A high-mobility small target detection method and system based on event signal

Abstract

The present invention discloses a method and system for detecting highly maneuverable small targets based on event signals. The method comprises: obtaining a first monitoring image of a moving target output by an event camera and a second monitoring image of a moving target output by a traditional camera; registering the first monitoring image using a turntable motion parameter to obtain a first registered image; eliminating the background information of the first registered image and highlighting the moving target using a moving target enhancement method based on multi-frame fusion to obtain a first enhanced image; locating the position of the moving target in the first enhanced image; calculating the position of the moving target in the second monitoring image, and identifying the moving target in the second monitoring image using a preset classification algorithm, and outputting the position information and category of the moving target; and controlling the turntable to rotate according to the position information of the moving target. The present invention is suitable for detecting highly maneuverable small moving targets under low-altitude complex backgrounds, and has good detection, tracking and recognition capabilities.

Description

Technical Field

[0001] The present invention relates to the technical field of moving target detection, and in particular to a method and system for detecting small highly maneuverable targets based on event signals. Background Art

[0002] There is a demand for low-altitude small flying object detection in many fields, such as drone intrusion detection for security purposes, and bird activity detection at airports for bird strike prevention. Specifically, it is necessary to be able to detect targets in a timely manner, lock and track targets, and identify targets. In common detection scenarios, radar is first used to perform preliminary detection of large-scale targets, and then optoelectronic equipment is used for precise detection and confirmation.

[0003] Existing photoelectric detection technology mainly relies on traditional cameras. The detection frame rate of traditional cameras is generally 20-25FPS, which is not enough to lock and track highly maneuverable targets. High-speed cameras can increase the detection frame rate to thousands or even tens of thousands. However, a large amount of data poses a huge challenge to real-time signal processing.

[0004] In comparison, event cameras have shown technical advantages. First, the detection frame rate is high, which can reach millions. Second, the amount of data is small, and only the changed area generates event signals. However, event camera signal processing is difficult. On the one hand, it is because the target information is incomplete. Only the changed area generates event signals, which makes it difficult for the signal to fully describe the target. On the other hand, there is too much interference information. The low-altitude background will generate a large number of interference signals when the optoelectronic device rotates. Being able to effectively process event signals is still a difficulty in detecting highly maneuverable small targets. Summary of the invention

[0005] In view of the defects in the prior art, the present invention provides a high-maneuverability small target detection method and system based on event signals to improve the accuracy of high-maneuverability small target detection in low-altitude complex environments.

[0006] First aspect

[0007] The present invention provides a highly maneuverable small target detection method based on event signals, comprising:

[0008] Acquire a first monitoring image of a moving target output by an event camera and a second monitoring image of a moving target output by a traditional camera;

[0009] Acquiring the motion parameters of the turntable; the event camera and the traditional camera are respectively arranged on the turntable;

[0010] Controlling the turntable to rotate according to the motion parameters of the turntable;

[0011] Using the turntable motion parameters to register the first monitoring image to obtain a first registered image;

[0012] Using a moving target enhancement method based on multi-frame fusion to eliminate background information of the first registered image and highlight the moving target, to obtain a first enhanced image;

[0013] Locate the position of the moving target in the first enhanced image; calculate the position of the moving target in the second monitoring image according to the position of the moving target in the first enhanced image, the second monitoring image, and the relative position relationship between the event camera and the traditional camera, and identify the moving target in the second monitoring image using a preset classification algorithm, and output the position information and category of the moving target;

[0014] The turntable is controlled to rotate according to the position information of the moving target.

[0015] Preferably, the registering the first monitoring image using the turntable motion parameter to obtain a first registered image specifically includes:

[0016] Receiving turntable parameters sent by the turntable control unit, and calculating background optical flow using the turntable motion parameters;

[0017] The background optical flow is used to register the first monitoring image according to a preset registration method to obtain a first registered image for eliminating background signals.

[0018] Preferably, the method of using a moving target enhancement method based on multi-frame fusion to eliminate background information of the first registered image and highlight the moving target to obtain the first enhanced image specifically includes:

[0019] The preset inter-frame registration algorithm is used to register multiple frames of images, and the differential operation is introduced to preliminarily obtain the contour of the moving target;

[0020] The image containing the motion contour is further registered between frames using a preset registration algorithm, and a first enhanced image based on event signal registration and multi-frame fusion is output through a subtraction and weighted fusion strategy.

[0021] Preferably, the method further comprises:

[0022] The first monitoring image is pre-processed by densification using a local mean operator to obtain a pre-processed first monitoring image.

