Target detection method and system based on Gm-APD range image and intensity image

Abstract

The invention provides a target detection method and system based on a Gm-APD range image and an intensity image, and belongs to the technical field of image recognition, and the method specifically comprises the steps: obtaining a range image and an intensity image, detected by a Gm-APD array, of a long-distance unmanned plane target; performing initial seed point selection in the range profile based on a preset seed point selection rule; setting a region growth rule, adding pixel points meeting the region growth rule in the range profile into a growth queue, carrying out region growth, and stopping the region growth when a growth stop condition is met; the area of a region after growth is calculated, and when the area of the region is within a preset range, the region is an unmanned aerial vehicle target region; and performing contour reconstruction of the unmanned aerial vehicle target based on the distance image and the intensity image corresponding to the unmanned aerial vehicle target area. According to the scheme, good boundary information and segmentation results can be reserved, the characteristics of a distance image and an intensity image are strictly fitted, parameters can be flexibly adjusted according to the image characteristics, and adaptive segmentation is carried out.

Description

Technical Field

[0001] The present application relates to the field of image recognition technology, and in particular to a target detection method and system based on Gm-APD range image and intensity image. Background Art

[0002] In recent years, with the increasing maturity of Unmanned Aerial Vehicle (UAV) technology, its application areas are also gradually expanding. For example, in the military field, it is used for aerial deployment to reduce the casualties of pilots; in the civil and commercial fields, it can be used for cargo transportation, emergency search and rescue, and as a relay system. However, the management of UAVs is still imperfect, and most of them are flying illegally, posing a serious threat to the privacy of citizens and the safety of specific places. How to effectively detect and warn such targets has become a hot issue. However, due to the small size of UAVs and the large search space, it is not easy to detect UAVs. Compared with daytime, the lack of lighting information at night makes this problem more difficult.

[0003] Traditional laser radar can detect high-speed moving targets at long distances, but it is helpless in detecting small targets at relatively close distances. The proposed laser radar system uses a Geiger-mode avalanche photodiode Gm-APD as a detector and photon counting technology, which can easily detect weak light signals. However, the distance image and intensity image detected by the Gm-APD array are very different from those of traditional radars: one is the imaging method, and the other is the noise formation form. There is a large amount of lost information and randomly distributed irregular distance anomaly noise in the distance image detected by the Gm-APD array. At the same time, the image resolution obtained by using a 64×64 Gm-APD array to detect small drone targets 5 kilometers away is too small, making it difficult to distinguish and restore many details.

[0004] At present, foreign countries started early in the processing technology of range image and intensity image detected by Gm-APD array, while China started later but has a higher enthusiasm, and has invested more in the improvement and optimization of algorithms, and the effect is constantly improving. For the existing range image and intensity image processing technology, while denoising, it will damage the target's detailed information. Therefore, finding a noise suppression algorithm that can effectively improve the noise suppression ability without damaging the ability to protect the target's detailed information is still an important issue for subsequent research. Summary of the invention

[0005] In view of this, an embodiment of the present application provides a target detection method and system based on Gm-APD range image and intensity image, which at least partially solves the problem that the existing range image and intensity image processing technology will damage the target's detailed information while performing denoising.

[0006] In a first aspect, an embodiment of the present application provides a target detection method based on Gm-APD range image and intensity image, the method comprising:

[0007] Obtain the range image and intensity image of the long-distance UAV target detected by the Gm-APD array;

[0008] Selecting an initial seed point in the range image based on a preset seed point selection rule, wherein the initial seed point serves as a critical point between the target area of ​​the drone and the noise point;

[0009] Setting a region growing rule, adding pixel points in the range image that meet the region growing rule to a growing queue, performing region growing, and stopping region growing when a growth stop condition is reached;

[0010] Calculate the area of ​​the grown region. If the area of ​​the region is within the preset range, the region is determined to be the target area of ​​the drone.

[0011] Based on the range image and the intensity image corresponding to the drone target area, the contour of the drone target is reconstructed.

[0012] According to a specific implementation of the embodiment of the present application, the preset seed point selection rule includes:

[0013] The pixel grayscale value gradient of every two pixels in the range image is calculated, and the pixel with the largest gradient amplitude is indexed as the initial seed point.

