Image acquisition camera for deep learning of sports actions

Abstract

The invention belongs to the technical field of motion capture, and particularly relates to an image acquisition camera for deep learning of sports motions, which comprises a motion capture camera, an adjusting holder and an image acquisition system for controlling the motion capture camera to operate, the adjusting holder is formed by combining a first arm, a second arm, a third arm, a fourth arm and a camera table, an inner gas injector used for cooling the motion capture camera by means of kinetic energy generated when the cradle head operates is assembled in the third machine arm, and an anti-coincidence capture unit, an infrared light detection unit and a motion acquisition unit are arranged in the image acquisition system. The infrared light detection unit is internally provided with a power-on self-test function unit and an abnormal reflection point detection unit. According to the invention, the acquisition range of a single camera can be expanded, the input number of equipment can be reduced, the influence of pixel point superposition on the accuracy can be reduced, in addition, the method also has a function of checking abnormal reflective points, and the acquisition error is greatly reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of motion capture, and particularly relates to an image acquisition camera for deep learning of sports actions. Background Art

[0002] Motion capture is to set trackers at key parts of a moving object; trackers are set at key parts of the moving object, and the Motion capture system captures the positions of the trackers, and then obtains the data of three-dimensional space coordinates after being processed by a computer; when the data is recognized by the computer, it can be applied in fields such as animation production, gait analysis, biomechanics, and ergonomics.

[0003] Problems existing in the prior art: Existing motion capture cameras have the following disadvantages. The motion capture camera has a more accurate acquisition effect compared to inertial acquisition devices, but its acquisition range is limited and generally cannot be changed. When comprehensive acquisition is required, multiple motion capture cameras need to be arranged in the venue. Although multiple motion capture cameras can handle the problem of pixel point coincidence, the economic cost required is too high, making the cost of the motion capture camera venue too expensive; in addition, during the use of the motion capture camera, it receives the reflected light of the infrared rays emitted by its own infrared lamp ring. Therefore, the use of the motion capture camera has additional restrictive conditions for the venue. First, no sunlight can enter the venue, because the infrared light in sunlight will be captured by the motion capture camera after reflection. Second, there cannot be any additional objects with a reflective effect in the venue, because the light reflected by them does not belong to the pixel points reflected by the devices worn on the bodies of the action personnel. The existing motion capture cameras cannot overcome the above restrictive conditions. Therefore, this device does not have the function of excluding abnormal pixel points. However, during actual use, it is inevitable that abnormal infrared reflections will be captured by the motion capture camera, and unexpected situations cannot be eradicated. But because the device itself does not have the ability to exclude abnormalities, the subsequent synthesized images will inevitably have a large error, and additional time will be required for resetting later. Summary of the Invention

[0004] The purpose of the present invention is to provide an image acquisition camera for deep learning of sports actions, which can expand the acquisition range of a single camera, reduce the number of devices invested, and can reduce the impact of pixel point coincidence on accuracy. In addition, it also has the function of detecting abnormal reflection points, greatly reducing the acquisition error.

[0005] The technical solution adopted by the present invention is specifically as follows: An image acquisition camera for deep learning of sports movements includes a motion capture camera, an adjustment pan-tilt head, and an image acquisition system for controlling the operation of the motion capture camera. A camera is installed inside the lens of the motion capture camera, and an infrared lamp ring is arranged around the camera. The adjustment pan-tilt head is composed of a first arm, a second arm, a third arm, a fourth arm, and a camera platform. Among them, the mounting base forms a rotational assembly relationship with one end of the first arm, and the second arm forms a rotational assembly relationship with the other end of the first arm. One end of the third arm forms a rotational assembly relationship with the other end of the second arm, and one end of the fourth arm forms a rotational assembly relationship with the other end of the third arm. The camera platform is rotatably installed at the other end of the fourth arm, and the motion capture camera is rotatably assembled inside the camera platform; An internal injector for cooling the motion capture camera by means of the kinetic energy during the operation of the adjustment pan-tilt head is assembled inside the third arm; An anti-overlap capture unit, an infrared light detection unit, and an action capture unit are arranged inside the image acquisition system. The anti-overlap capture unit is used to drive the adjustment pan-tilt head to adjust the shooting angle of the motion capture camera when pixel point overlap is detected in the captured image. The infrared light detection unit is used to detect abnormal infrared light reflection points within the shooting range. The action capture unit is used to perform normal action capture work; A pixel point information acquisition unit, a pixel point alignment unit, an overlap area determination unit, a coverage point analysis unit, and a lens adjustment unit are arranged inside the anti-overlap capture unit. The pixel point information acquisition unit is used to collect pixel point information in real time. The pixel point alignment unit is used to connect multiple pixel points with the same motion law. The overlap area determination unit is used to identify the pixel point overlap situation in the captured image. The coverage point analysis unit is used to analyze the pixel point overlap image and calculate the motion path L at the time of overlap. The lens adjustment unit is used to move the motion capture camera to a non-overlapping pixel point shooting angle according to the path L information; A power-on self-check function unit and an abnormal reflection point detection unit are arranged inside the infrared light detection unit. The power-on self-check function unit is used to enable the abnormal reflection point detection function when the device is powered on. The abnormal reflection point detection unit is used to detect abnormal reflection points during the action capture process; An infrared lamp group shutdown unit, an infrared light reception trigger unit, a full-range lens movement trigger unit, and a self-check abnormal prompt unit are arranged inside the power-on self-check function unit. The infrared lamp group shutdown unit is used to temporarily turn off the infrared lamp ring when the device is powered on. The infrared light reception trigger unit is used to turn on the infrared light reception function. The full-range lens movement trigger unit is used to control the adjustment pan-tilt head to move the motion capture camera in all directions within the shootable range. The self-check abnormal prompt unit is used to issue a prompt when an abnormal reflection point is found; Inside the abnormal reflection point detection unit, there are a pixel column outside determination unit, a different pixel column determination unit, a lens emergency stop trigger unit, and an abnormal reflection point prompt unit. The pixel column outside determination unit is used to determine whether the redundant emission points belong to the pixel column according to the quantity information when the pixel points do not completely coincide. The different pixel column determination unit is used to determine whether the redundant emission points belong to the pixel column according to the pixel column shape when the pixel points completely coincide. The lens emergency stop trigger unit is used to stop the movement of the motion capture camera in time when an abnormal reflection point is found. The abnormal reflection point prompt unit is used to give a prompt when an abnormal reflection point is found.

