UAV projectile drop point collection system and method for test and evaluation

Abstract

A system and method for collecting the drop point of a UAV for test and identification belongs to the field of target positioning technology. In the system, the first rotor UAV and the second rotor UAV fly and rotate to a set position according to the command of the ground station and the obstacle avoidance strategy encountered during the flight, and then the first target detection device obtains the first image information including the drop point image, and transmits it to the ground station through the first data transmission module, and the second target detection device obtains the second image information including the drop point image, and transmits it to the ground station through the data transmission module; the ground station determines the location of the drop point according to the first hovering position of the first rotor UAV, the posture of the first target detection device, the second hovering position of the second rotor UAV, the posture of the second target detection device, the position of the drop point image in the first image, and the position of the drop point image in the second image through the binocular intersection measurement method. The present invention is not limited by the environment and is easy to operate and use.

Description

Technical Field

[0001] The present invention relates to a system and method for collecting the impact point of a drone missile for test and identification, and belongs to the technical field of artificial intelligence. Background Art

[0002] The Chinese invention patent application with publication number CN113126126A discloses a full-time automatic target reporting system and its ammunition landing point positioning method, the system includes: the system includes: GPS positioning equipment, a main control machine and a first measuring station; the main control machine is connected to the GPS positioning equipment, the main control machine includes a comprehensive evaluation unit, a dual-station positioning unit, a display and control unit, a main control network communication unit and a main control time control unit, the comprehensive evaluation unit interacts with the display and control unit, the main control time control unit interacts with the display and control unit, the dual-station positioning unit interacts with the display and control unit, and the main control network communication unit interacts with the display and control unit; the first measuring station includes a first electric The system further comprises a measuring station, which comprises an electric pan-tilt, a time control unit, a GPS unit, a pan-tilt control unit, a high-speed digital signal processing unit, a visible light camera, a first infrared camera, a power supply unit and a first network communication unit. The main control network communication unit interacts with the first pan-tilt control unit. The first pan-tilt control unit is respectively connected to the first electric pan-tilt, the first high-speed digital signal processing unit and the first power supply unit. The first high-speed digital signal processing unit is respectively connected to the first time control unit and the first network communication unit. The first time control unit is connected to the first GPS unit. The first high-speed digital signal processing unit is connected to the first visible light camera and the first infrared camera. The system further comprises a measuring station, which comprises an electric pan-tilt, a time control unit, a GPS unit, a pan-tilt control unit, a high-speed digital signal processing unit, a visible light camera, an infrared camera, a power supply unit and a network communication unit. The main control network communication unit interacts with the pan-tilt control unit. The pan-tilt control unit is respectively connected to the electric pan-tilt, the high-speed digital signal processing unit and the power supply unit. The high-speed digital signal processing unit is respectively connected to the time control unit and the network communication unit. The time control unit is connected to the GPS unit. The high-speed digital signal processing unit is connected to the visible light camera and the infrared camera.

[0003] However, the technical solution disclosed in the patent application has the following disadvantages: the setting of the first measuring station and the measuring station is subject to environmental restrictions: different terrain and environmental conditions, especially in harsh terrain environments such as mountainous areas and wetlands, make it difficult to find a suitable location for setting up equipment that meets the requirements. Summary of the invention

[0004] The purpose of the present invention is to provide a UAV bullet drop point acquisition system and method for test identification, which is not restricted by the environment, is easy to operate and use, and dynamically adjusts the segmentation parameters according to the local features of the image to achieve a more accurate image area of ​​the bullet drop point from the image.