[0014] According to a specific implementation of the embodiment of the present application, the preset seed point selection rule also includes:

[0015] There are at least a preset number of pixels of the same nature around the initial seed point, and the pixel grayscale value of the initial seed point is not zero;

[0016] When the number of pixels of the same nature around the initial seed point does not meet the preset number requirement, or the pixel grayscale value of the initial seed point is zero, a pixel with the second largest index gradient amplitude is selected as the initial seed point;

[0017] This process is repeated until the best initial seed point is found.

[0018] According to a specific implementation of the embodiment of the present application, the preset number is set to 4.

[0019] According to a specific implementation of the embodiment of the present application, the region growing rule is set as:

[0020] The pixel points in the range image whose pixel grayscale value difference with the initial seed point is less than a preset threshold are added to the growth queue.

[0021] According to a specific implementation of the embodiment of the present application, the preset threshold is set to 10.

[0022] According to a specific implementation of the embodiment of the present application, when the growth stop condition is reached, stopping the regional growth includes:

[0023] Each pixel point in the growth queue is used as a seed. With each seed as the center, the pixels around each seed are visited, and the pixels that meet the regional growth rule are continuously added to the growth queue. At the same time, the visited seeds in the growth queue are marked until all the seeds in the growth queue are marked, and then the growth is stopped.

[0024] According to a specific implementation of the embodiment of the present application, the preset range is set to 4-500.

[0025] According to a specific implementation method of an embodiment of the present application, the method for selecting the initial seed point includes a human-computer interaction method and an algorithm autonomous method.

[0026] In a second aspect, an embodiment of the present application further provides a target detection system based on a Gm-APD range image and an intensity image, the system comprising:

[0027] An acquisition module is used to acquire the distance image and intensity image of the long-distance UAV target detected by the Gm-APD array;

[0028] A selection module, used to select an initial seed point in the range image based on a preset seed point selection rule, wherein the initial seed point serves as a critical point between the target area of ​​the drone and the noise point;

[0029] A region growing module, used to set a region growing rule, add the pixel points in the range image that meet the region growing rule to a growing queue, perform region growing, and stop region growing when a growth stop condition is reached;

[0030] A calculation module is used to calculate the area of ​​the grown region. When the area of ​​the region is within a preset range, the region is determined to be a target area of ​​the drone.

[0031] The contour reconstruction module is used to reconstruct the contour of the drone target based on the range image and the intensity image corresponding to the drone target area.

[0032] Beneficial effects:

[0033] The target detection method and system based on Gm-APD range image and intensity image in the embodiments of the present application process the range image and intensity image obtained by single-photon detection based on an improved region growing algorithm, which can retain good boundary information and segmentation results, and does not require prior image preprocessing or providing prior knowledge. It only needs to clarify the rules for seed point selection to perform segmentation, which effectively improves the noise suppression capability without compromising the ability to protect target detail information.

[0034] The improved region growing algorithm proposed in the present invention strictly conforms to the characteristics of the distance image and the intensity image, performs seed point selection and setting of growth rules, and can propose corresponding solutions according to actual needs to solve actual needs.

[0035] The present invention only processes the detected distance image and can simultaneously solve the problem of extracting corresponding data of the intensity image. The algorithm idea is simple and easy to understand, and parameters can be flexibly adjusted according to image characteristics to perform segmentation adaptively. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 is a flow chart of a target detection method based on Gm-APD range image and intensity image according to an embodiment of the present invention;

[0038] Figure 2 A typical structural diagram of a range image and an intensity image of a UAV according to an embodiment of the present invention;

[0039] Figure 3 A diagram showing a regional growth principle according to an embodiment of the present invention;

[0040] Figure 4 is a schematic diagram of a distance image and an intensity image of a high-rise building according to an embodiment of the present invention;

[0041] Figure 5 4 is a structural block diagram of a target detection system based on Gm-APD range image and intensity image according to an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0043] The following describes the implementation methods of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0044] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present application, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0045] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present application. The illustrations only show the components related to the present application rather than being drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed at will, and the component layout may also be more complicated.

[0046] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0047] In the first aspect, the present application embodiment provides a method for detecting a target based on a Gm-APD range image and an intensity image. Figures 1 to 5 Describe in detail.

[0048] This application uses the image characteristics of the long-distance UAV target range image and intensity image detected by the 64*64Gm-APD array to propose an improved region growing algorithm for target recognition. The typical structure of the 64*64 UAV range image and intensity image detected by the Gm-APD array is as follows: Figure 2 shown.