[0006] Inside the covered point analysis unit, there are a coincidence point coordinate information processing unit, a covered point calculation unit, and a motion path L calculation unit. The coincidence point coordinate information processing unit is used to analyze the coordinate information of the coincident pixel points. The covered point calculation unit is used to analyze the covered pixel points and pixel columns. The motion path L calculation unit is used to analyze one of the motion paths L of the pixel column according to the change of the coordinate information.

[0007] Inside the lens adjustment unit, there are a path L information copying unit and a motor group driving unit. The path L information copying unit is used to copy the path L information analyzed by the motion path L calculation unit. The motor group driving unit is used to control and adjust the pan-tilt head to make the motion capture camera move along the path L.

[0008] One end of the first robotic arm is fixedly installed with a first motor inside, and the output shaft of the first motor is fixedly connected to the mounting seat. The other end of the first robotic arm is fixedly installed with a second motor inside, and the output shaft of the second motor is fixedly connected to one end of the second robotic arm.

[0009] The other end of the second robotic arm is fixedly installed with a third motor inside, and the output shaft of the third motor is fixedly connected to one end of the third robotic arm.

[0010] One end of the fourth robotic arm is fixedly installed with a fourth motor inside, and the output shaft of the fourth motor is fixedly connected to the other end of the third robotic arm. And the other end of the fourth robotic arm is fixedly installed with a fifth motor inside, and the output shaft of the fifth motor is fixedly connected to the camera platform.

[0011] One side of the camera platform is fixedly installed with a sixth motor inside, and the output shaft of the sixth motor is fixedly connected to one side of the motion capture camera.

[0012] A first gear is fixedly installed on the outer wall of the second robotic arm at the junction with the third robotic arm. A first gear is also fixedly installed on the outer wall of the fourth robotic arm at the junction with the third robotic arm.

[0013] Both ends inside the third robotic arm are fixedly installed with internal air injectors. One end outer wall of the internal air injector is rotationally assembled with a second gear, and the second gear is arranged outside the third robotic arm and meshes with the corresponding first gear. A crankshaft is rotationally installed inside the internal air injector, and the crankshaft is fixedly connected to the corresponding second gear. The end of the crankshaft is rotationally connected to a rotating arm, and the end of the rotating arm is connected to a piston, and the piston is slidably installed inside the internal air injector.

[0014] Both sides of one end outer wall of the internal air injector are respectively fixedly connected with an air suction pipe and an air outlet pipe. The air suction pipe and the air outlet pipe both extend outside the third robotic arm, and the two air outlet pipes are jointly connected to the housing of the motion capture camera.

[0015] The technical effects achieved by the present invention are as follows: In the present invention, the investment in the adjustable pan-tilt is beneficial to expanding the acquisition range of the motion capture camera, reducing the number of motion capture cameras invested, and lowering the economic cost required.

[0016] In the present invention, while the adjustable pan-tilt is operating, heat dissipation work can be carried out on the motion capture camera. The airflow directly injected into the housing of the motion capture camera can efficiently take away the heat generated by its work, with a good heat dissipation effect. Moreover, this heat dissipation process is achieved by utilizing the kinetic energy during the relative rotation between the corresponding robotic arms and does not require additional power equipment to complete, thus not generating additional power consumption.

[0017] In the present invention, through the function of preventing overlapping captures of the device, the pictures with overlapping pixel points in the pictures captured by the motion capture camera are reduced, which can effectively solve the problems of missed detection of motion acquisitions and incorrect pictures caused by overlapping pixel points in traditional motion capture technologies, ensuring the accuracy of acquisitions. In addition, it greatly expands the acquisition range of a single motion capture camera. Compared with the form of fixedly installing multiple traditional cameras, this anti-overlapping capture function indirectly reduces the number of motion capture cameras that need to be invested, lowering the economic cost of motion capture technology.

