A vehicle target detection method and device based on SAR image

Abstract

The present invention provides a vehicle target detection method and device based on SAR images, which improves the vehicle target detection performance under non-uniform strong clutter conditions and reduces the influence of non-uniform ground background on vehicle target detection by performing multiple overlapping segmentation on SAR images.

Description

Technical Field

[0001] The invention belongs to the technical field of radar target detection, and in particular relates to a vehicle target detection method and device based on SAR images. Background Art

[0002] Whether in the military or civilian fields, the acquisition platforms and channels of remote sensing SAR images are becoming more and more diverse, and the quality of SAR images is getting higher and higher. Remote sensing SAR image acquisition also has the advantages of all-day and all-weather, so its application is growing rapidly. Vehicle target detection based on SAR images is one of the key technologies currently shared in the military and civilian fields, especially the challenge of vehicle target detection under non-uniform strong ground clutter conditions.

[0003] Classic SAR image-based vehicle target detection methods usually use constant false alarm processing technology based on statistical models. Compared with machine learning-based target detection methods, they have the advantages of no need for training samples, relatively simple reasoning and decision-making, and low algorithm deployment costs. However, when faced with non-uniform strong ground clutter conditions, the direct use of SAR image vehicle target detection methods based on statistical models has the problem of high and low signal-to-clutter ratios, resulting in a simultaneous deterioration of false alarm probability and detection probability. Summary of the invention

[0004] The technical problem to be solved by the present invention is how to improve the robustness of vehicle target detection based on SAR images and reduce the false alarm probability. A vehicle target detection method and device based on SAR images are proposed.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A vehicle target detection method based on SAR images comprises the following steps:

[0007] Step 1: Acquire SAR images;

[0008] Step 2: segmenting the SAR image in a manner of multiple overlapping segmentation frames;

[0009] Step 3: Perform vehicle target detection on the SAR image within each segmented area, where the segmented area refers to the area segmented by each segmentation box;

[0010] Step 4: Fuse the vehicles detected by the SAR image targets in all segmented areas and give the vehicle target detection results.

[0011] Furthermore, the multiple overlaps refer to that each segmentation frame sequentially performs overlapping segmentation on the SAR image in radial and lateral dimensions according to a certain overlap ratio.

[0012] Furthermore, the overlap ratio refers to the ratio of the overlapped portion of the next segmentation frame and the current segmentation frame to the current segmentation frame.

[0013] Furthermore, the size of the segmentation frame for segmenting the SAR image is configured according to the maximum vertical and horizontal dimensions of the vehicle target to be detected. The specific size of the segmentation frame is

[0014] ,

[0015] in Indicates the multiple of the segmentation box size compared to the maximum size of the vehicle target to be detected. Indicates the magnification of the number of radial resolution units of the SAR image occupied by the vehicle target to be detected, Indicates the magnification of the number of lateral resolution units of the SAR image occupied by the vehicle target to be detected, and They respectively represent the number of radial and lateral resolution units occupied by the vehicle target to be detected in the SAR image.

[0016] Furthermore, before performing vehicle target detection on the SAR image in each segmented area in step 3, target enhancement is performed on the SAR image in each segmented area, and then vehicle target detection is performed on the target enhanced SAR image in each segmented area.

[0017] Furthermore, performing target enhancement on the SAR image in each segmented region in step 3 refers to selecting a corresponding target enhancement method to perform target enhancement according to the histogram distribution of the SAR image amplitude values ​​in each segmented region.

[0018] Furthermore, the method for performing vehicle target detection is to adopt a constant false alarm processing method based on a statistical model.

[0019] Furthermore, the method for fusing the vehicles detected by the SAR image targets in all segmented areas is:

[0020] For each resolution unit in the SAR image, except for the edge resolution unit of the entire SAR image, the detection results of each resolution unit in different coverage segmentation frames are comprehensively detected and judged using the majority voting decision fusion operator;

[0021] For the edge resolution unit of the entire SAR image, the detection result is directly output without fusion.