[0049] In one embodiment, referring to Figure 1 , based on the Gm-APD range image and intensity image target detection method, specifically including the following steps:

[0050] Step S101, obtaining the distance image and intensity image of the long-distance UAV target detected by the Gm-APD array;

[0051] Step S102, selecting an initial seed point in the range image based on a preset seed point selection rule, wherein the initial seed point serves as a critical point between the drone target area and the noise point;

[0052] Step S103, setting a region growing rule, adding the pixel points in the range image that meet the region growing rule to a growing queue, performing region growing, and stopping region growing when a growth stop condition is reached;

[0053] Step S104, calculating the area of ​​the grown region, and when the area of ​​the region is within a preset range, determining that the region is a target area of ​​the drone;

[0054] Step S105 : reconstructing the contour of the drone target based on the range image and the intensity image corresponding to the drone target area.

[0055] Region growing is a method that uses data characteristics as a judgment standard, compares the characteristics of any point with the surrounding points, merges data points with similar properties, and then realizes region recognition and segmentation. For each region, a seed point is first specified as the starting point for growth, and then the pixels in the area around the seed point are compared with the seed point, and the points with similar properties are merged and continue to grow outward until no pixels that meet the conditions are included. The principle diagram of region growing is as follows Figure 3 shown.

[0056] The current original region growing algorithm has problems such as low efficiency, over-segmentation and arbitrary seed point selection when performing region segmentation. Therefore, this application improves the region growing algorithm by optimizing the seed point selection rules.

[0057] The distance image of the long-distance UAV target detected by the Gm-APD array shows that the distance values ​​of the target area are close, not exceeding 10, and basically form a closed area. The distance value is quite different from the background noise, which can reach more than 100. The intensity difference between the background noise and the target area boundary on the intensity image detected is close, and it cannot be segmented by setting a threshold. At the same time, comparing the distance image and the intensity image only from the composition aspect, it can be found that the distance image and the intensity image formed by both the regional target and the background noise only differ in value, but the distance image and the intensity image formed at the same time have the same formation position.

[0058] Based on the above analysis of the range image and intensity image of general UAV, a region growing algorithm based on the optimization seed point selection rule is proposed.

[0059] In one embodiment, the preset seed point selection rule includes:

[0060] The pixel grayscale value gradient of every two pixels in the range image is calculated, and the pixel with the largest gradient amplitude is indexed as the initial seed point.

[0061] The preset seed point selection rule also includes:

[0062] There are at least a preset number of pixels of the same nature around the initial seed point, and the pixel grayscale value of the initial seed point is not zero;

[0063] When the number of pixels of the same nature around the initial seed point does not meet the preset number requirement, or the pixel grayscale value of the initial seed point is zero, a pixel with the second largest index gradient amplitude is selected as the initial seed point;

[0064] This process is repeated until the best initial seed point is found.

[0065] In specific implementation, the preset seed point selection process specifically includes the following steps:

[0066] a) According to the difference in pixel values ​​between the background noise and the target, the gradient of the 64*64 matrix is ​​first calculated, and the pixel with the largest gradient amplitude is indexed as the initial seed element. The initial seed point is intended to be the critical point between the target area and the noise point;

[0067] b) The initial seed element constructed based on the background noise and the target area must meet the following requirements: there are at least 4 pixels of the same nature around the initial seed element and the gray value of the initial seed pixel is not zero;

[0068] c) If the initial seed point does not meet the condition in b), the initial seed point is set to the pixel point where the second largest value of the matrix gradient amplitude is located;

[0069] Repeat steps b) and c) until the best initial seed point is found.

[0070] In one embodiment, the preset number is set to 4.

[0071] In one embodiment, the region growing rule is set as:

[0072] The pixel points in the range image whose pixel grayscale value difference with the initial seed point is less than a preset threshold are added to the growth queue.

[0073] Furthermore, the preset threshold is set to 10.

[0074] In the specific implementation, for the growth criterion: according to the analysis of a large amount of noisy range image data, the grayscale value difference between the target area and the background noise can reach more than 10, while the grayscale value difference inside the target area is small, so the judgment of the growable seeds in the growth process is set to: the difference in grayscale value of the pixel with the initial seed point, that is, if the grayscale difference is less than the threshold, the adjacent pixel is added to the growth queue.