[0018] In the present invention, the motion capture acquisition device is enabled to have the function of self-checking for abnormal reflection points at startup, increasing the functions of the motion capture acquisition device, avoiding the situation where abnormal reflection points existing at the beginning cannot be identified and excluded at the initial stage of motion capture work, solving the problem that the abnormal reflection points existing at the beginning directly affect the motion capture acquisition work, greatly reducing the acquisition error, and further improving the acquisition accuracy.

[0019] In the present invention, the motion capture acquisition device is enabled to have the function of excluding abnormal reflection points, which further increases the functions of the motion capture acquisition device. It can timely identify abnormal pixel points generated by abnormal infrared light reflection points during the motion capture work, solve the problem that abnormal reflection points directly affect the motion capture acquisition work, greatly reduce the acquisition error in the motion capture work, and further improve the acquisition accuracy. Brief Description of the Drawings

[0020] Figure 1 is the front view structural diagram of the image acquisition camera provided by the embodiment of the present invention; Figure 2 is the rear view structural diagram of the image acquisition camera provided by the embodiment of the present invention; Figure 3 is the system diagram of the image acquisition system provided by the embodiment of the present invention; Figure 4 is the flowchart of the anti-overlap capture function provided by the embodiment of the present invention; Figure 5 is the flowchart of the infrared light detection function provided by the embodiment of the present invention; Figure 6 is the three-dimensional coordinate diagram of the pixel points before and after the formation of the path L provided by the embodiment of the present invention; Figure 7 is the schematic diagram of two abnormal reflection points that appear during the motion capture provided by the embodiment of the present invention; Figure 8 is the cross-sectional structural diagram of the third robotic arm provided by the embodiment of the present invention; Figure 9 is the internal structural diagram of the internal injector provided by the embodiment of the present invention.

[0021] In the drawings, the list of components represented by each reference numeral is as follows: 1. Motion capture camera; 101. Camera; 102. Infrared lamp ring; 2. Adjusting pan-tilt head; 201. Arm 1; 202. Motor 1; 203. Mounting base; 204. Arm 2; 205. Motor 2; 206. Motor 3; 207. Arm 3; 208. Arm 4; 209. Motor 4; 210. Camera stage; 211. Motor 6; 3. Gear 1; 4. Internal injector; 401. Suction pipe; 402. Outlet pipe; 403. Gear 2; 404. Crankshaft; 405. Rotating arm; 406. Piston; 5. Image acquisition system; 501. Anti-overlap capture unit; 502. Infrared light detection unit; 503. Action acquisition unit; 504. Pixel point information acquisition unit; 505. Pixel point alignment unit; 506. Overlap area determination unit; 507. Covered point analysis unit; 508. Lens adjustment unit; 509. Overlap point coordinate information processing unit; 510. Covered point calculation unit; 511. Motion path L calculation unit; 512. Path L information copying unit; 513. Motor group drive unit; 514. Power-on self-check function unit; 515. Infrared lamp group shutdown unit; 516. Infrared light reception trigger unit; 517. Full-range lens movement trigger unit; 518. Self-check exception prompt unit; 519. Abnormal reflection point detection unit; 520. Pixel column outer determination unit; 521. Opposite-sex pixel column determination unit; 522. Lens emergency stop trigger unit; 523. Abnormal reflection point prompt unit. Detailed implementation mode

[0022] In order to make the purpose and advantages of the present invention clearer, the present invention will be specifically described below in conjunction with embodiments. It should be understood that the following text is only used to describe one or several specific implementation manners of the present invention, and does not strictly limit the specific protection scope claimed by the present invention.

[0023] As Figures 1-8 shown, an image acquisition camera for deep learning of sports actions includes a motion capture camera 1, an adjusting pan-tilt head 2, and an image acquisition system 5 for controlling the operation of the motion capture camera 1. A camera 101 is installed inside the lens of the motion capture camera 1, and an infrared lamp ring 102 is arranged around the camera 101.

[0024] Embodiment 1: Refer to the attached Figure 1 and Figure 2, the adjustable pan-tilt 2 is composed of arm one 201, arm two 204, arm three 207, arm four 208 and camera platform 210. Among them, the mounting base 203 forms a rotational assembly relationship with one end of arm one 201, and arm two 204 forms a rotational assembly relationship with the other end of arm one 201. One end of arm three 207 forms a rotational assembly relationship with the other end of arm two 204, and one end of arm four 208 forms a rotational assembly relationship with the other end of arm three 207. The camera platform 210 is rotatably installed at the other end of arm four 208, and the motion capture camera 1 is rotationally assembled inside the camera platform 210. Inside one end of arm one 201, a motor one 202 is fixedly installed, and the output shaft of the motor one 202 is fixedly connected to the mounting base 203. Inside the other end of arm one 201, a motor two 205 is fixedly installed, and the output shaft of the motor two 205 is fixedly connected to one end of arm two 204. Inside the other end of arm two 204, a motor three 206 is fixedly installed, and the output shaft of the motor three 206 is fixedly connected to one end of arm three 207. Inside one end of arm four 208, a motor four 209 is fixedly installed, and the output shaft of the motor four 209 is fixedly connected to the other end of arm three 207. Inside the other end of arm four 208, a motor five is fixedly installed, and the output shaft of the motor five is fixedly connected to the camera platform 210. Inside one side of the camera platform 210, a motor six 211 is fixedly installed, and the output shaft of the motor six 211 is fixedly connected to one side of the motion capture camera 1.