[0005] To achieve the above-mentioned purpose, the present invention provides a UAV drop point collection system for test and identification, which includes a first rotor UAV, a second rotor UAV and a ground station, the first rotor UAV includes a first target detection device, a first processor, a first target detection device and a first data transmission module; the second rotor UAV includes a second target detection device, a second processor, a second target detection device and a second data transmission module; the first rotor UAV and the second rotor UAV respectively fly and rotate to a first position and a second position with a set interval apart according to the instructions of the ground station and the obstacle avoidance strategy adopted by the obstacles encountered during the flight, and then the first target detection device obtains first image information including the drop point image, the first processor processes the first image information and then transmits it to the ground station through the first data transmission module, and the second target detection device processes the first image information and then transmits it to the ground station through the first data transmission module. The detection device acquires second image information including an image of a bullet drop point, the second processor processes the second image information and then transmits it to the ground station through a second data transmission module; the ground station determines the position of the bullet drop point by a binocular intersection measurement method according to the first position of the first rotor UAV, the posture of the first target detection device, the second position of the second rotor UAV, the posture of the second target detection device, the position of the bullet drop point image in the first image, and the position of the bullet drop point image in the second image; the ground station includes a central processing unit, the central processing unit includes an image processing module, the image processing module includes a segmentation unit, a first calculation unit, a second calculation unit and a bullet drop point extraction unit, wherein the segmentation unit divides a frame of image acquired by the first target detection device and the second target detection device into Regions, , , where K is the number of pixels in a row and a column of the pixel window, C is the number of pixels in a row of a frame image, and D is the number of pixels in a column of a frame image. represents a floor function; the first calculation unit calculates the first value of the pixel window in the a-th row and the b-th column according to the following formula:

[0006] ,

[0007] In the formula, , The center pixel of the pixel window at row a and column b The value of Pixel The value of

[0008] The second calculation unit calculates the second value of the pixel window at the ath row and the bth column according to the following formula:

[0009] ;

[0010] Comparison of bullet drop point extraction units and ,when When , the pixel value in the pixel window of the ath row and the bth column is set to 1, otherwise it is set to 0, so that the one-frame image acquired by the first target detection device and the one-frame image acquired by the second target detection device are binarized, and further from the binarized images, the position of the bullet landing point image in the first image and the position of the bullet landing point image in the second image are obtained.

[0011] To achieve the above-mentioned purpose, the present invention also provides a method for collecting the drop point of drone bombs for test identification, which comprises the following steps:

[0012] Step 1: The ground station sends instructions to the first rotor UAV and the second rotor UAV, so that they fly and stop at the first position and the second position separated by a set distance, respectively, and keep the first target detection device and the second target detection device carried by them in a fixed posture, and then respectively obtain the first image information and the second image information including the image of the impact point, and then transmit them to the ground station through the first data transmission module and the second data transmission module respectively;

[0013] Step 2: The ground station uses an image processor to process the first image and the second image acquired by the first target detection device and the second target detection device respectively through a segmentation unit using a pixel window including K×K pixels to segment a frame of the image acquired by the first target detection device and the second target detection device into Regions, , , where K is the number of pixels in a row and a column of the pixel window, J is the number of pixels in a column of the pixel window, C is the number of pixels in a row of a frame image, and D is the number of pixels in a column of a frame image. represents the floor function;

[0014] Step 3: Calculate the first value of the pixel window at row a and column b according to the following formula:

[0015] ,

[0016] In the formula, , The center pixel of the pixel window at row a and column b The value of Pixel The value of

[0017] Step 4: Calculate the second value of the pixel window at row a and column b according to the following formula:

[0018] ;

[0019] Step 5: Compare and ,when When , the pixel value in the pixel window of the ath row and the bth column is set to 1, otherwise it is set to 0, so that a frame of image acquired by the first target detection device and the second target detection device is binarized, and further the position of the bullet landing point image in the first image and the position of the bullet landing point image in the second image are acquired from the binarized image;

[0020] Step 6: The ground station determines the position of the bullet drop point by binocular intersection measurement based on the first position of the first rotor UAV, the posture of the first target detection device, the second position of the second rotor UAV, the posture of the second target detection device, the position of the bullet drop point image in the first image, and the position of the bullet drop point image in the second image.

[0021] Compared with the prior art, the drone missile drop point acquisition system and method for test and identification provided by the present invention is not restricted by the environment, is easy to operate and use, and dynamically adjusts the segmentation parameters according to the local features of the image to achieve more accurate extraction of the drop point image from the image. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a block diagram of the composition of the drone bomb drop point collection system for test and identification provided by the first embodiment of the present invention.

[0023] Figure 2 It is a block diagram of the composition of any drone provided by the first embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0025] Figure 1 FIG. 1 is a block diagram of a drone bomb drop point acquisition system for test and identification provided by a first embodiment of the present invention. Figure 1 As shown, the first embodiment of the present invention provides a first rotor UAV, a second rotor UAV and a ground station, the first rotor UAV includes a first target detection device, a first positioning module, a first navigation system, a first processor, a first data transmission module, a first flight controller and a first flight mechanism, and the second rotor UAV includes a second target detection device, a second positioning module, a second navigation system, a second processor, a second data transmission module, a second flight controller and a second flight mechanism.

