obtain the change regions; S6. Obtain the detection results and optimize them in combination with context information.

2. The method for detecting weak changes based on multi-temporal synthetic aperture radar images according to claim 1, characterized in that: The specific method of step S1 is: collect two multi-temporal SAR images of the same region at different times, respectively recorded as and They represent the image data collected at time t1 and t2 respectively; in order to facilitate subsequent processing, the collected images should maintain consistent geometric perspective and sensor parameter settings.

3. The method for detecting weak changes based on multi-temporal synthetic aperture radar images according to claim 1, characterized in that: The specific method of step S2 is as follows: SAR image preprocessing: the original SAR image and Perform preprocessing to increase the difference between signal and background; this step includes: (a) Multiscale decomposition: Decompose the SAR image into different scales to better separate the target and background information; use methods such as wavelet transform or Laplace pyramid decomposition to obtain multiscale image representation; The formula is: Among them, I s (x, y) is the image after multi-scale image decomposition, W i is the weight of each scale, s i is the scale parameter, f(I(x,y),s i ) is the scale decomposition function; (b) Noise removal: Use a filter to remove noise, such as mean filtering, median filtering or adaptive filtering; the formula is: I denoise (x,y)=g(I s (x,y)) Among them, I denoise (x, y) represents the denoised image, g(·) represents the filter function; (c) Contrast enhancement: Use histogram equalization or contrast stretching technology to enhance the contrast of the image. To highlight the area of ​​weak changes; the formula is: I enhance (x,y)=h(I denoise (x,y)) h(·) is the contrast enhancement operation.

4. The method for detecting weak changes based on multi-temporal synthetic aperture radar images according to claim 1, characterized in that: The specific method of step S3 is: due to terrain fluctuations, satellite attitude changes, etc., it is necessary to collect pre-processed images at different times. and Perform high-precision registration; select one of the two images with relatively better image quality as the reference image for registration; The registration technology based on feature point matching is used to accurately overlay SAR images taken at different times to ensure that corresponding pixels represent the same surface point, which is convenient for subsequent weak trace detection and analysis; Feature point matching can use algorithms such as scale-invariant feature transform (SIFT) and speeded up robust features (SURF); the formula is expressed as: Where T(·) is the registration transformation, Represents the aligned image.

5. The method for detecting weak changes based on multi-temporal synthetic aperture radar images according to claim 1, characterized in that: The specific method of step S4 is: assuming that the image pair after registration is The target area is separated from the background using a feature-based algorithm to generate a difference map. The basic method for generating a difference map is to obtain a difference map by calculating the coherence statistics of the registered image and extract the changed area. The calculation formula is as follows: It is a difference map, where each pixel represents the degree of change between two phases.

6. The method for detecting weak changes based on multi-temporal synthetic aperture radar images according to claim 1, characterized in that: The specific method of step S5 is: Change difference map analysis: by analyzing the difference map The target is identified by the change characteristics in the image; the statistical threshold is used to classify the change area; the threshold classification formula is as follows: Among them, M(x, y) is the change mask image, and θ is the coherence threshold of change detection. By setting this threshold, significant change areas can be identified.

7. The method for detecting weak changes based on multi-temporal synthetic aperture radar images according to claim 1, characterized in that: The specific method of step S6 is: optimizing the detection result by using the context information, including an optimization method based on spatiotemporal consistency, the formula of which is as follows, to further improve the robustness and accuracy of the detection; M opt (x,y)=φ(M(x,y),neigh(x,y)) Among them, M opt (x, y) represents the optimized change mask, neigh(x, y) is the neighborhood information around the pixel, and φ(·) is the spatial consistency optimization operation.

Images