[0025] According to the above structure, the adjustable pan-tilt 2 is installed through the mounting base 203. The motor one 202 is used to control the rotational operation of the motor one 202. The motor two 205 is used to control the relative rotation between arm one 201 and arm two 204. The motor three 206 is used to control the relative rotation between arm two 204 and arm three 207. The motor four 209 is used to control the relative rotation between arm three 207 and arm four 208. The motor five is used to control the relative rotation between arm four 208 and the camera platform 210. Finally, the motor six 211 can be used to control the angle of the motion capture camera 1. Compared with the traditional fixed-structure motion capture camera 1, the introduction of this adjustable pan-tilt 2 is beneficial to expanding the acquisition range of the motion capture camera 1, reducing the number of motion capture cameras 1 invested, and reducing the economic cost required.

[0026] The working principle of the present invention is: The adjustable pan-tilt 2 is installed through the mounting base 203. The motor one 202 is used to control the rotational operation of the motor one 202. The motor two 205 is used to control the relative rotation between arm one 201 and arm two 204. The motor three 206 is used to control the relative rotation between arm two 204 and arm three 207. The motor four 209 is used to control the relative rotation between arm three 207 and arm four 208. The motor five is used to control the relative rotation between arm four 208 and the camera platform 210. Finally, the motor six 211 can be used to control the angle of the motion capture camera 1.

[0027] Example Two: Refer to the attached Figure 3 , inside the image acquisition system 5, there are an anti-overlap capture unit 501, an infrared light detection unit 502, and an action acquisition unit 503. When the anti-overlap capture unit 501 detects pixel point overlap in the captured image, it is used to drive the adjustment pan-tilt 2 to adjust the shooting angle of the motion capture camera 1. The infrared light detection unit 502 is used to check for abnormal infrared light reflection points within the shooting range, and the action acquisition unit 503 is used to perform normal action acquisition work.

[0028] Refer to the attached Figure 3 , Figure 4 and Figure 6 , inside the anti-overlap capture unit 501, there are a pixel point information acquisition unit 504, a pixel point alignment unit 505, an overlap area determination unit 506, a covered point position analysis unit 507, and a lens adjustment unit 508. The pixel point information acquisition unit 504 is used to collect pixel point information in real time. The pixel point alignment unit 505 is used to connect multiple pixel points with the same motion law. The overlap area determination unit 506 is used to identify pixel point overlap in the captured image. The covered point position analysis unit 507 is used to analyze the pixel point overlap image and calculate the motion path L at the time of overlap. The lens adjustment unit 508 is used to move the motion capture camera 1 to a non-overlapping pixel shooting angle according to the path L information; Refer to the attached Figure 3 , Figure 4 and Figure 6 , inside the covered point position analysis unit 507, there are a coincidence point position coordinate information processing unit 509, a covered point position calculation unit 510, and a motion path L calculation unit 511. The coincidence point position coordinate information processing unit 509 is used to analyze the coordinate information of the overlapping pixel points. The covered point position calculation unit 510 is used to analyze the covered pixel points and pixel columns. The motion path L calculation unit 511 is used to analyze one motion path L of the pixel column according to the change of the coordinate information; Inside the lens adjustment unit 508, there are a path L information copying unit 512 and a motor group driving unit 513. The path L information copying unit 512 is used to copy the path L information analyzed by the motion path L calculation unit 511, and the motor group driving unit 513 is used to control the adjustment pan-tilt 2 to make the motion capture camera 1 move along the path L.

[0029] According to the above structure, taking Figure 6 as an example, the pixel point information acquisition unit 504 collects pixel point change information in real time. The pixel point alignment unit 505 regularizes the corresponding number of pixel points into a pixel column through the human body limb model, such as Figure 6As shown, this figure is a pixel model of a hand. Among them, pixel columns P1, P2, P3, P4, and P5 composed of different numbers of pixel points respectively correspond to five fingers. During the movement of the performer, the overlapping area determination unit 506 can identify the situation of pixel points overlapping in the picture. For example Figure 6 As shown in figure b of Figure 6 , in this figure, column P3 and column P4 overlap, and column P4 and column P5 overlap. When the overlapping situation occurs, the overlapping point position coordinate information processing unit 509 analyzes the coordinate information of the overlapping pixel points, that is, the coordinate information of each pixel point in columns P3, P4, and P5. The covered point position calculation unit 510 can immediately analyze the covered pixel points and pixel columns. In figure b of this figure, it can be analyzed that column P4 is partially covered by column P3, and column P5 is partially covered by column P4. Immediately afterwards, the covering point position analysis unit 507 analyzes and calculates the movement path L at the time of overlap according to the coordinate changes of the corresponding pixel points. Subsequently, the information of path L will be immediately copied to the path L information copying unit 512 of the lens adjustment unit 508. Finally, the motor group driving unit 513 controls each motor in the adjustment pan-tilt 2, and finally realizes the movement of the motion capture camera 1 along path L and ensures that the camera 101 is aligned with the limbs of the performer. In this way, the picture with overlapping pixel points in the captured picture of the motion capture camera 1 is reduced. Through the above process, the problems of missed detection of action collection and picture errors caused by pixel point overlap in traditional motion capture technology can be effectively solved, ensuring the accuracy of collection. In addition, the collection range of a single motion capture camera 1 is greatly expanded. Compared with the form of fixed installation of multiple traditional cameras, this anti-overlap capture function indirectly reduces the number of motion capture cameras 1 that need to be invested, reducing the economic cost of motion capture technology.

