Video tracking control algorithm

Abstract

The invention provides a video tracking control algorithm, which comprises the following steps of: S1, after an embedded box algorithm program is started, pulling a video stream from a ball machine and identifying a video picture frame by frame; s2, after the recognition algorithm recognizes the target, the coordinates X and Y of the target with the center point of the screen as the original point and the height H and width W of the target are returned; and S3, the current field angle Az of the dome camera is read by the tracking algorithm, and the angle Ap that the dome camera should rotate in the horizontal direction when the target is in the center of the screen is calculated according to the target coordinate X. According to the video tracking control algorithm provided by the invention, through the algorithm deployed at the front-end equipment, the network delay can be eliminated, the video can be analyzed in real time, the identification distance is long, and the accuracy is high; the control response to the video equipment is fast; according to the algorithm, the pitch angle, the azimuth angle and the focal length of the video equipment are controlled through the size and the position of target recognition by the video equipment, so that target tracking is completed.

Description

Technical Field

[0001] The present invention relates to the field of video tracking technology, and in particular to a video tracking control algorithm. Background Art

[0002] The existing video tracking solutions include algorithms deployed in video devices, such as home smart cameras or conference cameras. Such video devices have target tracking functions and are generally used indoors to track and shoot people. They are characterized by tracking moving targets, a single background image, and little or no adjustment of the focal length. Security video dome cameras from manufacturers such as Hikvision also deploy video recognition and target tracking algorithms inside, which are triggered by smart events. After a regular area is drawn, the target is recognized and tracked after entering. According to actual test results, the recognition distance of such dome cameras with built-in recognition algorithms does not exceed 200 meters. When the target moves out of the recognition range, it will not be recognized and tracked, and the target will quickly move out of the video screen.

[0003] Another video tracking solution is to deploy an identification and tracking algorithm on a server in the central computer room. The server has powerful hardware performance, a strong algorithm, and good target recognition effects, and can identify distant targets. However, this solution often involves a long distance between the front-end video acquisition device and the central computer room, resulting in a large network delay. The video images captured by the video device are transmitted back to the server in the central computer room. After the server algorithm identifies the image, it issues a control command to the video device. At this time, the actual position of the target has deviated from the position of the target in the image. This solution has a poor tracking effect and a slow real-time response. It can only track distant and slow targets for a short period of time, while fast and close targets are easily lost.

[0004] According to the description of the closest existing technical solution, the tracking algorithm deployed in the video device has the following defects: 1. It can only be used indoors, does not need to adjust the focal length, has a single usage scenario, and has poor outdoor scene recognition effect; 2. The video device used outdoors has a short recognition distance. When the target moves out of the recognition range, the tracking of the target is interrupted. The tracking algorithm deployed in the server in the central computer room has the following defects: 1. The video stream is returned to the central computer room for target recognition, and the network delay is large. When the target recognition is completed, the position of the target has changed, which is not conducive to tracking by the tracking algorithm; 2. After the target recognition is completed, when the tracking algorithm controls the video device, when the video device adjusts the pitch angle, azimuth angle and focal length, there is a problem of slow response and easy frame loss; because of the above two defects, the tracking effect is that it can only track targets with slow angular velocity relative to the video device in the distance. In order to see the target features clearly, it is easy to lose the target when the focal length is enlarged.

[0005] Therefore, it is necessary to provide a video tracking control algorithm to solve the above technical problems. Summary of the invention

[0006] The present invention provides a video tracking control algorithm, which solves the problems of single usage scenario of indoor equipment and short recognition distance of outdoor video equipment, as well as large network delay, slow control response speed and easy loss of target of tracking algorithms deployed on servers in a central computer room.