[0022] The present invention also proposes a vehicle target detection device based on SAR images, comprising the following modules:

[0023] Image acquisition module: used to acquire SAR images;

[0024] Segmentation module: used for segmenting the SAR image in a manner of multiple overlapping segmentation frames;

[0025] Target enhancement module: used to enhance the target of the SAR image in each segmented area, where the segmented area refers to the area segmented by each segmentation box;

[0026] Target detection module: used to detect vehicle targets in each segmented area of ​​the SAR image after target enhancement;

[0027] Detection output module: used to fuse the vehicles detected by SAR image targets in all segmented areas and give vehicle target detection results.

[0028] By adopting the above technical solution, the present invention has the following beneficial effects:

[0029] The present invention provides a vehicle target detection method and device based on SAR images, which improves the vehicle target detection performance under non-uniform strong clutter conditions and reduces the influence of non-uniform ground background on vehicle target detection by performing multiple overlapping segmentation on SAR images. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a flow chart of the system of the present invention;

[0031] Figure 2 This is a schematic diagram of multiple overlaps. DETAILED DESCRIPTION

[0032] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the 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.

[0033] Figure 1 and Figure 2 The present invention shows a vehicle target detection method based on SAR images, such as Figure 1 As shown, the following steps are included:

[0034] Step 1: Acquire SAR image.

[0035] In this embodiment, vehicle target detection is performed through SAR images, and a specific solution is given to the problem that the current SAR image vehicle target detection method, when facing non-uniform strong ground clutter conditions, directly uses the SAR image vehicle target detection method based on the statistical model, which has the problem of fluctuating signal-to-clutter ratio, resulting in the simultaneous deterioration of the false alarm probability and the detection probability.

[0036] Step 2: Segment the SAR image in a manner of multiple overlapping segmentation frames.

[0037] In this embodiment, Figure 2 As shown, each color frame represents a segmentation frame, and the adjacent two color frames are overlapped according to a certain overlap ratio and then sequentially segmented outwards. The multiple overlaps refer to the overlapping segmentation of the SAR image from the radial and lateral dimensions by each segmentation frame in sequence according to a certain overlap ratio. The overlap ratio refers to the proportion of the overlap between the next segmentation frame and the current segmentation frame to the current segmentation frame. This embodiment reduces the influence of the non-uniform ground background on vehicle target detection by segmenting the SAR image, and improves the vehicle target detection performance under non-uniform strong clutter conditions.

[0038] In this embodiment, the size of the segmentation frame for segmenting the SAR image is configured according to the maximum vertical and horizontal dimensions of the vehicle target to be detected. The specific size of the segmentation frame is

[0039] ,

[0040] in Indicates the multiple of the segmentation box size compared to the maximum size of the vehicle target to be detected. Indicates the magnification of the number of radial resolution units of the SAR image occupied by the vehicle target to be detected, Indicates the magnification of the number of lateral resolution units of the SAR image occupied by the vehicle target to be detected, and They respectively represent the number of radial and lateral resolution units occupied by the vehicle target to be detected in the SAR image. , . and Respectively represent the maximum length and width of the physical size of the vehicle target to be detected, in meters; and They represent the radial and lateral resolutions of the SAR image, also in meters; represents the ceiling function. Set the segmentation box size to In order to avoid target fragmentation and subsequent constant false alarm processing based on statistical models, when performing regional segmentation on SAR images, The selection of should make the segmentation area as large as possible than the maximum longitudinal and lateral dimensions of the vehicle target to be detected; on the other hand, in order to reduce the non-uniformity of the background, The choice of should make the segmentation area as small as possible. To strike a balance between the two, the default value can be .

[0041] In this embodiment, Figure 2As shown in the figure, when the SAR image is multi-overlapped and segmented according to a certain overlap ratio, the overlap ratio The larger the value, the more segmentation boxes there are, and the more robust the subsequent decision fusion for target detection is, but the computational complexity is higher. To strike a balance between the two, the default value can be used in the radial and horizontal dimensions respectively. The overlapping method is used for segmentation, so except for the edge of the SAR image, all other resolution units are The number of split frames included.