[0023] Preferably, locating the position of the moving target in the first enhanced image specifically includes:

[0024] The area of ​​the spots is detected in the first enhanced image, and the largest spot is selected as the moving target; the position of the largest spot is the position of the moving target.

[0025] Second aspect

[0026] The present invention provides a high-mobility small target detection system based on event signals, comprising an event camera, a traditional camera, a turntable and a computing module; the event camera is arranged at a first position of the turntable; the traditional camera is arranged at a second position of the turntable; the event camera and the traditional camera rotate with the rotation of the turntable;

[0027] The event camera is used to output a first monitoring image of the moving target;

[0028] The conventional camera is used to output a second monitoring image of the moving target;

[0029] The calculation module comprises:

[0030] A parameter acquisition unit, used for acquiring the motion parameters of the turntable;

[0031] A turntable control unit, used for controlling the turntable to rotate according to the motion parameters of the turntable;

[0032] An image registration unit, configured to receive the turntable parameters sent by the turntable control unit, and register the first monitoring image using the turntable motion parameters to obtain a first registered image;

[0033] An image enhancement unit, configured to eliminate background information of the first registered image and highlight the moving target by using a moving target enhancement method based on multi-frame fusion, so as to obtain a first enhanced image;

[0034] a target recognition unit, configured to locate the position of the moving target in the first enhanced image; and further configured to calculate the position of the moving target in the second monitoring image according to the position of the moving target in the first enhanced image, the second monitoring image, and the relative position relationship between the event camera and the traditional camera, and to recognize the moving target in the second monitoring image using a preset classification algorithm, and output the position information and category of the moving target;

[0035] The turntable control unit is further used to receive the position information and category of the moving target sent by the target recognition unit, and control the turntable to rotate according to the position information of the moving target.

[0036] The beneficial effects of the present invention are:

[0037] The high frame rate and low data redundancy of the event camera are used to detect and track small targets with high maneuverability, and guide the traditional camera to complete the recognition task of the target; and an event signal registration method based on the motion parameters of the turntable is introduced to eliminate background noise in complex environments; a moving target enhancement algorithm based on multi-frame fusion is also used to further eliminate background noise, highlight moving targets, and improve accuracy. The present invention uses event cameras and traditional cameras to work together to detect and track moving targets, which is particularly suitable for detecting small targets with high maneuverability under complex backgrounds at low altitudes, and has good detection, tracking and recognition capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0039] Figure 1 A schematic diagram of a process of detecting a highly maneuverable small target based on event signals according to an embodiment of the present invention;

[0040] Figure 2 The present invention is a schematic diagram of the structure of a high-maneuverability small target detection system based on event signals according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The following embodiments of the technical solution of the present invention are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.

[0042] It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should have the common meanings understood by those skilled in the art to which the present invention belongs.

[0043] like Figure 1 As shown, the embodiment of the present invention also provides a high-maneuverability small target detection method based on event signals, comprising the following steps:

[0044] S1, acquiring a first monitoring image of a moving target output by an event camera and a second monitoring image of a moving target output by a traditional camera;

[0045] S2, obtaining the motion parameters of the turntable; the event camera and the traditional camera are respectively set on the turntable; the motion parameters of the turntable can be obtained by the suspicious target position information sent by the radar, and the coordinate system is converted to obtain the motion parameters of the turntable; when there is no radar signal, the turntable motion parameters pre-set in the system can be obtained;

[0046] S3, controlling the turntable to rotate according to the turntable motion parameters;

[0047] S4, registering the first monitoring image using the turntable motion parameter to obtain a first registered image;

[0048] S5, using a moving target enhancement method based on multi-frame fusion to eliminate background information of the first registered image and highlight the moving target, thereby obtaining a first enhanced image;

[0049] S6, locating the position of the moving target in the first enhanced image; calculating the position of the moving target in the second monitoring image according to the position of the moving target in the first enhanced image, the second monitoring image, and the relative position relationship between the event camera and the traditional camera, and identifying the moving target in the second monitoring image using a preset classification algorithm, and outputting the position information and category of the moving target;

[0050] S7, controlling the turntable to rotate according to the position information of the moving target, so that the event camera and the traditional camera continue to track the moving target.