[0007] In order to solve the above technical problems, the present invention provides a video tracking control algorithm, comprising the following steps:

[0008] S1. After the embedded box algorithm program is started, it pulls the video stream from the ball camera and recognizes the video frame by frame;

[0009] S2. After the recognition algorithm recognizes the target, it returns the coordinates X, Y of the target with the center of the screen as the origin and the height H and width W of the target;

[0010] S3, the tracking algorithm reads the current field of view angle Az of the ball camera, and calculates the angle Ap that the ball camera should rotate in the horizontal direction when the target is at the center of the screen according to the target coordinate X; and calculates the angle At that the ball camera should rotate in the pitch direction when the target is at the center of the screen according to the target coordinate Y;

[0011] S4, the tracking algorithm reads the current focal length value Z of the ball camera, and calculates the focal length Z' that needs to be adjusted according to the target size;

[0012] S5. Calculate the horizontal rotation speed Sp and the pitch rotation speed St according to Ap and At calculated above;

[0013] S6. The tracking algorithm controls the ball camera to rotate at a horizontal speed Sp and a pitch speed St, and reads the horizontal angle and pitch angle of the ball camera every 50ms to recalculate new Ap and At;

[0014] S7, read the algorithm recognition result every 300ms, and repeat the above calculation according to the new target result X, Y, W, H;

[0015] S8. Finally, the target is tracked at the center of the screen. As the target moves, the ball camera automatically follows the target and adjusts the focal length according to the distance of the target.

[0016] Preferably, the formula of the angle Ap is Ap=X*Az / 1920, and the formula of the angle At is At=Y*Az / 1080.

[0017] Preferably, when the target H>W (personnel target), Z'=1080 / 4*Z / H, and when the target W>H (car or ship target), Z'=1920 / 4*Z / W, and the focal length value of the ball camera is adjusted by Z'.

[0018] Preferably, the horizontal rotation speed Sp=Pp*Ap / 0.3, and the pitch rotation speed St=Pt*At / 0.3.

[0019] Preferably, the startup Pp and Pt are speed coefficients, which are related to the rotation speed of different models of ball cameras. It is necessary to obtain a time of 0.3 of 300ms based on actual scene tests.

[0020] Preferably, new Sp and St are calculated according to the new Ap and At, that is, the horizontal speed and pitch speed of the ball camera are adjusted every 50ms.

[0021] Preferably, a ball camera is required in step S1, and the ball camera includes a mounting plate, a fixing column is fixedly installed on the bottom of the mounting plate, a top plate is fixedly installed on the bottom of the outer side of the fixing column, a protective transparent cover is fixedly installed on the bottom of the top plate, a first driving motor is fixedly installed on the bottom of the fixing column, and a camera assembly is fixedly installed on the output end of the first driving motor.

[0022] Preferably, a mounting groove is provided on the top of the top plate, a second driving motor is fixedly mounted on the inner side surface of the mounting groove, and a driving gear is fixedly mounted on the output end of the second driving motor.

[0023] Preferably, the outer side surface of the fixed column is rotatably connected to a rotating circle, a groove is provided at the bottom of the inner side surface of the rotating circle, a gear ring is fixedly installed on the inner side surface of the groove, the gear ring is meshed with the outer side surface of the driving gear, a connecting frame is fixedly installed on one side of the rotating circle, a through groove is provided on one side of the connecting frame, a wiping cotton is provided on the other side of the connecting frame, and one side of the wiping cotton is attached to the side surface of the protective transparent cover.

[0024] Preferably, a second Velcro is fixedly installed on the other side of the connecting frame, a first Velcro is adhered to one side of the second Velcro, and the first Velcro is fixedly installed on the other side of the wiping cotton.

[0025] Compared with the related art, the video tracking control algorithm provided by the present invention has the following beneficial effects:

[0026] The present invention provides a video tracking control algorithm. By deploying the algorithm on the front-end device, the network delay can be eliminated, the video can be analyzed in real time, the recognition distance is long, and the accuracy is high; the control response of the video device is fast; the algorithm controls the pitch angle, azimuth angle and focal length of the video device through the size and position of the target recognized by the video device to complete the tracking of the target. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A flowchart of a first embodiment of a video tracking control algorithm provided by the present invention;

[0028] Figure 2 A schematic structural diagram of a second embodiment of a video tracking control algorithm provided by the present invention;

[0029] Figure 3 for Figure 1 A schematic diagram of the top plate cross-section structure shown;

[0030] Figure 4 for Figure 3 An enlarged schematic diagram of part A is shown;

[0031] Figure 5 A schematic structural diagram of a third embodiment of a video tracking control algorithm provided by the present invention.