[0075] In one embodiment, when the growth stop condition is reached, stopping the regional growth includes:

[0076] Each pixel point in the growth queue is used as a seed. With each seed as the center, the pixels around each seed are visited, and the pixels that meet the regional growth rule are continuously added to the growth queue. At the same time, the visited seeds in the growth queue are marked until all the seeds in the growth queue are marked, and then the growth is stopped.

[0077] In the specific implementation, for the growth stop condition: first traverse the pixel points around the initial seed point, add the pixel points that meet the growth criteria to the growth queue, then take each seed in the growth queue as the center, visit the surrounding pixel points, continue to add the seeds that meet the growth criteria to the growth queue, and mark the visited seeds in the growth queue at the same time, until all the seeds in the growth queue are marked, that is, stop growing.

[0078] The above data processing based on the improved region growing algorithm only performs preliminary denoising. The detected range image and intensity image may contain other targets, so the detection of drone targets still requires identification. In this application, a method based on a fixed threshold is used to determine the presence or absence of drone targets.

[0079] Consider that the segmented area may be a noise connected area and other non-UAV target areas, such as Figure 4 Shown are the range image and intensity image of the detected high-rise building.

[0080] In the specific implementation, the area of ​​the grown region is first calculated, and then the threshold of the area is set. A fixed threshold is set based on a large amount of experimental data. It is known that the pixels formed by noise are scattered and the area is small, while the pixels formed by high-rise buildings are much larger than the pixels formed by drones, generally reaching more than 100 pixels. Based on the above analysis of the area of ​​the target connected area, and for a 64*64 image, the maximum number of regional pixels is 4096. Therefore, the maximum area threshold of the drone target in the range image is set to no more than 500, and the minimum area threshold is not less than 4.

[0081] In one embodiment, the preset range is set to 4-500.

[0082] In one embodiment, the initial seed point selection method includes a human-computer interaction method and an algorithm autonomous method.

[0083] Regarding the selection of seed points, two methods are generally used: human-computer interaction and algorithm autonomy. However, in order to reduce the influence of subjective factors, algorithm autonomy is currently used more often.

[0084] In a second aspect, the present application also provides a target detection system based on Gm-APD range image and intensity image, referring to Figure 5 , the system comprising:

[0085] An acquisition module 501 is used to acquire a range image and an intensity image of a long-distance UAV target detected by a Gm-APD array;

[0086] A selection module 502 is used to select an initial seed point in the range image based on a preset seed point selection rule, and the initial seed point is used as a critical point between the drone target area and the noise point;

[0087] The region growing module 503 is used to set a region growing rule, add the pixel points in the range image that meet the region growing rule to the growing queue, perform region growing, and stop the region growing when the growth stop condition is reached;

[0088] The calculation module 504 is used to calculate the area of ​​the grown region. When the area of ​​the region is within a preset range, the region is determined to be a target area of ​​the drone.

[0089] The contour reconstruction module 505 is used to reconstruct the contour of the drone target based on the range image and the intensity image corresponding to the drone target area.

[0090] In one embodiment, the selection module 502 is further used to:

[0091] The pixel grayscale value gradient of every two pixels in the range image is calculated, and the pixel with the largest gradient amplitude is indexed as the initial seed point.

[0092] In one embodiment, the selection module 502 is further used to:

[0093] There are at least a preset number of pixels of the same nature around the initial seed point, and the pixel grayscale value of the initial seed point is not zero;

[0094] When the number of pixels of the same nature around the initial seed point does not meet the preset number requirement, or the pixel grayscale value of the initial seed point is zero, a pixel with the second largest index gradient amplitude is selected as the initial seed point;

[0095] This process is repeated until the best initial seed point is found.

[0096] In one embodiment, the selection module 502 is further used to: set the preset number to 4.

[0097] In one embodiment, the region growing module 503 is further used to:

[0098] The pixel points in the range image whose pixel grayscale value difference with the initial seed point is less than a preset threshold are added to the growth queue.

[0099] In one embodiment, the region growing module 503 is further configured to: set the preset threshold to 10.

[0100] In one embodiment, the region growing module 503 is further used to:

[0101] Each pixel point in the growth queue is used as a seed. With each seed as the center, the pixels around each seed are visited, and the pixels that meet the regional growth rule are continuously added to the growth queue. At the same time, the visited seeds in the growth queue are marked until all the seeds in the growth queue are marked, and then the growth is stopped.

[0102] In one embodiment, the calculation module 504 is further configured to: set the preset range to 4-500.