[0051] Wherein, step S4 comprises:

[0052] Receive the turntable parameters sent by the turntable control unit, and use the turntable motion parameters to calculate the background optical flow. Specifically, during the process of the event camera rotating with the turntable, the low-altitude background produces pixel value changes in the imaging plane of the event camera, outputs complex background signals, and all background signals have the same optical flow field. Since the moving target has moved relative to the background, its optical flow field is different from the background. The embodiment of the present invention utilizes this feature, estimates the background optical flow using the turntable motion parameters, and aligns the preprocessed first monitoring image on this basis. Since the moving target has a relative displacement, the embodiment of the present invention can eliminate the background signal and highlight the moving target by performing a differential operation on the registered image. When the turntable rotates at a constant speed, the background optical flow in the horizontal and vertical directions can be expressed as:

[0053]

[0054]

[0055] Among them, (f x ,f y ) is the background optical flow, (w P ,w T ) is the angular velocity of the turntable in the horizontal and vertical directions, (s x ,s y ) is the pixel distance, Δt is the preset interval time, and in the embodiment of the present invention, Δt is set to 30ms.

[0056] The first monitoring image is registered using the background optical flow according to the preset registration method to obtain the first registered image, which is used to eliminate the background signal. The position of the background in the first monitoring image at different times can be restored to the original imaging position through the registration operation, while the moving target cannot return to the original imaging position through the registration operation. The registration process can be expressed as:

[0057]

[0058] Among them, I1 and I2 represent two images at different times in the image sequence, I'2 is the image after I2 is aligned to I1, and W and H are the width and height of the image respectively.

[0059] Step S5 includes:

[0060] The preset inter-frame registration algorithm is used to register multiple frames of images, and the differential operation is introduced to preliminarily obtain the contour of the moving target;

[0061] The preset registration algorithm is further used to perform inter-frame registration on the image containing the motion contour, and through a subtraction and weighted fusion strategy, a first enhanced image based on event signal registration and multi-frame fusion is output.

[0062] Since the event signal is sparse, the event signal positions generated by the same background in the two frames are different, which is manifested as misalignment or loss. The registration difference between the two frames is not enough to eliminate the background noise. In addition, the target motion distance between the two frames is small, the contour of the moving target is not obvious after the difference, and the detection is difficult. The embodiment of the present invention adopts an algorithm based on multi-frame fusion to enhance the moving target while introducing less noise. First, the adjacent frames in the image sequence are used to perform a registration difference operation to eliminate the background signal and obtain the contour of the moving target; at this time, the moving target in the image should contain 1 edge. Secondly, if the difference result is directly added, the contour of the moving target is enhanced, but the background noise is also enhanced. Therefore, the image obtained by the difference is further inter-frame registration, and the subtraction operation is performed on the registered image to enhance the edge of the moving target without enhancing the noise. In theory, the moving target in the image should contain 2 edges at this time, and the moving target is enhanced. Finally, since the background noise has been suppressed after the above operation, the weighted fusion of the above-mentioned enhanced image containing 2 edges can highlight and enhance the moving target. In the embodiment of the present invention, the weighting coefficient is set to 1.

[0063] Specifically, for a continuous K-frame event signal image {X n ,X n+1 ,…,X n+K-1}(The value of K is generally 3-4 depending on the turntable accuracy). First, the adjacent frames in the image sequence are used for registration difference operation to eliminate background information and obtain the contour of the moving target, as shown below:

[0064] Vi =|wrap(X i )-X i-1 |,n<i≤n+K-1

[0065] Among them, wrap(.) is the registration function, V i It is the difference image, i.e. the contour of the moving target;

[0066] Secondly, in order to enhance the moving target and eliminate the background noise in Vi, all Vi images are further aligned to the first frame in the Vi sequence, and then the first frame image in Vi is subtracted. The process can be expressed as:

[0067] Z i =|wrap(V i )-V n+1 |,n+1<i≤n+K-1

[0068] Among them, Zi is the image obtained by difference;

[0069] Finally, all Zi in the sequence are weighted and fused to obtain the first enhanced image. In the present invention, the weight coefficient is set to 1;

[0070] The weighting process is expressed as:

[0071] R n+K-1 =∑Z i .

[0072] In step 6, locating the position of the moving target in the first enhanced image specifically includes:

[0073] The area of ​​the spots is detected in the first enhanced image, and the largest spot is selected as the moving target; the position of the largest spot is the position of the moving target.

[0074] Specifically, in the enhanced frame R n+K-1 Detect the area of ​​the spots and select the largest spot as the moving target:

[0075] Obj=Argmax(Area blobs ), Area blobs >A

[0076] A is the screening threshold. Only spots with an area larger than A are calculated to reduce computational overhead. The location of the largest spot is (x1, y1, w1, h1), where (x1, y1) is the middle of the detection box, and w1 and h1 are the width and height of the detection box.