[0030] The working principle of the present invention is: after the action performer wears the corresponding equipment, the motion capture camera 1 can capture the movement of pixel points. Among them, the pixel point information acquisition unit 504 collects the pixel point change information in real time, and the pixel point forming column unit 505 regularizes the corresponding number of pixel points into a pixel column through the human body limb model, such as Figure 6 As shown, this figure is a pixel model of a hand. Among them, pixel columns P1, P2, P3, P4, and P5 composed of different numbers of pixel points respectively correspond to five fingers. During the movement of the performer, the overlapping area determination unit 506 can identify the situation of pixel points overlapping in the picture. For example Figure 6As shown in Figure b, in this figure, column P3 coincides with column P4, and column P4 coincides with column P5. When the coincidence occurs, the coincidence point position coordinate information processing unit 509 analyzes the coordinate information of the coincident pixel points, that is, the coordinate information of each pixel point in columns P3, P4, and P5. The covered point position calculation unit 510 can immediately analyze the covered pixel points and pixel columns. In this Figure b, it can be analyzed that column P4 is partially covered by column P3, and column P5 is partially covered by column P4. Immediately afterwards, the covered point position analysis unit 507 analyzes and calculates the movement path L at the time of coincidence according to the coordinate changes of the corresponding pixel points. Subsequently, the information of path L will be immediately copied to the path L information copying unit 512 of the lens adjustment unit 508. Finally, the motor group driving unit 513 controls each motor in the adjustment pan-tilt 2, and finally realizes the movement of the motion capture camera 1 along the path L and can ensure that the camera 101 is aligned with the limbs of the performing personnel. In this way, the pixel point coincidence in the captured image of the motion capture camera 1 is reduced.

[0031] Embodiment 3: Refer to the appendix Figure 3 and Figure 5 As shown in the figure, the internal of the infrared light detection unit 502 is provided with a power-on self-check function unit 514 and an abnormal reflection point detection unit 519. The power-on self-check function unit 514 is used to enable the abnormal reflection point detection function when the device is powered on; the internal of the power-on self-check function unit 514 is provided with an infrared lamp group shutdown unit 515, an infrared light reception trigger unit 516, a lens full-range movement trigger unit 517, and a self-check abnormal prompt unit 518. The infrared lamp group shutdown unit 515 is used to temporarily turn off the infrared lamp ring 102 when the device is powered on. The infrared light reception trigger unit 516 is used to turn on the infrared light reception function. The lens full-range movement trigger unit 517 is used to control the adjustment pan-tilt 2 to move the motion capture camera 1 in all directions within the shootable range. The self-check abnormal prompt unit 518 is used to give a prompt when an abnormal reflection point is found.

[0032] According to the above structure, when the device is just powered on, the power-on self-check function will be triggered. First, when powering on, the infrared lamp group shutdown unit 515 temporarily turns off the infrared lamp ring 102. Secondly, the infrared light reception trigger unit 516 turns on the infrared light reception function. Immediately afterwards, the lens full-range movement trigger unit 517 controls and adjusts the pan-tilt 2 to make the motion capture camera 1 move in all directions within the shootable range. When infrared reflection is detected, it can be determined at this time that the reflection is not the light emitted by the infrared lamp ring 102, and thus it can be directly determined as an abnormal reflection point. Immediately afterwards, the motion capture camera 1 makes an emergency stop, and the self-check abnormality prompt unit 518 issues a prompt for self-check abnormality. Through the above process, the motion capture acquisition device is enabled to perform power-on self-check for abnormal reflection points, increasing the functions of the motion capture acquisition device, avoiding the situation where abnormal reflection points existing at the beginning cannot be identified and excluded at the initial stage of motion capture work, solving the problem that the abnormal reflection points existing at the beginning directly affect the motion capture acquisition work, greatly reducing the acquisition error, and further improving the acquisition accuracy.

[0033] The working principle of the present invention is as follows: when the device is just powered on, the power-on self-check function will be triggered. First, when powering on, the infrared lamp group shutdown unit 515 temporarily turns off the infrared lamp ring 102. Secondly, the infrared light reception trigger unit 516 turns on the infrared light reception function. Immediately afterwards, the lens full-range movement trigger unit 517 controls and adjusts the pan-tilt 2 to make the motion capture camera 1 move in all directions within the shootable range. When infrared reflection is detected, it can be determined at this time that the reflection is not the light emitted by the infrared lamp ring 102, and thus it can be directly determined as an abnormal reflection point. Immediately afterwards, the motion capture camera 1 makes an emergency stop, and the self-check abnormality prompt unit 518 issues a prompt for self-check abnormality.