[0026] The first navigation system identifies the unit vector of the first rotor UAV's line of sight to an obstacle, the distance from the first rotor UAV to the obstacle, and the distance from the first rotor UAV to the target location based on data provided by the first target detection device, and outputs the data to the first processor; the first navigation system also monitors the position of the first rotor UAV, the distance to the obstacle, and the distance to the target location based on data provided by the first positioning module, and determines the unit vector of the first rotor UAV's line of sight to the obstacle, the rotation rate of the first rotor UAV's line of sight to the obstacle, the velocity vector of the first rotor UAV to the obstacle, and the velocity vector of the first rotor UAV to the target location based on the data provided by the first target detection device, and then provides them to the first processor.

[0027] The second navigation system identifies the unit vector of the second rotor UAV's line of sight to the obstacle, the distance from the second rotor UAV to the obstacle, and the distance from the second rotor UAV to the target location based on data provided by the second target detection device, and outputs them to the first processor; the first navigation system also monitors the position, distance to the obstacle, and distance to the target location of the second rotor UAV based on data provided by the second positioning module, and determines the unit vector of the second rotor UAV's line of sight to the obstacle, the rotation rate of the second rotor UAV's line of sight to the obstacle, and the speed vector of the second rotor UAV to the obstacle based on the data provided by the second target detection device, and then provides them to the second processor. The first rotor UAV and the second rotor UAV have different target locations, which are the first position and the second position, respectively.

[0028] The first rotor UAV and the second rotor UAV fly and stop at the set first position and second position according to the instructions of the ground station and the obstacle avoidance strategy encountered during the flight, and then the first target detection device obtains the first image information including the image of the landing point, and the first processor processes the first image information and transmits it to the ground station through the first data transmission module; the first positioning module is used to obtain the position information and time information of the first rotor UAV; the first processor also obtains the instructions of the ground station through the first data transmission module, and provides flight control instructions to the first flight controller according to the instructions generated by the first control network, and the first flight mechanism flies according to the instructions provided by the first flight controller, and the instructions generated by the first control network include the received instructions and the obstacle avoidance strategy during the flight.

[0029] The second target detection device obtains second image information including an image of the bullet landing point, and the second processor processes the second image information and transmits it to the ground station through the second data transmission module; the second positioning module is used to obtain the position information of the second rotor UAV; the second processor also obtains instructions from the ground station through the second data transmission module, and provides flight control instructions to the second flight controller based on the instructions generated by the second control network. The second flight mechanism flies according to the instructions provided by the second flight controller, and the instructions generated by the second control network include received instructions and obstacle avoidance strategies during flight.

[0030] The ground station includes a central processor, a third data transmission module, a display, an input / output device, and a database. The third data transmission module is used to receive image information including the drop point sent by the first rotor UAV and the second rotor UAV. The central processor also sends instructions to the first rotor UAV and the second rotor UAV through the third data transmission module. The central processor also receives the first image information obtained by the first target detection device from the first rotor UAV through the second data transmission module, and receives the second image information obtained by the second target detection device from the second rotor UAV, and converts the first image information into a first image and then displays it on the display, and converts the second image information into a second image and then displays it on the display; the central processor of the ground station determines the location of the drop point by binocular intersection measurement according to the first position of the first rotor UAV, the posture of the first target detection device, the second position of the second rotor UAV, the posture of the second target detection device, the location of the drop point image in the first image, and the location of the drop point image in the second image. The input / output device includes a keyboard, a mouse, a printer, etc. The keyboard and the mouse are used to input instructions. The printer is used to print out the output content. The database includes map information, and users can retrieve geographical information from the map from the ground station to the drop point, such as plains, mountains, cities, forests, etc.

[0031] In the present invention, the target detection device includes a visible camera and an infrared camera.

