Search

CN-121999512-A - Limb key point recognition device and method thereof

CN121999512ACN 121999512 ACN121999512 ACN 121999512ACN-121999512-A

Abstract

A limb key point identification device and a method thereof are provided, wherein the limb key point identification device comprises a camera and an arithmetic circuit. The camera is used for shooting image data. The image data includes a plurality of frames. The computing circuit receives the plurality of frames from the camera to perform a first operation on a first frame of the plurality of frames, wherein the first operation locates a human body in the first frame with a plurality of keypoints, and to perform a second operation on the current frame in response to the current frame of the plurality of frames being not the first frame and having a confidence level higher than a confidence level, wherein the second operation tracks the plurality of keypoints of the current frame and adjusts the confidence level based on the plurality of keypoints of the current frame.

Inventors

  • LIN HONGXUAN
  • Weng Ruiqun
  • LI YUZHE

Assignees

  • 台达电子工业股份有限公司

Dates

Publication Date
20260508
Application Date
20241105

Claims (16)

  1. 1. A limb keypoint identification device comprising: a camera for capturing image data, wherein the image data comprises a plurality of frames, and An arithmetic circuit receiving the plurality of frames from the camera for: Performing a first operation on a first frame of the plurality of frames, wherein the first operation locates a human body in the first frame using a plurality of keypoints; In response to a current frame of the plurality of frames being not the first frame and having a confidence level greater than a confidence level, performing a second operation on the current frame, wherein the second operation tracks the plurality of keypoints of the current frame and adjusts the confidence level according to the plurality of keypoints of the current frame.
  2. 2. The limb keypoint identification device of claim 1, wherein the arithmetic circuit performs the first operation on the current frame in response to the confidence being not higher than the confidence criterion.
  3. 3. The limb keypoint identification device of claim 1, wherein the second operation further comprises: Comparing the plurality of key points of the current frame with the plurality of key points of a frame previous to the current frame to calculate a difference value, and Subtracting the difference from the confidence level to adjust the confidence level.
  4. 4. The limb keypoint identification device of claim 1, wherein the operation circuit further calculates an amount of displacement between one of the plurality of keypoints of a frame previous to the current frame and a corresponding one of the plurality of keypoints of the current frame to adjust the confidence.
  5. 5. The limb keypoint identification device of claim 1, wherein the arithmetic circuit is further to: Selecting a first line according to the plurality of key points of the current frame; Selecting a second line from a frame previous to the current frame corresponding to the first line; measuring the first line to calculate a length; comparing the first line with the second line to calculate the length variation, and And adjusting the confidence according to at least one of the length and the length variation.
  6. 6. The limb keypoint identification device of claim 1, wherein the arithmetic circuit is further to: Selecting a first line according to the plurality of key points of the current frame; Selecting a second line from a frame previous to the current frame corresponding to the first line; selecting a third line according to the plurality of key points of the current frame; Selecting a fourth line from a frame previous to the current frame corresponding to the third line; comparing the length ratios of the first and third wires with the length ratios of the second and fourth wires to calculate a length ratio variation, and The confidence is adjusted according to at least one of the length ratios of the first and third lines and the length ratio variation.
  7. 7. The limb keypoint identification device of claim 1, wherein the arithmetic circuit is further to: Selecting a first line according to the plurality of key points of the current frame; Selecting a second line from a frame previous to the current frame corresponding to the first line; selecting a third line according to the plurality of key points of the current frame; Selecting a fourth line from a frame previous to the current frame corresponding to the third line; comparing the angles of the first line and the third line with the angles of the second line and the fourth line to calculate an angle change amount, and And adjusting the confidence according to the angle change amount.
  8. 8. The limb keypoint identification device of claim 1, wherein the arithmetic circuit is further to: Selecting a first line according to the plurality of key points of the current frame; Selecting a second line from a frame previous to the current frame corresponding to the first line; selecting a third line according to the plurality of key points of the current frame; Selecting a fourth line from a frame previous to the current frame corresponding to the third line; comparing the distances of the first line and the second line with the distances of the third line and the fourth line to calculate a distance variation amount, and And adjusting the confidence according to the distance variation.
  9. 9. A limb keypoint identification method comprising: Capturing image data, wherein the image data comprises a plurality of consecutive frames; Performing a first operation on a first frame of the plurality of frames, wherein the first operation locates a human body in the first frame using a plurality of keypoints; In response to a current frame of the plurality of frames being not the first frame and having a confidence level greater than a confidence level, performing a second operation on the current frame, wherein the second operation tracks the plurality of keypoints of the current frame and adjusts the confidence level according to the plurality of keypoints of the current frame.
  10. 10. The limb keypoint identification method of claim 9, wherein the first operation is performed on the current frame in response to the confidence being not higher than the confidence criterion.
  11. 11. The limb keypoint identification method of claim 9, wherein the second operation further comprises: Comparing the plurality of key points of the current frame with the plurality of key points of a frame previous to the current frame to calculate a difference value, and Subtracting the difference from the confidence level to adjust the confidence level.
  12. 12. The limb keypoint identification method of claim 9, further comprising: The displacement amount between one of the plurality of key points of the previous frame of the current frame and the corresponding one of the plurality of key points of the current frame is calculated to adjust the confidence.
  13. 13. The limb keypoint identification method of claim 9, further comprising: Selecting a first line according to the plurality of key points of the current frame; Selecting a second line from a frame previous to the current frame corresponding to the first line; measuring the first line to calculate a length; comparing the first line with the second line to calculate the length variation, and And adjusting the confidence according to at least one of the length and the length variation.
  14. 14. The limb keypoint identification method of claim 9, further comprising: Selecting a first line according to the plurality of key points of the current frame; Selecting a second line from a frame previous to the current frame corresponding to the first line; selecting a third line according to the plurality of key points of the current frame; Selecting a fourth line from a frame previous to the current frame corresponding to the third line; comparing the length ratios of the first and third wires with the length ratios of the second and fourth wires to calculate a length ratio variation, and And adjusting the confidence according to the length proportion variation.
  15. 15. The limb keypoint identification method of claim 9, further comprising: Selecting a first line according to the plurality of key points of the current frame; Selecting a second line from a frame previous to the current frame corresponding to the first line; selecting a third line according to the plurality of key points of the current frame; Selecting a fourth line from a frame previous to the current frame corresponding to the third line; comparing the angles of the first line and the third line with the angles of the second line and the fourth line to calculate an angle change amount, and And adjusting the confidence according to the angle change amount.
  16. 16. The limb keypoint identification method of claim 9, further comprising: Selecting a first line according to the plurality of key points of the current frame; Selecting a second line from a frame previous to the current frame corresponding to the first line; selecting a third line according to the plurality of key points of the current frame; Selecting a fourth line from a frame previous to the current frame corresponding to the third line; comparing the distances of the first line and the second line with the distances of the third line and the fourth line to calculate a distance variation amount, and And adjusting the confidence according to the distance variation.