[0032] Numbers in the figure: 1. Mounting plate, 11. Fixed column, 12. Top plate, 13. Mounting groove, 14. First drive motor, 15. Camera assembly, 16. Protective transparent cover, 2. Second drive motor, 21. Driving gear, 3. Rotating circle, 31. Groove, 32. Gear ring, 4. Connecting frame, 41. Through groove, 42. Wiping cotton, 5. First Velcro, 51. Second Velcro. DETAILED DESCRIPTION

[0033] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0034] First embodiment

[0035] Please refer to Figure 1 ,in, Figure 1 A flow chart of a first embodiment of a video tracking control algorithm provided by the present invention. A video tracking control algorithm comprises the following steps:

[0036] S1. After the embedded box algorithm program is started, it pulls the video stream from the ball camera and recognizes the video frame by frame;

[0037] S2. After the recognition algorithm recognizes the target, it returns the coordinates X, Y of the target with the center of the screen as the origin and the height H and width W of the target;

[0038] S3, the tracking algorithm reads the current field of view angle Az of the ball camera, and calculates the angle Ap that the ball camera should rotate in the horizontal direction when the target is at the center of the screen according to the target coordinate X; and calculates the angle At that the ball camera should rotate in the pitch direction when the target is at the center of the screen according to the target coordinate Y;

[0039] S4, the tracking algorithm reads the current focal length value Z of the ball camera, and calculates the focal length Z' that needs to be adjusted according to the target size;

[0040] S5. Calculate the horizontal rotation speed Sp and the pitch rotation speed St according to Ap and At calculated above;

[0041] S6. The tracking algorithm controls the ball camera to rotate at a horizontal speed Sp and a pitch speed St, and reads the horizontal angle and pitch angle of the ball camera every 50ms to recalculate new Ap and At;

[0042] S7, read the algorithm recognition result every 300ms, and repeat the above calculation according to the new target result X, Y, W, H;

[0043] S8. Finally, the target is tracked at the center of the screen. As the target moves, the ball camera automatically follows the target and adjusts the focal length according to the distance of the target.

[0044] The formula of the angle Ap is Ap=X*Az / 1920, and the formula of the angle At is At=Y*Az / 1080.

[0045] When the target H>W (personnel target), Z'=1080 / 4*Z / H, when the target W>H (car or ship target), Z'=1920 / 4*Z / W, and the focus value of the ball camera is adjusted by Z'.

[0046] The horizontal rotation speed Sp=Pp*Ap / 0.3, and the pitch rotation speed St=Pt*At / 0.3.

[0047] The startup Pp and Pt are speed coefficients, which are related to the rotation speed of different models of ball cameras. It is necessary to test the actual scene to obtain a time of 0.3 for 300ms.

[0048] The new Sp and St are calculated according to the new Ap and At, that is, the horizontal speed and pitch speed of the ball camera are adjusted every 50ms.

[0049] Using the video recognition algorithm deployed in the front-end embedded system, the size and position of people, cars, and ships in each frame of the video stream are identified. The size of the target on the screen is divided into the pixel height H of the target and the pixel width W of the target. The position of the target on the screen is divided into the horizontal coordinate X from the center point of the target to the center point of the screen and the vertical coordinate Y from the center point of the target to the center point of the screen. The target tracking algorithm calculates the rotation speed of the azimuth and pitch of the video device based on the difference between the target coordinates XY in the previous and next two frames, and calculates the focal length of the video device based on the difference between the target size HW in the previous and next two frames;