[0032] In the first embodiment, the central processing unit includes an image processing module, which includes a segmentation unit, a first calculation unit, a second calculation unit and a bullet drop point extraction unit, wherein the segmentation unit divides a frame of image acquired by the first target detection device and a frame of image acquired by the second target detection device into Regions, , , where K is the number of pixels in a row and a column of the pixel window, C is the number of pixels in a row of a frame image, and D is the number of pixels in a column of a frame image. represents a floor function; the first calculation unit calculates the first value of the pixel window in the a-th row and the b-th column according to the following formula:

[0033] ,

[0034] In the formula, , The center pixel of the pixel window at row a and column b The value of Pixel The value of

[0035] The second calculation unit calculates the second value of the pixel window at the ath row and the bth column according to the following formula:

[0036] ;

[0037] Comparison of bullet drop point extraction units and ,when When , the pixel value in the pixel window of the ath row and the bth column is set to 1, otherwise it is set to 0, so that the one-frame image acquired by the first target detection device and the one-frame image acquired by the second target detection device are binarized, and further from the binarized images, the position of the bullet landing point image in the first image and the position of the bullet landing point image in the second image are obtained.

[0038] In the first implementation, the first UAV further includes a first flight controller and a first flight mechanism, the first processor provides the flight instruction provided by the control network to the first flight controller, and the first flight controller generates an instruction to control the flight of the first flight mechanism according to the flight instruction. The second UAV further includes a second flight controller and a second flight mechanism, the second processor provides the flight instruction provided by the control network to the second flight controller, and the second flight controller generates an instruction to control the flight of the second flight mechanism according to the flight instruction.

[0039] The first embodiment of the present invention uses the above technical solution to dynamically adjust the segmentation parameters according to the local features of the image, so as to more accurately extract the image of the landing point from the image.

[0040] Figure 2 is a block diagram of any drone provided by the first embodiment of the present invention, such as Figure 2 As shown, the first processor and the second processor both include a control network , the first reward function and the second reward function set , , is the flight instruction of the UAV at time k, is the state of the UAV at the target location at time k, , To control the parameters of the network, is the velocity vector of the UAV relative to the target at time k, is the distance from the drone to the destination at time k The distance from the drone to the destination at time k+1 difference; is the state of the drone at time k with respect to J obstacles , , is the sight unit vector of the UAV to the jth obstacle at time k, is the rotation rate of the UAV’s line of sight to the jth obstacle at time k, The velocity vector of the drone relative to the jth obstacle at time k, is the distance from the UAV to the jth obstacle at time k; is the parameter of the first reward function, is the parameter set of the J second reward function, , where the drone follows the flight instructions provided by the control network The first reward function is used to evaluate the state of the UAV with respect to the target at time k and give a first reward, and the J second reward functions in the second reward function set are respectively used to evaluate the state of the UAV with respect to J obstacles at time k and give a second reward respectively, thereby forming a second reward data set; the control network updates its own parameters according to the first reward and the second reward data set.

[0041] In the first embodiment, With J The synthetic loss is the running speed vector of the UAV. In the present invention, the obstacles also include: birds encountered by the UAV during flight, other UAVs, etc.

[0042] In the first embodiment, the training of the drone includes the following process:

[0043] S1-1: Make , ;

[0044] S1-2: , T is the number of training times; ;

[0045] S1-3: Get the state of the drone relative to the target at time k and the states of J obstacles ;

[0046] S1-4: Get the flight instructions of the drone according to the following formula :

[0047] ;

[0048] S1-5: According to flight instructions Estimate the state of the drone at the target location at the next time k+1 and the states of J obstacles , and get the first reward and the second reward dataset , ;

[0049] S1-6: , ;

[0050] S1-7: Judgement, if If it is zero, execute S1-8, if not, return to step S1-4;

[0051] S1-8: Make the first cumulative return , ; Otherwise, make the first cumulative return , so that the second cumulative return ;

[0052] S1-9: ;

[0053] S1-10: If , then execute step S1-12, otherwise execute step S1-11;

[0054] S1-11: ; ;

[0055] ;

[0056] ;

[0057] ,

[0058] Then return to step S1-10, and is the learning rate;

[0059] S1-12: , , , and judge Is it true? If yes, end the training. If not, return to step S1-2. is the maximum number of training times.

[0060] In the first embodiment, the distance from the drone to the target at time k is The distance from the drone to the destination at time k+1 difference When is positive, is a positive value, otherwise it is zero; the distance between the drone and the jth obstacle When it is greater than the second threshold, is a positive value, otherwise it is zero.