[0077] Further combined with the relative position relationship between the event camera and the traditional camera, the position (x2, y2, w2, h2) of the moving target in the second monitoring image is calculated.

[0078] When performing target recognition, firstly, w2 and h2 are sampled in the second monitoring image according to the ratio of 1, 1.1, and 1.2 to obtain the second target image block;

[0079] Secondly, the preset target classification algorithm is used to classify and identify the moving target on the second target image block, and the classification result with the highest confidence in the three scales is taken as the classification result for output to obtain the position information and category of the moving target.

[0080] Finally, the target recognition unit sends the position information and category of the moving target to the turntable control unit, and the turntable control unit is used to control the turntable rotation tracking according to the position information and category of the moving target.

[0081] A highly maneuverable small target detection method based on event signals provided by an embodiment of the present invention also includes: using a local mean operator to perform densification preprocessing on the first monitoring image to obtain the preprocessed first monitoring image. The embodiment of the present invention uses an event camera to directly output the event_in mode to output a frame of event image at an interval of 30ms. The image is sparse in both time and space, resulting in the texture edge of the projected image I also having sparse characteristics. Therefore, the embodiment of the present invention introduces a local mean operator to make the grayscale of the texture edge of the image more continuous and the pixels more dense:

[0082]

[0083] Where m and n are the sizes of the local mean operator, and k represents the mean operator. When m and n are 3, the effect is the best.

[0084] like Figure 2 As shown, based on the above-mentioned high-mobility small target detection method based on event signals, an embodiment of the present invention also provides a high-mobility small target detection system based on event signals, including an event camera, a traditional camera, a turntable and a computing module. The event camera is set at a first position of the turntable, and the traditional camera is set at a second position of the turntable. The event camera and the traditional camera rotate with the rotation of the turntable.

[0085] The event camera is used to output the first monitoring image of the moving target. The event camera is the core component of this system, responsible for monitoring the scene and continuously outputting asynchronous event signals. The traditional camera is used to output the second monitoring image of the moving target; the computing module includes: a preprocessing unit, a parameter acquisition unit, a turntable control unit, an image registration unit, an image enhancement unit, and a target recognition unit.

[0086] The preprocessing unit is used to preprocess the first monitoring image using a local mean operator to obtain a preprocessed first monitoring image. The embodiment of the present invention uses an event camera to directly output the event_in mode to output a frame of event image at an interval of 30ms. The image is sparse in both time and space, resulting in the texture edge of the projected image I also having sparse characteristics. Therefore, the embodiment of the present invention introduces a local mean operator to make the grayscale of the texture edge of the image more continuous and the pixels more dense:

[0087]

[0088] Where m and n are the sizes of the local mean operator, and k represents the mean operator. When m and n are 3, the effect is the best.

[0089] The parameter acquisition unit is used to obtain the turntable motion parameters. Specifically, the parameter acquisition unit can obtain the turntable motion parameters based on the suspicious target position information sent by the radar and perform coordinate system conversion. When there is no radar signal, the parameter acquisition unit can obtain the turntable motion parameters preset in the system and send them to the turntable control unit. The turntable control unit controls the turntable to rotate according to the preset turntable motion parameters.

[0090] The image registration unit is used to receive the turntable parameters sent by the turntable control unit, and to register the preprocessed first monitoring image using the turntable motion parameters to obtain a first registered image. In the process of the event camera rotating with the turntable, the low-altitude background produces pixel value changes in the imaging plane of the event camera, and outputs complex background signals, and all background signals have the same optical flow field. Since the moving target has moved relative to the background, its optical flow field is different from the background. The embodiment of the present invention utilizes this feature, estimates the background optical flow using the turntable motion parameters, and registers the preprocessed first monitoring image on this basis. Since the moving target has a relative displacement, the embodiment of the present invention can eliminate the background signal and highlight the moving target by performing a differential operation on the registered image. When the turntable rotates at a constant speed, the background optical flow in the horizontal and vertical directions can be expressed as:

[0091]

[0092]

[0093] Among them, (f x ,f y ) is the background optical flow, (w P ,w T ) is the angular velocity of the turntable in the horizontal and vertical directions, (s x ,s y ) is the pixel distance, Δt is the preset interval time, and in the embodiment of the present invention, Δt is set to 30ms.