[0034] Embodiment 4: Refer to the appendix Figure 3 and Figure 7 The abnormal reflection point detection unit 519 is used to detect abnormal reflection points during the action acquisition process; inside the abnormal reflection point detection unit 519, there are a pixel column outer determination unit 520, a different pixel column determination unit 521, a lens emergency stop trigger unit 522, and an abnormal reflection point prompt unit 523. The pixel column outer determination unit 520 is used to determine whether the redundant emission point belongs to the pixel column according to the quantity information when the pixel points do not completely overlap. The different pixel column determination unit 521 is used to determine whether the redundant emission point belongs to the pixel column according to the pixel column shape when the pixel points completely overlap. The lens emergency stop trigger unit 522 is used to stop the movement of the motion capture camera 1 in time when an abnormal reflection point is found. The abnormal reflection point prompt unit 523 is used to issue a prompt when an abnormal reflection point is found.

[0035] According to the above structure, during the action acquisition process, when redundant pixel points appear near a certain column of pixel columns, the abnormal reflection point detection unit 519 starts to operate. As Figure 7 shown, when the corresponding pixel column is not covered, such as Figure 7 in column P1, the abnormal pixel point Y1 near column P1 is caused by the camera capturing an abnormal reflection point. At this time, the off-pixel-column determination unit 520 can determine that point Y1 does not belong to column P1 based on the pixel point quantity information of column P1, and further determine that Y1 is an abnormal pixel point; when the corresponding pixel column is covered, such as Figure 7 in column P5, the abnormal pixel point Y2 near column P5 is caused by the camera capturing an abnormal reflection point. Since a certain pixel point in column P5 is covered, Y2 forms a new array with other pixel points in column P5, but this new array is of a different type and does not conform to the normal line shape of a finger. At this time, the different-type pixel-column determination unit 521 can determine that point Y2 does not belong to column P5 based on the normal shape information of column P5, and further determine that Y2 is an abnormal pixel point. When an abnormal pixel point is detected, the lens emergency stop trigger unit 522 stops the movement of the motion capture camera 1, and the abnormal reflection point prompt unit 523 issues a prompt. Through the above process, the motion capture acquisition device has the function of excluding abnormal reflection points, and the function of the motion capture acquisition device is increased. It can timely identify abnormal pixel points caused by abnormal infrared light reflection points during the motion capture work process, solve the problem that abnormal reflection points directly affect the motion capture acquisition work, greatly reduce the acquisition error in the motion capture work, and further improve the acquisition accuracy.

[0036] The working principle of the present invention is: during the action acquisition process, when redundant pixel points appear near a certain column of pixel columns, the abnormal reflection point detection unit 519 starts to operate. As Figure 7 shown, when the corresponding pixel column is not covered, such as Figure 7 in column P1, the abnormal pixel point Y1 near column P1 is caused by the camera capturing an abnormal reflection point. At this time, the off-pixel-column determination unit 520 can determine that point Y1 does not belong to column P1 based on the pixel point quantity information of column P1, and further determine that Y1 is an abnormal pixel point; when the corresponding pixel column is covered, such as Figure 7 in column P5, the abnormal pixel point Y2 near column P5 is caused by the camera capturing an abnormal reflection point. Since a certain pixel point in column P5 is covered, Y2 forms a new array with other pixel points in column P5, but this new array is of a different type and does not conform to the normal line shape of a finger. At this time, the different-type pixel-column determination unit 521 can determine that point Y2 does not belong to column P5 based on the normal shape information of column P5, and further determine that Y2 is an abnormal pixel point. When an abnormal pixel point is detected, the lens emergency stop trigger unit 522 stops the movement of the motion capture camera 1, and the abnormal reflection point prompt unit 523 issues a prompt.

[0037] Embodiment Five: Refer to the appendix Figure 8 , an internal air injector 4 for cooling the motion capture camera 1 by means of the kinetic energy when the adjustment pan-tilt 2 operates is assembled inside the arm three 207; a first gear 3 is fixedly installed on the outer wall of the arm two 204 at the connection with the arm three 207, and a first gear 3 is also fixedly installed on the outer wall of the arm four 208 at the connection with the arm three 207; both sides of one end of the air injector 4 are fixedly connected with an air suction pipe 401 and an air outlet pipe 402 respectively, both the air suction pipe 401 and the air outlet pipe 402 extend outside the arm three 207, and the two air outlet pipes 402 are jointly connected to the housing of the motion capture camera 1.

[0038] Refer to the appendix Figure 9 , air injectors 4 are fixedly installed at both ends inside the arm three 207, a second gear 403 is rotatably assembled on the outer wall of one end of the air injector 4, and the second gear 403 is arranged outside the arm three 207 and meshes with the corresponding first gear 3, a crankshaft 404 is rotatably installed inside the air injector 4, and the crankshaft 404 is fixedly connected to the corresponding second gear 403, the end of the crankshaft 404 is rotatably connected to a rocker arm 405, and the end of the rocker arm 405 is connected to a piston 406, and the piston 406 is slidably installed inside the air injector 4.