[0061] The first embodiment of the present invention is achieved through the above-mentioned technical solution: the first rotor UAV and the second rotor UAV can autonomously avoid multiple obstacles, thereby flying to the first position and the second position respectively, and obtaining the first image and the second image including the image of the bullet landing point through the first target detection device and the second target detection device respectively, and the obstacles are such as buildings, trees, etc.

[0062] Second embodiment

[0063] The second embodiment of the present invention only describes the contents that are different from the first embodiment, and the same contents are not repeated.

[0064] The second embodiment of the present invention provides a method for collecting the impact point of a drone bomb for test identification, which comprises the following steps:

[0065] Step 1: The ground station sends instructions to the first rotor UAV and the second rotor UAV, so that they fly and stop at the first position and the second position separated by a set distance, respectively, and keep the first target detection device and the second target detection device carried by them in a fixed posture, and then respectively obtain the first image information and the second image information including the image of the impact point, and then transmit them to the ground station through the first data transmission module and the second data transmission module respectively;

[0066] Step 2: The ground station uses an image processor to process the first image and the second image acquired by the first target detection device and the second target detection device respectively through a segmentation unit to segment a frame of image acquired by the first target detection device and a frame of image acquired by the second target detection device into Regions, , , where K is the number of pixels in a row and a column of the pixel window, J is the number of pixels in a column of the pixel window, C is the number of pixels in a row of a frame image, and D is the number of pixels in a column of a frame image. represents the floor function;

[0067] Step 3: Calculate the first value of the pixel window at row a and column b according to the following formula:

[0068] ,

[0069] In the formula, , The center pixel of the pixel window at row a and column b The value of Pixel The value of

[0070] Step 4: Calculate the second value of the pixel window at row a and column b according to the following formula:

[0071] ;

[0072] Step 5: Compare and ,when When , the pixel value in the pixel window of the ath row and the bth column is set to 1, otherwise it is set to 0, so that a frame of image acquired by the first target detection device and the second target detection device is binarized, and further the position of the bullet landing point image in the first image and the position of the bullet landing point image in the second image are acquired from the binarized image;

[0073] Step 6: The ground station determines the position of the bullet drop point by binocular intersection measurement based on the first position of the first rotor UAV, the posture of the first target detection device, the first position of the second rotor UAV, the posture of the second target detection device, the position of the bullet drop point image in the first image, and the position of the bullet drop point image in the second image.

[0074] The method for collecting the drop point of a drone bomb for test identification provided by the second embodiment of the present invention also includes the following steps:

[0075] Step 7: Drone Based on Control Network Provide flight instructions to fly, is the flight instruction of the UAV at time k, is the state of the UAV at the target location at time k, , To control the parameters of the network, is the velocity vector of the UAV relative to the target at time k, is the distance between the UAV and the target at time k The distance between the drone and the target at time k+1 difference; , , is the sight unit vector of the UAV to the jth obstacle at time k, is the rotation rate of the UAV’s line of sight to the jth obstacle at time k, The velocity vector of the drone relative to the jth obstacle at time k, is the distance between the UAV and the jth obstacle at time k;

[0076] Step 8: Pass the first reward function Evaluate the state of the target location at time k and give the first reward to the drone. is the parameter of the first reward function;

[0077] Step 9: Pass the second reward function set The J second reward functions in respectively evaluate the state of the drone to the J obstacles at time k and give the second reward respectively, thereby forming a second reward data set. , is the parameter set of the J second reward function, ;

[0078] Step 10: The control network updates its own parameters based on the first reward and second reward data sets.

[0079] The beneficial effects of the method for collecting the drop points of drone bombs for test and identification provided by the second embodiment of the present invention are the same as those of the first embodiment and will not be described again here.

[0080] Third embodiment

[0081] A third embodiment of the present invention provides a computer program product, which includes computer program codes. The computer program codes can be called by a processor to execute the method of the embodiment.

[0082] The beneficial effects of the computer program product provided by the third embodiment of the present invention are the same as those of the second embodiment, and will not be described again here.

[0083] In addition, the terms "first", "first", "second", "third", "fourth", "fif ...