[0094] The preprocessed first monitoring image is offset by using the background optical flow, and the registration is completed according to the preset registration method to obtain the first registered image. The position of the background in the first monitoring image at different times can be restored to the original imaging position through the registration operation, while the moving target cannot return to the original imaging position through the registration operation. The registration process can be expressed as:

[0095]

[0096] Among them, I1 and I2 represent two images at different times in the image sequence, I'2 is the image after I2 is aligned to I1, and W and H are the width and height of the image respectively.

[0097] The image enhancement unit is used to eliminate the background information of the first registered image and highlight the moving target by using a moving target enhancement method based on multi-frame fusion to obtain a first enhanced image. Since the event signal is sparse, the event signal positions generated by the same background in the two frames are different, which is manifested as misalignment or loss, and the registration difference between the two frames is not enough to eliminate the background noise. In addition, the target motion distance between the two frames is small, the contour of the moving target is not obvious after the difference, and the detection is difficult. The embodiment of the present invention adopts an algorithm based on multi-frame fusion to enhance the moving target while introducing less noise. First, the adjacent frames in the image sequence are used to perform a registration difference operation to eliminate the background signal and obtain the contour of the moving target; at this time, the moving target in the image should contain 1 edge. Secondly, if the difference results are directly added, the contour of the moving target is enhanced, but the background noise is also enhanced. Therefore, the image obtained by the difference is further inter-frame registered, and the registered image is subtracted to enhance the edge of the moving target without enhancing the noise. In theory, the moving target in the image should contain 2 edges at this time, and the moving target is enhanced. . Finally, since the background noise has been suppressed after the above operation, weighted fusion is performed on the enhanced image containing two edges to highlight and enhance the moving target. In the embodiment of the present invention, the weighted coefficient is set to 1.

[0098] Specifically, for a continuous K-frame event signal image {X n ,X n+1 ,…,X n+K-1}(The value of K is generally 3-4 depending on the turntable accuracy). First, the adjacent frames in the image sequence are used for registration difference operation to eliminate background information and obtain the contour of the moving target, as shown below:

[0099] V i =|wrap(X i )-X i-1 |,n<i≤n+K-1

[0100] Among them, wrap(.) is the registration function, V i It is the difference image, i.e. the contour of the moving target;

[0101] Secondly, to strengthen the sports goal and eliminate V i background noise, further all V i Image registration to V i The first frame in the sequence, then subtract V i The first frame image in . The process can be expressed as:

[0102] Z i =|wrap(V i )-V n+1 |,n+1<i≤n+K-1

[0103] Among them, Z i The image obtained by doing the difference;

[0104] Finally, for all Z in the sequence i Perform weighted fusion to obtain a first enhanced image, wherein the weighted coefficient is set to 1 in the present invention;

[0105] The weighting process is expressed as:

[0106] R n+K-1 =∑Z i .

[0107] The target recognition unit is used to locate the position of the moving target in the first enhanced image; it is also used to calculate the position of the moving target in the second monitoring image according to the position of the moving target in the first enhanced image, the second monitoring image, and the relative position relationship between the event camera and the traditional camera, and use a preset classification algorithm to identify the moving target in the second monitoring image, and output the position information and category of the moving target. n+K-1 The moving target is relatively prominent, and the embodiment of the present invention adopts a spot detection algorithm to complete target recognition and positioning.

[0108] Specifically, in the enhanced frame R n+K-1 Detect the area of ​​the spots and select the largest spot as the moving target:

[0109] Obj=Argmax(Area blobs ), Area blobs >A

[0110] A is the screening threshold. Only spots with an area larger than A are calculated to reduce computational overhead. The location of the largest spot is (x1, y1, w1, h1), where (x1, y1) is the middle of the detection box, and w1 and h1 are the width and height of the detection box.

[0111] Further combined with the relative position relationship between the event camera and the traditional camera, the position (x2, y2, w2, h2) of the moving target in the second monitoring image is calculated.

[0112] When performing target recognition, firstly, w2 and h2 are sampled in the second monitoring image according to the ratio of 1, 1.1, and 1.2 to obtain the second target image block;

[0113] Secondly, the preset target classification algorithm is used to classify and identify the moving target on the second target image block, and the classification result with the highest confidence in the three scales is taken as the classification result for output to obtain the position information and category of the moving target.

[0114] Finally, the target recognition unit sends the position information and category of the moving target to the turntable control unit, and the turntable control unit is used to control the turntable rotation tracking according to the position information and category of the moving target.