[0050] The physical connection basic structure of the present invention is composed of: a ball camera and an embedded algorithm box, and the algorithm box deploys video recognition algorithm and target tracking algorithm programs. After the algorithm box is started, the recognition algorithm starts to run, and the video images collected by the ball camera are recognized frame by frame. When a target enters the video image of the ball camera, the video recognition algorithm calculates the target recognition result, and the target tracking algorithm controls the ball camera according to the target recognition result, and adjusts the azimuth, pitch angle and field angle of the ball camera to track the target in real time;

[0051] The present invention adopts a video recognition and tracking algorithm deployed in the front algorithm box, and uses edge computing technology to locally process the data processing and control instructions of the edge video device, which greatly eliminates network delays, realizes millisecond-level completion of video picture processing, millimeter-level response of control instructions, and quick and sensitive control of the angle and focal length of the ball camera. The tracking algorithm can perform video tracking on targets of different distances and speeds;

[0052] The video recognition algorithm of the present invention is an algorithm specially developed for long-distance business scenarios. After long-term data optimization, it has a high recognition accuracy. The pixel requirement for personnel targets is 64*64, and the pixel requirement for vehicle and ship targets is 100*100. The effective recognition distance depends on the focal length of the video device. At present, a ball camera with a focal length of 192mm can complete the recognition of personnel at 1.2km and the recognition of vehicles and ships at 2km.

[0053] The working principle of a video tracking control algorithm provided by the present invention is as follows:

[0054] After the embedded box algorithm program is started, the video stream is pulled from the ball camera and the video screen is recognized frame by frame; after the recognition algorithm recognizes the target, it returns the target coordinates X, Y with the center of the screen as the origin and the height H and width W of the target; S3, the tracking algorithm reads the current field of view angle Az of the ball camera, and calculates the angle Ap that the ball camera should rotate horizontally when the target is in the center of the screen according to the target coordinate X, and the formula is Ap=X*Az / 1920; calculates the angle At that the ball camera should rotate in the pitch direction when the target is in the center of the screen according to the target coordinate Y, and the formula is At=Y*Az / 1080; the tracking algorithm reads the current focal length value Z of the ball camera, and calculates the focal length Z' that needs to be adjusted according to the size of the target. When the target H>W (personnel target), Z'=1080 / 4*Z / H, when the target W>H (car or ship target), Z'=1920 / 4*Z / W, and adjust the focal length value of the ball camera with Z'; according to The Ap and At calculated above are used to calculate the horizontal rotation speed Sp = Pp*Ap / 0.3 and the pitch rotation speed St = Pt*At / 0.3. Pp and Pt are started as speed coefficients, which are related to the rotation speed of different models of ball cameras. It is necessary to test the actual scene to determine that 0.3 is 300ms. The tracking algorithm controls the ball camera to rotate at the horizontal speed Sp and the pitch speed St. The horizontal angle and pitch angle of the ball camera are read every 50ms to recalculate the new Ap and At. The new Sp and St are calculated based on the new Ap and At, that is, the horizontal speed and pitch speed of the ball camera are adjusted every 50ms. The algorithm recognition result is read every 300ms, and the above calculation is repeated based on the new target results X, Y, W, H. Finally, the tracking target is achieved at the center of the screen. As the target moves, the ball camera automatically follows the target and automatically adjusts the focal length according to the distance of the target.

[0055] Compared with the related art, the video tracking control algorithm provided by the present invention has the following beneficial effects:

[0056] By deploying algorithms on front-end devices, network delays can be eliminated, real-time video analysis can be performed, long-distance recognition can be achieved, and accuracy can be increased. The control response to video equipment is fast. The algorithm uses the size and position of the target identified by the video equipment to control the pitch angle, azimuth angle, and focal length of the video equipment to complete target tracking.

[0057] Second embodiment

[0058] Please refer to Figure 2 , Figure 3 and Figure 4 Based on a video tracking control algorithm provided by the first embodiment of the present application, the second embodiment of the present application proposes another video tracking control algorithm. The second embodiment is only a preferred method of the first embodiment, and the implementation of the second embodiment will not affect the independent implementation of the first embodiment.