[0084] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A drone impact point collection system for test and identification, characterized in that: The invention comprises a first rotor UAV, a second rotor UAV and a ground station, wherein the first rotor UAV comprises a first target detection device, a first processor and a first data transmission module; the second rotor UAV comprises a second target detection device, a second processor and a second data transmission module; the first rotor UAV and the second rotor UAV respectively fly and rotate to a first position and a second position separated by a set interval according to the command of the ground station and the obstacle avoidance strategy adopted by the obstacles encountered during the flight, and then the first target detection device obtains first image information including an image of a bullet landing point, the first processor processes the first image information and then transmits it to the ground station through the first data transmission module, the second target detection device obtains second image information including an image of a bullet landing point, and the second processor processes the second image information The information is processed and then transmitted to the ground station through the second data transmission module; the ground station determines the position of the bullet drop point according to the first position of the first rotor UAV, the posture of the first target detection device, the second position of the second rotor UAV, the posture of the second target detection device, the position of the bullet drop point image in the first image, and the position of the bullet drop point image in the second image through binocular intersection measurement; the ground station includes a central processing unit, the central processing unit includes an image processing module, the image processing module includes a segmentation unit, a first calculation unit, a second calculation unit and a bullet drop point extraction unit, wherein the segmentation unit divides a frame of image acquired by the first target detection device and a frame of image acquired by the second target detection device into Regions, , , where K is the number of pixels in a row and a column of the pixel window, C is the number of pixels in a row of a frame image, and D is the number of pixels in a column of a frame image. represents a floor function; the first calculation unit calculates the first value of the pixel window in the a-th row and the b-th column according to the following formula: , In the formula, , The center pixel of the pixel window at row a and column b The value of Pixel The value of The second calculation unit calculates the second value of the pixel window at the ath row and the bth column according to the following formula: ； Comparison of bullet drop point extraction units and ,when When , the pixel value in the pixel window of the ath row and the bth column is set to 1, otherwise it is set to 0, so that the one-frame image acquired by the first target detection device and the one-frame image acquired by the second target detection device are binarized, and further from the binarized images, the position of the bullet landing point image in the first image and the position of the bullet landing point image in the second image are obtained.

2. The drone impact point collection system for test and identification according to claim 1 is characterized in that: The first processor and the second processor each include a control network , the first reward function and the second reward function set , , is the flight instruction of the UAV at time k, is the state of the UAV at the target location at time k, , To control the parameters of the network, is the velocity vector of the UAV relative to the target at time k, is the distance from the drone to the destination at time k The distance from the drone to the destination at time k+1 difference; is the state of the drone at time k with respect to J obstacles , , is the sight unit vector of the UAV to the jth obstacle at time k, is the rotation rate of the UAV’s line of sight to the jth obstacle at time k, The velocity vector of the drone relative to the jth obstacle at time k, is the distance from the UAV to the jth obstacle at time k; is the parameter of the first reward function, is the parameter set of the J second reward function, , where the drone follows the flight instructions provided by the control network The first reward function is used to evaluate the state of the UAV with respect to the target at time k and give a first reward, and the J second reward functions in the second reward function set are respectively used to evaluate the state of the UAV with respect to J obstacles at time k and give a second reward respectively, thereby forming a second reward data set; the control network updates its own parameters according to the first reward and the second reward data set.

3. The drone impact point collection system for test and identification according to claim 2 is characterized in that: The training of drones includes the following process: S1-1: Make , ; S1-2: , T is the number of training times; ; S1-3: Get the state of the drone relative to the target at time k and the states of J obstacles ; S1-4: Get the flight instructions of the drone according to the following formula : ； S1-5: According to flight instructions Estimate the state of the drone at the target location at the next time k+1 and the states of J obstacles , and get the first reward and the second reward dataset , ; S1-6： ， ; S1-7: Judgement, if If it is zero, execute S1-8, if not, return to step S1-4; S1-8: Make the first cumulative return , so that the second cumulative return ; Otherwise, make the first cumulative return , so that the second cumulative return ; S1-9： ; S1-10: If , then execute step S1-12, otherwise execute step S1-11; S1-11： ; ; ; , Then return to step S1-10, and is the learning rate; S1-12: , , , and judge Is it true? If yes, end the training. If not, return to step S1-2. is the maximum number of training times.

4. The drone impact point collection system for test and identification according to claim 3 is characterized in that: The distance from the drone to the destination at time k The distance from the drone to the destination at time k+1 difference When is positive, is a positive value, otherwise it is zero; the distance between the drone and the jth obstacle When it is greater than the second threshold, is a positive value, otherwise it is zero.