[0042] Step 3: Perform vehicle target detection on the SAR image within each segmented area, where the segmented area refers to the area segmented by each segmentation box.

[0043] In this embodiment, before performing vehicle target detection on the SAR image in each segmented area in step 3, target enhancement is also required for the SAR image in each segmented area, followed by vehicle target detection on the SAR image in each segmented area after target enhancement. Performing target enhancement on the SAR image in each segmented area means performing target enhancement using a corresponding target enhancement method according to the histogram distribution of the SAR image amplitude values ​​in each segmented area. By performing target enhancement on the SAR image in each segmented area, the problem of synchronous deterioration of the target false alarm probability and the detection probability caused by the low signal-to-clutter ratio of the vehicle target under the background of strong ground clutter is enhanced. The target enhancement method includes linear, piecewise linear or nonlinear contrast stretching functions commonly used in the prior art.

[0044] Assume that a segmented area The histogram distribution of the internal SAR image amplitude value is: ,in ,in and Respectively represent the minimum and maximum amplitude values ​​of all resolution units in the segmented area. For this purpose, the classic piecewise linear function can be used for contrast stretching, which can enhance the detection probability of potential low signal-to-clutter ratio targets while minimizing the false alarm probability.

[0045] The piecewise linear function is as follows:

[0046]

[0047] in Indicates the segmented area A resolution unit The SAR image amplitude value, Represents the resolution unit after piecewise linear transformation SAR image amplitude value. Respectively represent the two gray threshold values ​​corresponding to the piecewise linear transformation, and there is , It is used to control the two gray value inflection points after contrast enhancement by piecewise linear function transformation; and there exists By performing piecewise linear function on the SAR image of the segmented area to complete contrast stretching, the signal-to-clutter ratio of the vehicle target under the background of strong ground clutter is improved.

[0048] In the present embodiment, the method for vehicle target detection is to adopt a constant false alarm processing method based on a statistical model. Usually, the number of resolution units of the target in each SAR image segmentation area accounts for a small proportion, and their probability distribution is hardly changed. Therefore, it can be reasonably assumed that the amplitude value of the resolution unit where the target is located has little effect on the histogram distribution of all pixel amplitude values ​​in the segmentation area. Therefore, the SAR image after the target enhancement in each segmentation area obtained in step 3 is used to perform target detection using a constant false alarm processing method based on a statistical model.

[0049] Step 4: Fuse the vehicles detected by the SAR image targets in all segmented areas and give the vehicle target detection results.

[0050] In this embodiment, the method for fusing the vehicles detected by the SAR image targets in all segmented areas is:

[0051] For each resolution unit in the SAR image, except for the edge resolution unit of the entire SAR image, the detection results of each resolution unit in different coverage segmentation frames are comprehensively detected and judged using decision fusion operators such as majority voting. shared by The segmentation boxes cover each segmentation box. When the detection conclusions of the resolution unit corresponding to each segmentation box are The segmentation box is judged to contain the target and , then the resolution unit in the original SAR image can be determined For the edge resolution unit of the entire SAR image, the detection result is directly output without fusion.

[0052] The multiple overlapping segmentation method is adopted so that each resolution unit in the whole SAR image, except for the edge resolution unit of the whole SAR image, consists of at least 2~ Therefore, except for the edgemost resolution unit of the entire SAR image, the detection results of each resolution unit in different coverage segmentation frames are comprehensively detected and judged using decision fusion operators such as majority voting, which further reduces the detection unreliability caused by the non-uniformity of the image after regional segmentation and improves the robustness of the vehicle target detection method.