[0059] Specifically, the difference of a video tracking control algorithm provided by the second embodiment of the present application is that, in a video tracking control algorithm, a ball camera is required in step S1, and the ball camera includes a mounting plate, a fixing column is fixedly installed on the bottom of the mounting plate, a top plate is fixedly installed on the bottom of the outer side of the fixing column, a protective transparent cover is fixedly installed on the bottom of the top plate, a first drive motor is fixedly installed on the bottom of the fixing column, and a camera assembly is fixedly installed on the output end of the first drive motor.

[0060] A mounting groove is formed on the top of the top plate, a second driving motor is fixedly mounted on the inner side of the mounting groove, and a driving gear is fixedly mounted on the output end of the second driving motor.

[0061] The outer side surface of the fixed column is rotatably connected with a rotating circle, a groove is provided at the bottom of the inner side surface of the rotating circle, a gear ring is fixedly installed on the inner side surface of the groove, the gear ring is meshed with the outer side surface of the driving gear, a connecting frame is fixedly installed on one side of the rotating circle, a through groove is provided on one side of the connecting frame, a wiping cotton is provided on the other side of the connecting frame, and one side of the wiping cotton is attached to the side surface of the protective transparent cover.

[0062] The camera angle of the camera assembly 15 passes through the middle of the through slot 41, and the speed at which the first drive motor 14 drives the camera assembly 15 to rotate is the same as the speed at which the second drive motor 2 drives the connecting frame 4 to rotate, so that the camera angle of the camera assembly 15 is always in the middle of the through slot 41.

[0063] The working principle of a video tracking control algorithm provided by the present invention is as follows:

[0064] When in use, the rotation of the first drive motor 14 will drive the camera assembly 15 to rotate, and when the first drive motor 14 rotates, it will also drive the second drive motor 2 to rotate, so that the second drive motor 2 drives the driving gear 21 to rotate, and then the driving gear 21 drives the ring gear 32 to rotate, and then the ring gear 32 drives the rotating circle 3 to rotate on the outer side of the fixed column 11, and then drives the connecting frame 4 and the wiping cotton 42 to rotate, and then the wiping cotton 42 cleans the dust on the protective transparent cover 16.

[0065] Compared with the related art, the video tracking control algorithm provided by the present invention has the following beneficial effects:

[0066] By means of the cooperation of the second driving motor 2, the driving gear 21, the rotating circle 3, the groove 31, the gear ring 32, the connecting frame 4, the through groove 41 and the wiping cotton 42, when in use, the second driving motor 2 can drive the wiping cotton 42 to clean the dust on the protective transparent cover 16, thereby avoiding dust obstruction and affecting the shooting effect of the camera assembly 15.

[0067] Third embodiment

[0068] Please refer to Figure 5 Based on a video tracking control algorithm provided by the first embodiment of the present application, the third embodiment of the present application proposes another video tracking control algorithm. The third embodiment is only a preferred method of the first embodiment, and the implementation of the third embodiment will not affect the independent implementation of the first embodiment.

[0069] Specifically, the difference of a video tracking control algorithm provided in the third embodiment of the present application is that, in a video tracking control algorithm, a second Velcro is fixedly installed on the other side of the connecting frame, a first Velcro is adhered to one side of the second Velcro, and the first Velcro is fixedly installed on the other side of the wiping cotton.

[0070] The working principle of a video tracking control algorithm provided by the present invention is as follows:

[0071] After being used for a long time, when the wiping cotton 42 is severely worn and needs to be replaced, the user can separate the first Velcro 5 and the second Velcro 51 to disassemble and replace the wiping cotton 42.

[0072] Compared with the related art, the video tracking control algorithm provided by the present invention has the following beneficial effects:

[0073] By cooperating with each other through structures such as the first Velcro 5 and the second Velcro 51 , the first Velcro 5 and the second Velcro 51 are adhered to each other during use, so that the user can disassemble and replace the wiping cotton 42 conveniently.