Description

Limb key point recognition device and method thereof Technical Field The present invention relates to a device for identifying key points of a limb and a method for detecting the same, and more particularly, to a device for identifying key points of a limb and a method for detecting the same by using an artificial intelligent identification technology and an object tracking technology. Background In the prior art, artificial intelligence recognition techniques may be used to set critical points of a limb in a single image frame to accurately locate features of a human body in the image frame. If a segment of continuous image is identified by using the conventional artificial intelligence identification technology, only the segment of image can be identified frame by frame. Because the operation time required by the artificial intelligence recognition technology is long, the operation efficiency of the system is poor. How to maintain the technical advantages of accurately positioning the characteristic parts of the human body by the artificial intelligent recognition technology and improve the operation efficiency is an important subject to be processed by the person skilled in the art. Disclosure of Invention The present disclosure provides a limb key point recognition device. The device comprises a camera and an arithmetic circuit. The camera is used for shooting image data. The image data includes a plurality of frames. The computing circuit receives the plurality of frames from the camera to perform a first operation on a first frame of the plurality of frames, wherein the first operation locates a human body in the first frame with a plurality of keypoints, and to perform a second operation on the current frame in response to the current frame of the plurality of frames being not the first frame and having a confidence level higher than a confidence level, wherein the second operation tracks the plurality of keypoints of the current frame and adjusts the confidence level based on the plurality of keypoints of the current frame. The disclosure further provides a method for identifying key points of limbs. The method includes capturing image data, wherein the image data includes a plurality of frames in succession, performing a first operation on a first frame of the plurality of frames, wherein the first operation locates a human body in the first frame using a plurality of keypoints, and performing a second operation on the current frame in response to the current frame of the plurality of frames being not the first frame and having a confidence level above a confidence level, wherein the second operation tracks the plurality of keypoints of the current frame and adjusts the confidence level based on the plurality of keypoints of the current frame. The limb key point identification device of the invention uses an artificial intelligent identification technology in the first operation and an object tracking technology in the second operation, so that the invention combines the advantages of the two technologies, not only can set key points to accurately position the human body in the continuous images, but also can save the time required by operation. Drawings FIG. 1 is a schematic diagram of a limb key point recognition device according to an embodiment of the disclosure; FIG. 2 is a schematic diagram illustrating the ordering of a plurality of frames of image data according to one embodiment of the disclosure; FIG. 3A is a diagram illustrating a human body recognition operation according to an embodiment of the present disclosure; FIG. 3B is a diagram illustrating a human body recognition operation according to an embodiment of the present disclosure; Fig. 4 is a flowchart of a limb key point identification method according to an embodiment of the disclosure. Description of the reference numerals 100 Limb key point identification device 110 Video camera 120 Arithmetic circuit HBI_1 image data HB1 human body MHB 1, MHB 2 human body image AIM 1, AIM 2, AIM m, first operation frame OIM_ 1[1-OIM_1 [4], OIM_ 2[1-OIM_2 [3], OIM_m1-OIM_m5 ] of the second operation frame 310. 320 Human body recognition operation M1, M2: (frame) P0 to P32 key points L2_1 l2_2 l2_1, l2_2 l2_3, l2_4: wire (C) N distance threshold LN ratio threshold Limb key point identification method 400 S410, S420, S430, S440, S450 step Detailed Description Embodiments of the present disclosure will be described below with reference to the accompanying drawings. In the drawings, like reference numbers indicate identical or similar elements or method flows. Referring to fig. 1 and fig. 2, fig. 1 is a schematic diagram of a limb key point recognition device 100 according to an embodiment of the disclosure. Fig. 2 is a schematic diagram illustrating the ordering of a plurality of frames of the image data hbi_1 according to one embodiment of the present disclosure. In fig. 1, the limb key point recognition device 100 includes a camera 110 and an ar