[0053] The present invention also proposes a vehicle target detection device based on SAR images, comprising the following modules:

[0054] Image acquisition module: used to acquire SAR images;

[0055] Segmentation module: used for segmenting the SAR image in a manner of multiple overlapping segmentation frames;

[0056] Target enhancement module: used to enhance the target of the SAR image in each segmented area, where the segmented area refers to the area segmented by each segmentation box;

[0057] Target detection module: used to detect vehicle targets in each segmented area of ​​the SAR image after target enhancement;

[0058] Detection output module: used to fuse the vehicles detected by SAR image targets in all segmented areas and give vehicle target detection results.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A vehicle target detection method based on SAR images, characterized in that: The following steps are involved: Step 1: Acquire SAR images; Step 2: segmenting the SAR image in a manner of multiple overlapping segmentation frames; Step 3: Perform vehicle target detection on the SAR image within each segmented area, where the segmented area refers to the area segmented by each segmentation box; Step 4: Fuse all vehicles detected by SAR image targets in all segmented areas and give the vehicle target detection result; The method for fusing the vehicles detected by the SAR image targets in all segmented areas is as follows: for each resolution unit in the SAR image, except for the edge resolution unit of the entire SAR image, the detection results of each resolution unit in different coverage segmentation frames are comprehensively detected and judged by using the majority voting decision fusion operator; For the edge resolution unit of the entire SAR image, the detection result is directly output without fusion; The multiple overlap means that each segmentation frame sequentially overlaps and segments the SAR image in two dimensions, radial and lateral, according to a certain overlap ratio.

2. The vehicle target detection method according to claim 1, characterized in that: The overlap ratio refers to the ratio of the overlapped portion of the next segmentation frame and the current segmentation frame to the current segmentation frame.

3. The vehicle target detection method according to claim 2, characterized in that: The size of the segmentation frame for segmenting the SAR image is configured according to the maximum vertical and horizontal dimensions of the vehicle target to be detected. The specific size of the segmentation frame is ， in Indicates the multiple of the segmentation box size compared to the maximum size of the vehicle target to be detected. Indicates the magnification of the number of radial resolution units of the SAR image occupied by the vehicle target to be detected, Indicates the magnification of the number of lateral resolution units of the SAR image occupied by the vehicle target to be detected, and They respectively represent the number of radial and lateral resolution units occupied by the vehicle target to be detected in the SAR image.

4. The vehicle target detection method according to claim 3, characterized in that: Before performing vehicle target detection on the SAR image in each segmented area in step 3, target enhancement is performed on the SAR image in each segmented area, and then vehicle target detection is performed on the SAR image in each segmented area after target enhancement.

5. The vehicle target detection method according to claim 4, characterized in that: In step 3, target enhancement of the SAR image in each segmented area refers to using a target enhancement method to perform target enhancement according to the histogram distribution of the SAR image amplitude values ​​in each segmented area.

6. The vehicle target detection method according to claim 5, characterized in that: The method for vehicle target detection is to adopt a constant false alarm processing method based on a statistical model.

7. A vehicle target detection device based on SAR images, characterized in that: Includes the following modules: Image acquisition module: used to acquire SAR images; Segmentation module: used for segmenting the SAR image in a manner of multiple overlapping segmentation frames; the multiple overlapping means that each segmentation frame sequentially overlaps and segments the SAR image in radial and lateral dimensions according to a certain overlapping ratio; Target enhancement module: used to perform target enhancement on the SAR image within each segmented area, where the segmented area refers to the area segmented by each segmentation frame; Target detection module: used to detect vehicle targets in each segmented area of ​​the SAR image after target enhancement; Detection output module: used to fuse the vehicles detected by the SAR image targets in all segmented areas and give the vehicle target detection results. The method for fusing the vehicles detected by the SAR image targets in all segmented areas is: For each resolution unit in the SAR image, except for the edge resolution unit of the entire SAR image, the detection results of each resolution unit in different coverage segmentation frames are comprehensively detected and judged using the majority voting decision fusion operator; For the edge resolution unit of the entire SAR image, the detection result is directly output without fusion.

Images