[0039] According to the above structure, the adjustment pan-tilt 2 is used to control the movement of the motion capture camera 1. Among them, when relative rotation occurs between the arm two 204 and the arm three 207 and between the arm three 207 and the arm four 208, the corresponding second gear 403 will rotate around the first gear 3 and rotate itself at the same time. When the second gear 403 rotates itself, it will drive the crankshaft 404 to rotate inside the air injector 4, and then drive the piston 406 to reciprocate through the rocker arm 405, so that positive and negative pressures are continuously formed inside the air injector 4. When the pressure inside the air injector 4 is negative, it can suck air through the air suction pipe 401, and when the pressure inside the air injector 4 is positive, the gas inside it can be injected into the housing of the motion capture camera 1 through the air outlet pipe 402. The heat generated by the internal mechanism of the motion capture camera 1 during operation is taken away by the flow of air, achieving the effect of heat dissipation. In this process, the heat dissipation work of the motion capture camera 1 can be carried out while the adjustment pan-tilt 2 is operating. The air flow directly injected into the housing of the motion capture camera 1 can efficiently take away the heat generated by its operation, having a good heat dissipation effect. Moreover, this heat dissipation process is realized by using the kinetic energy during the relative rotation between the corresponding arms, and no additional power equipment needs to be invested to complete it, so no additional power consumption will be generated.

[0040] The working principle of the present invention is as follows: The adjustment pan-tilt 2 is used to control the movement of the motion capture camera 1. Among them, when relative rotation occurs between the second arm 204 and the third arm 207, and between the third arm 207 and the fourth arm 208, the corresponding second gear 403 will rotate around the first gear 3 and rotate on its own at the same time. When the second gear 403 rotates on its own, it will drive the crankshaft 404 to rotate inside the internal injector 4, and then drive the piston 406 to reciprocate through the rotating arm 405, so that positive and negative pressures are continuously formed inside the internal injector 4. When the inside of the internal injector 4 is negative pressure, it can suck air through the suction pipe 401, and when the inside of the internal injector 4 is positive pressure, the gas inside it can be injected into the casing of the motion capture camera 1 through the outlet pipe 402, and the heat generated by the internal body of the motion capture camera 1 during operation can be taken away through the flow of air, achieving the effect of heat dissipation.

[0041] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention. The structures, devices, and operation methods not specifically described and explained in the present invention, unless otherwise specified and limited, are implemented according to the conventional means in the art.

Claims

1. An image acquisition camera for deep learning of sports movements, comprising a motion capture camera (1), an adjustment pan / tilt (2), and an image acquisition system (5) for controlling the operation of the motion capture camera (1), characterized in that: The adjusting pan-tilt platform (2) comprises a machine arm three (207); the machine arm three (207) is internally assembled with an internal gas injector (4) for cooling the motion capture camera (1) by utilizing the kinetic energy of the adjusting pan-tilt platform (2) during operation; The image acquisition system (5) is internally provided with an anti-coincidence capture unit (501), an infrared light detection unit (502) and a motion acquisition unit (503); The anti-overlap capture unit (501) is internally provided with a pixel point information acquisition unit (504), a pixel point alignment unit (505), an overlap region determination unit (506), a coverage point position analysis unit (507) and a lens adjustment unit (508), wherein the pixel point information acquisition unit (504) is used to acquire pixel point information in real time, the pixel point alignment unit (505) is used to connect a plurality of pixel points in the same motion law, the overlap region determination unit (506) is used to identify the overlap of pixel points in the captured image, the coverage point position analysis unit (507) is used to analyze the pixel point overlap image and calculate the motion path L when the pixel points overlap, and the lens adjustment unit (508) is used to move the motion capture camera (1) to a shooting angle where the pixel points do not overlap according to the path L information; The infrared light detection unit (502) is provided with a power-on self-check function unit (514) and an abnormal reflection point detection unit (519) inside. The power-on self-check function unit (514) is used to enable an abnormal reflection point troubleshooting function when the device is turned on, and the abnormal reflection point detection unit (519) is used to troubleshoot abnormal reflection points during the motion acquisition process.

2. The image acquisition camera for deep learning of sports movements according to claim 1, characterized in that: The anti-overlap capture unit (501) is used to drive the pan / tilt (2) to adjust the shooting angle of the motion capture camera (1) when pixel overlap is detected in the captured image; the infrared light detection unit (502) is used to check for abnormal infrared light reflection points within the shooting range; and the motion capture unit (503) is used to perform normal motion capture work.