[0074] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A video tracking control algorithm, characterized in that: The following steps are involved: S1. After the embedded box algorithm program is started, it pulls the video stream from the ball camera and recognizes the video frame by frame; S2. After the recognition algorithm recognizes the target, it returns the coordinates X, Y of the target with the center of the screen as the origin and the height H and width W of the target; S3, the tracking algorithm reads the current field of view angle Az of the ball camera, and calculates the angle Ap that the ball camera should rotate in the horizontal direction when the target is at the center of the screen according to the target coordinate X; and calculates the angle At that the ball camera should rotate in the pitch direction when the target is at the center of the screen according to the target coordinate Y; S4, the tracking algorithm reads the current focal length value Z of the ball camera, and calculates the focal length Z' that needs to be adjusted according to the target size; S5. Calculate the horizontal rotation speed Sp and the pitch rotation speed St according to Ap and At calculated above; S6. The tracking algorithm controls the ball camera to rotate at a horizontal speed Sp and a pitch speed St, and reads the horizontal angle and pitch angle of the ball camera every 50ms to recalculate new Ap and At; S7, read the algorithm recognition result every 300ms, and repeat the above calculation according to the new target result X, Y, W, H; S8. Finally, the target is tracked at the center of the screen. As the target moves, the ball camera automatically follows the target and adjusts the focal length according to the distance of the target.

2. A video tracking control algorithm according to claim 1, characterized in that: The formula of the angle Ap is Ap=X*Az / 1920, and the formula of the angle At is At=Y*Az / 1080.

3. A video tracking control algorithm according to claim 2, characterized in that: When the target H>W (personnel target), Z'=1080 / 4*Z / H, when the target W>H (car or ship target), Z'=1920 / 4*Z / W, and the focus value of the ball camera is adjusted by Z'.

4. A video tracking control algorithm according to claim 3, characterized in that: The horizontal rotation speed Sp=Pp*Ap / 0.3, and the pitch rotation speed St=Pt*At / 0.

3.

5. A video tracking control algorithm according to claim 4, characterized in that: The startup Pp and Pt are speed coefficients, which are related to the rotation speed of different models of ball cameras. It is necessary to test the actual scene to obtain a time of 0.3 for 300ms.

6. A video tracking control algorithm according to claim 4, characterized in that: The new Sp and St are calculated according to the new Ap and At, that is, the horizontal speed and pitch speed of the ball camera are adjusted every 50ms.

7. A video tracking control algorithm according to claim 1, characterized in that: A ball camera is required in the step S1, and the ball camera includes a mounting plate, a fixing column is fixedly mounted on the bottom of the mounting plate, a top plate is fixedly mounted on the bottom of the outer side of the fixing column, a protective transparent cover is fixedly mounted on the bottom of the top plate, a first driving motor is fixedly mounted on the bottom of the fixing column, and a camera assembly is fixedly mounted on the output end of the first driving motor.

8. A video tracking control algorithm according to claim 7, characterized in that: A mounting groove is formed on the top of the top plate, a second driving motor is fixedly mounted on the inner side of the mounting groove, and a driving gear is fixedly mounted on the output end of the second driving motor.

9. A video tracking control algorithm according to claim 8, characterized in that: The outer side surface of the fixed column is rotatably connected with a rotating circle, a groove is provided at the bottom of the inner side surface of the rotating circle, a gear ring is fixedly installed on the inner side surface of the groove, the gear ring is meshed with the outer side surface of the driving gear, a connecting frame is fixedly installed on one side of the rotating circle, a through groove is provided on one side of the connecting frame, a wiping cotton is provided on the other side of the connecting frame, and one side of the wiping cotton is attached to the side surface of the protective transparent cover.

10. A video tracking control algorithm according to claim 9, characterized in that: A second Velcro is fixedly installed on the other side of the connecting frame, a first Velcro is adhered to one side of the second Velcro, and the first Velcro is fixedly installed on the other side of the wiping cotton.

Images