[0115] The embodiment of the present invention provides a method and system for detecting small high-maneuverable targets based on event signals. The method and system utilize the characteristics of high frame rate and low data redundancy of event cameras to detect and track small high-maneuverable moving targets, and guide traditional cameras to complete the recognition task of targets. The method also introduces an event signal registration method based on turntable motion parameters to eliminate background noise in complex environments. The method also utilizes a motion target enhancement algorithm based on multi-frame fusion to further eliminate background noise, highlight moving targets, and improve accuracy. The embodiment of the present invention utilizes event cameras to work in conjunction with traditional cameras to detect and track moving targets, and is particularly suitable for detecting small moving targets under complex backgrounds at low altitudes, and has good detection, tracking, and recognition capabilities.

[0116] 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 by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A highly maneuverable small target detection method based on event signals, characterized in that: include: Acquire a first monitoring image of a moving target output by an event camera and a second monitoring image of a moving target output by a traditional camera; Acquiring the motion parameters of the turntable; the event camera and the traditional camera are respectively arranged on the turntable; Controlling the turntable to rotate according to the motion parameters of the turntable; Using the turntable motion parameters to register the first monitoring image to obtain a first registered image; Using a moving target enhancement method based on multi-frame fusion to eliminate background information of the first registered image and highlight the moving target, to obtain a first enhanced image; Locating the position of the moving object in the first enhanced image; Calculate the position of the moving target in the second monitoring image according to the position of the moving target in the first enhanced image, the second monitoring image, and the relative position relationship between the event camera and the traditional camera, and identify the moving target in the second monitoring image using a preset classification algorithm, and output the position information and category of the moving target; The turntable is controlled to rotate according to the position information of the moving target.

2. The method for detecting a highly maneuverable small target based on event signals according to claim 1, characterized in that: The registering the first monitoring image using the turntable motion parameter to obtain the first registered image specifically includes: Receiving turntable parameters sent by the turntable control unit, and calculating background optical flow using the turntable motion parameters; The background optical flow is used to register the first monitoring image according to a preset registration method to obtain a first registered image for eliminating background signals.

3. The high-mobility small target detection method based on event signals according to claim 1 is characterized in that: The method of using a moving target enhancement method based on multi-frame fusion to eliminate background information of the first registered image and highlight the moving target to obtain a first enhanced image specifically includes: The preset inter-frame registration algorithm is used to register multiple frames of images, and the differential operation is introduced to preliminarily obtain the contour of the moving target; The preset registration algorithm is further used to perform inter-frame registration on the image containing the motion contour, and through a subtraction and weighted fusion strategy, a first enhanced image based on event signal registration and multi-frame fusion is output.

4. The method for detecting a highly maneuverable small target based on event signals according to claim 1, characterized in that: The method further comprises: The first monitoring image is pre-processed by densification using a local mean operator to obtain a pre-processed first monitoring image.

5. The method for detecting a highly maneuverable small target based on event signals according to claim 1, characterized in that: The step of locating the position of the moving target in the first enhanced image specifically includes: The area of ​​the spots is detected in the first enhanced image, and the largest spot is selected as the moving target; the position of the largest spot is the position of the moving target.

6. A highly maneuverable small target detection system based on event signals, characterized in that: The device comprises an event camera, a conventional camera, a turntable and a computing module; the event camera is arranged at a first position of the turntable; the conventional camera is arranged at a second position of the turntable; the event camera and the conventional camera rotate with the rotation of the turntable; The event camera is used to output a first monitoring image of the moving target; The conventional camera is used to output a second monitoring image of the moving target; The calculation module comprises: A parameter acquisition unit, used for acquiring the motion parameters of the turntable; A turntable control unit, used for controlling the turntable to rotate according to the motion parameters of the turntable; An image registration unit, configured to receive the turntable parameters sent by the turntable control unit, and register the first monitoring image using the turntable motion parameters to obtain a first registered image; An image enhancement unit, configured to eliminate background information of the first registered image and highlight the moving target by using a moving target enhancement method based on multi-frame fusion, so as to obtain a first enhanced image; a target recognition unit, configured to locate the position of the moving target in the first enhanced image; and further configured to calculate the position of the moving target in the second monitoring image according to the position of the moving target in the first enhanced image, the second monitoring image, and the relative position relationship between the event camera and the traditional camera, and to recognize the moving target in the second monitoring image using a preset classification algorithm, and output the position information and category of the moving target; The turntable control unit is further used to receive the position information and category of the moving target sent by the target recognition unit, and control the turntable to rotate according to the position information of the moving target.

Images