3. The image acquisition camera for deep learning of sports movements according to claim 2, characterized in that: The power-on self-check function unit (514) is internally provided with an infrared light group stop unit (515), an infrared light receiving trigger unit (516), a lens full-range movement trigger unit (517), and a self-check abnormality prompt unit (518); the infrared light group stop unit (515) is used to temporarily turn off the infrared light ring (102) when the device is turned on; the infrared light receiving trigger unit (516) is used to turn on the infrared light receiving function; the lens full-range movement trigger unit (517) is used to control and adjust the pan / tilt head (2) so that the motion capture camera (1) can move in all directions within the shooting range; and the self-check abnormality prompt unit (518) is used to issue a prompt when an abnormal reflection point is found; The abnormal reflection point detection unit (519) is internally provided with an out-of-pixel column determination unit (520), an opposite-sex pixel column determination unit (521), a lens emergency stop trigger unit (522), and an abnormal reflection point prompt unit (523). The out-of-pixel column determination unit (520) is used to determine whether the redundant emission points belong to the pixel column according to quantity information when the pixel points are not completely overlapped, the opposite-sex pixel column determination unit (521) is used to determine whether the redundant emission points belong to the pixel column according to the shape of the pixel column when the pixel points are completely overlapped, the lens emergency stop trigger unit (522) is used to stop the motion of the motion capture camera (1) in time when an abnormal reflection point is found, and the abnormal reflection point prompt unit (523) is used to issue a prompt when an abnormal reflection point is found.

4. The image acquisition camera for deep learning of sports movements according to claim 1, characterized in that: The covering point analysis unit (507) is internally provided with an overlapping point coordinate information processing unit (509), a covered point estimation unit (510) and a motion path L calculation unit (511); the overlapping point coordinate information processing unit (509) is used to analyze the coordinate information of overlapping pixel points; the covered point estimation unit (510) is used to analyze the covered pixel points and pixel columns; and the motion path L calculation unit (511) is used to analyze one of the motion paths L of the pixel column according to the change of the coordinate information.

5. The image acquisition camera for deep learning of sports movements according to claim 4, characterized in that: The lens adjustment unit (508) is internally provided with a path L information copying unit (512) and a motor group driving unit (513); the path L information copying unit (512) is used to copy the path L information analyzed by the motion path L calculation unit (511); and the motor group driving unit (513) is used to control and adjust the pan / tilt head (2) so that the motion capture camera (1) moves along the path L.

6. The image acquisition camera for deep learning of sports movements according to claim 1, characterized in that: A camera (101) is installed inside the lens of the motion capture camera (1), and an infrared light ring (102) is arranged around the camera (101); The adjusting pan-tilt platform (2) further comprises a machine arm 1 (201), a mounting base (203), a machine arm 2 (204), a machine arm 4 (208) and a camera platform (210), wherein the mounting base (203) and one end of the machine arm 1 (201) form a rotational assembly relationship, and the machine arm 2 (204) and the other end of the machine arm 1 (201) form a rotational assembly relationship, one end of the machine arm 3 (207) and the other end of the machine arm 2 (204) form a rotational assembly relationship, and one end of the machine arm 4 (208) and the other end of the machine arm 3 (207) form a rotational assembly relationship, the camera platform (210) is rotationally mounted on the other end of the machine arm 4 (208), and the motion capture camera (1) is rotationally assembled inside the camera platform (210); A motor 1 (202) is fixedly installed inside one end of the arm 1 (201), and the output shaft of the motor 1 (202) is fixedly connected to the mounting seat (203); a motor 2 (205) is fixedly installed inside the other end of the arm 1 (201), and the output shaft of the motor 2 (205) is fixedly connected to one end of the arm 2 (204).

7. The image acquisition camera for deep learning of sports movements according to claim 6, characterized in that: A motor 3 (206) is fixedly installed inside the other end of the machine arm 2 (204), and the output shaft of the motor 3 (206) is fixedly connected to one end of the machine arm 3 (207); A motor four (209) is fixedly installed inside one end of the arm four (208), and the output shaft of the motor four (209) is fixedly connected to the other end of the arm three (207), while a motor five is fixedly installed inside the other end of the arm four (208), and the output shaft of the motor five is fixedly connected to the camera platform (210).

8. The image acquisition camera for deep learning of sports movements according to claim 7, characterized in that: A sixth motor (211) is fixedly installed inside one side of the camera platform (210), and an output shaft of the sixth motor (211) is fixedly connected to one side of the motion capture camera (1).

9. The image acquisition camera for deep learning of sports movements according to claim 7, characterized in that: The outer wall of the second arm (204) at the junction with the third arm (207) is fixedly mounted with a gear one (3), and the outer wall of the fourth arm (208) at the junction with the third arm (207) is also fixedly mounted with a gear one (3).

10. The image acquisition camera for deep learning of sports movements according to claim 7, characterized in that: Both ends of the interior of the machine arm three (207) are fixedly installed with an internal gas injector (4); the outer wall of one end of the internal gas injector (4) is rotatably assembled with a gear two (403), and the gear two (403) is arranged outside the machine arm three (207) and meshes with the corresponding gear one (3); a crankshaft (404) is rotatably installed inside the internal gas injector (4), and the crankshaft (404) is fixedly connected to the corresponding gear two (403); the end of the crankshaft (404) is rotatably connected with a rotating arm (405), and the end of the rotating arm (405) is connected with a piston (406), and the piston (406) is slidably installed inside the internal gas injector (4); An air intake pipe (401) and an air outlet pipe (402) are fixedly connected to the outer walls on both sides of one end of the inner air injector (4), respectively; the air intake pipe (401) and the air outlet pipe (402) both extend to the outside of the third arm (207), and the two air outlet pipes (402) are connected to the housing of the motion capture camera (1) together.

Images