5. A method for collecting the impact point of drone bombs for test and identification, characterized in that: The steps include: Step 1: The ground station sends instructions to the first rotor UAV and the second rotor UAV, so that they fly and stop at the first position and the second position separated by a set distance, respectively, and keep the first target detection device and the second target detection device respectively carried by the first rotor UAV and the second rotor UAV in a fixed posture, and then respectively obtain the first image information and the second image information including the image of the impact point, and then transmit them to the ground station through the first data transmission module and the second data transmission module respectively; Step 2: The ground station uses an image processor to process the first image acquired by the first target detection device and the first image acquired by the second target detection device through a segmentation unit, using a pixel window including K×K pixels to segment a frame of the image acquired by the first target detection device and the second target detection device into Regions, , , where K is the number of pixels in a row and a column of the pixel window, J is the number of pixels in a column of the pixel window, C is the number of pixels in a row of a frame image, and D is the number of pixels in a column of a frame image. represents the floor function; Step 3: Calculate the first value of the pixel window at row a and column b according to the following formula: , In the formula, , The center pixel of the pixel window at row a and column b The value of Pixel The value of Step 4: Calculate the second value of the pixel window at row a and column b according to the following formula: ； Step 5: Compare and ,when When , the pixel value in the pixel window of the ath row and the bth column is set to 1, otherwise it is set to 0, so that a frame of image acquired by the first target detection device and the second target detection device is binarized, and further the position of the bullet landing point image in the first image and the position of the bullet landing point image in the second image are acquired from the binarized image; Step 6: The ground station determines the position of the bullet drop point by binocular intersection measurement based on the first position of the first rotor UAV, the posture of the first target detection device, the second position of the second rotor UAV, the posture of the second target detection device, the position of the bullet drop point image in the first image, and the position of the bullet drop point image in the second image.

6. The method for collecting the impact point of a drone bomb for testing and identification according to claim 5 is characterized in that: Also includes: Step 7: Drone Based on Control Network Provide flight instructions to fly, is the flight instruction of the UAV at time k, is the state of the UAV at the target location at time k, , To control the parameters of the network, is the velocity vector of the UAV relative to the target at time k, is the distance between the UAV and the target at time k The distance between the drone and the target at time k+1 difference; , , is the sight unit vector of the UAV to the jth obstacle at time k, is the rotation rate of the UAV’s line of sight to the jth obstacle at time k, The velocity vector of the drone relative to the jth obstacle at time k, is the distance between the UAV and the jth obstacle at time k; Step 8: Pass the first reward function Evaluate the state of the target location at time k and give the first reward to the drone. is the parameter of the first reward function; Step 9: Pass the second reward function set The J second reward functions in respectively evaluate the state of the drone to the J obstacles at time k and give the second reward respectively, thereby forming a second reward data set. , is the parameter set of the J second reward function, ; Step 10: The control network updates its own parameters based on the first reward and second reward data sets.

7. The method for collecting the impact point of a drone bomb for testing and identification according to claim 6 is characterized in that: The training of drones includes the following process: S1-1: Make , ; S1-2: , T is the number of training times; ; S1-3: Get the state of the drone relative to the target at time k and the states of J obstacles ; S1-4: Get the flight instructions of the drone according to the following formula : ； S1-5: According to flight instructions Estimate the state of the drone at the target location at the next time k+1 and the states of J obstacles , and get the first reward and the second reward dataset , ; S1-6： ， ; S1-7: Judgement, if If it is zero, execute S1-8, if not, return to step S1-4; S1-8: Make the first cumulative return , so that the second cumulative return ; Otherwise, make the first cumulative return , so that the second cumulative return ; S1-9： ; S1-10: If , then execute step S1-12, otherwise execute step S1-11; S1-11： ; ; ; , Then return to step S1-10, and is the learning rate; S1-12: , , , and judge Is it true? If yes, end the training. If not, return to step S1-2. is the maximum number of training times.

8. The method for collecting the impact point of a drone bomb for testing and identification according to claim 7 is characterized in that: The distance from the drone to the destination at time k The distance from the drone to the destination at time k+1 difference When is positive, is a positive value, otherwise it is zero; the distance between the drone and the jth obstacle When it is greater than the second threshold, is a positive value, otherwise it is zero.

Images