[0103] In one embodiment, the selection module 502 is further used for: the selection method of the initial seed point includes a human-computer interaction method and an algorithm autonomous method.

[0104] The embodiment provided by the present invention proposes a long-distance small target detection system to address the hazards brought by drones. A 64*64Gm-APD array detection is used to detect long-distance drone targets and locate drone targets.

[0105] The actual target range image and intensity image obtained by the 64*64Gm-APD array detection system are characterized by random distribution of background noise, low resolution of long-distance small target detection images, and less target detail information that is difficult to recover.

[0106] According to the analysis of the actual target range image and intensity image characteristics, it is proposed to process the original range image based on the improved region growing algorithm, setting the noise points as non-target parts instead of simply performing denoising. And by further judging the area of ​​the growing region, the detection of the presence or absence of the target is realized.

[0107] The present invention only preprocesses the actually collected single-frame target range image and intensity image, and detects the presence or absence of the target based on the preprocessing. This invention provides good sample data for subsequent joint target recognition between multiple frames.

[0108] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A target detection method based on Gm-APD range image and intensity image, characterized in that: The method comprises: Obtain the range image and intensity image of the long-distance UAV target detected by the Gm-APD array; Selecting an initial seed point in the range image based on a preset seed point selection rule, wherein the initial seed point serves as a critical point between the target area of ​​the drone and the noise point; Setting a region growing rule, adding pixel points in the range image that meet the region growing rule to a growing queue, performing region growing, and stopping region growing when a growth stop condition is reached; Calculate the area of ​​the grown region. If the area of ​​the region is within the preset range, the region is determined to be the target area of ​​the drone. Based on the range image and the intensity image corresponding to the drone target area, the contour of the drone target is reconstructed.

2. The target detection method based on Gm-APD range image and intensity image according to claim 1, characterized in that: The preset seed point selection rules include: The pixel grayscale value gradient of every two pixels in the range image is calculated, and the pixel with the largest gradient amplitude is indexed as the initial seed point.

3. The target detection method based on Gm-APD range image and intensity image according to claim 2, characterized in that: The preset seed point selection rule also includes: There are at least a preset number of pixels of the same nature around the initial seed point, and the pixel grayscale value of the initial seed point is not zero; When the number of pixels of the same nature around the initial seed point does not meet the preset number requirement, or the pixel grayscale value of the initial seed point is zero, a pixel with the second largest index gradient amplitude is selected as the initial seed point; This process is repeated until the best initial seed point is found.

4. The target detection method based on Gm-APD range image and intensity image according to claim 3 is characterized in that: The preset number is set to 4.

5. The target detection method based on Gm-APD range image and intensity image according to claim 1, characterized in that: The region growing rule is set as: The pixel points in the range image whose pixel grayscale value difference with the initial seed point is less than a preset threshold are added to the growth queue.

6. The target detection method based on Gm-APD range image and intensity image according to claim 5, characterized in that: The preset threshold is set to 10.

7. The target detection method based on Gm-APD range image and intensity image according to claim 1, characterized in that: When the growth stop condition is reached, stopping the regional growth includes: Each pixel point in the growth queue is used as a seed. With each seed as the center, the pixels around each seed are visited, and the pixels that meet the regional growth rule are continuously added to the growth queue. At the same time, the visited seeds in the growth queue are marked until all the seeds in the growth queue are marked, and then the growth is stopped.

8. The target detection method based on Gm-APD range image and intensity image according to claim 1, characterized in that: The preset range is set to 4-500.

9. The target detection method based on Gm-APD range image and intensity image according to claim 1, characterized in that: The initial seed point selection method includes a human-computer interaction method and an algorithm autonomous method.

10. A target detection system based on Gm-APD range image and intensity image, characterized in that: The system comprises: An acquisition module is used to acquire the distance image and intensity image of the long-distance UAV target detected by the Gm-APD array; A selection module, used to select an initial seed point in the range image based on a preset seed point selection rule, wherein the initial seed point serves as a critical point between the target area of ​​the drone and the noise point; A region growing module, used to set a region growing rule, add the pixel points in the range image that meet the region growing rule to a growing queue, perform region growing, and stop region growing when a growth stop condition is reached; A calculation module is used to calculate the area of ​​the grown region. When the area of ​​the region is within a preset range, the region is determined to be a target area of ​​the drone. The contour reconstruction module is used to reconstruct the contour of the drone target based on the range image and the intensity image corresponding to the drone target area.

Images