CN-121982493-A - Method, device, medium and equipment for testing unmarked visual dynamic capture system

CN121982493ACN 121982493 ACN121982493 ACN 121982493ACN-121982493-A

Abstract

The application relates to the technical field of computer vision, and particularly provides a method, a device, a medium and equipment for testing a non-marking visual dynamic capturing system, wherein the method can comprise the steps of acquiring first dynamic capturing data under an infrared dynamic capturing system and acquiring second dynamic capturing data under the non-marking visual dynamic capturing system in the same dynamic capturing area; the method comprises the steps of aligning joint key points in a key point corresponding table of first dynamic capturing data and second dynamic capturing data to obtain aligned key data, calculating a test index based on the aligned key data, wherein the test index is used for representing the precision of the unmarked visual dynamic capturing system, the test index comprises joint position errors and/or key point detection accuracy, the precision is in negative correlation with the joint position errors, and the precision is in positive correlation with the key point detection accuracy. The embodiment of the application can realize the accurate and effective test of the unmarked visual dynamic capturing system.

Inventors

ZHAO GANGQIANG
Xia Xile
WANG XIN
LIU YAODONG
LIN ZHONGYAN

Assignees

北京虚拟动点科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260122

Claims (12)

1. A method for testing a markerless visual dynamic capture system, comprising: in the same dynamic capturing area, acquiring first dynamic capturing data under an infrared dynamic capturing system and acquiring second dynamic capturing data under a non-marking visual dynamic capturing system, wherein the first dynamic capturing data comprises first rigid body data and first action data; aligning the joint key points in the key point corresponding table of the first dynamic capturing data and the second dynamic capturing data to obtain aligned key data; Calculating a test index based on the aligned key data, wherein the test index is used for representing the precision of the unmarked visual dynamic capture system, the test index comprises joint position errors and/or key point detection accuracy, the precision is in negative correlation with the joint position errors, and the precision is in positive correlation with the key point detection accuracy.
2. The method of claim 1, wherein the acquiring the first dynamic capture data under the infrared dynamic capture system comprises: after the calibration of the infrared dynamic capturing system is completed, acquiring the first rigid body data of the rigid body created under the coordinate system corresponding to the infrared dynamic capturing system; And the first action data of the dynamic capturing person is derived from the video data of the dynamic capturing person.
3. The method of claim 1 or 2, wherein the acquiring second dynamic capture data under the markerless visual dynamic capture system comprises: Determining a ground calibration area by reading a calibration file of the non-marking visual dynamic capturing system; Acquiring the second rigid body data of a plurality of rigid body mark points under the unmarked visual dynamic capturing system in the ground calibration area; and extracting the second action data from the video data of the capturing person, wherein the first action data and the second action data are acquired based on the video data of the same frame.
4. The method of claim 1 or 2, wherein before aligning the joint keypoints in the keypoint correspondence table of the first dynamic capture data and the second dynamic capture data to obtain aligned keypoint data, the method further comprises: Setting the acquisition frame rate of the infrared dynamic capturing system and the unmarked visual dynamic capturing system; acquiring motion data with the same frame number from the first motion data and the second motion data based on the acquisition frame rate; and extracting the joint key points in the action data with the same frame number, and constructing the key point corresponding table.
5. The method as claimed in claim 1 or 2, wherein said aligning the joint keypoints in the keypoint correspondence table of the first dynamic capture data and the second dynamic capture data to obtain aligned key data includes: Determining a coordinate alignment matrix and a time pair Ji Canliang between the infrared dynamic capture system and the label-free visual dynamic capture system; And aligning the joint key points by using the coordinate alignment matrix and the time pair Ji Canliang to obtain aligned key data.
6. The method of claim 5, wherein the coordinate alignment matrix is obtained by: Acquiring a first position coordinate and a second position coordinate of the same rigid body mark point in a coordinate system of the infrared dynamic capturing system and a coordinate system of the non-mark visual dynamic capturing system respectively; and determining a rotation matrix and a translation matrix in a coordinate alignment matrix through the first position coordinates and the second position coordinates.
7. The method of claim 5, wherein the time alignment parameter is obtained by: Slicing the first action data and the second action data based on a preset window and a window sliding range to obtain a plurality of data slices; Calculating a distance average value corresponding to the action offset in the plurality of data slices; And taking the action offset corresponding to the minimum value in the distance average value as the time pair Ji Canliang.
8. The method of claim 1 or 2, wherein the calculating a test index based on the post-alignment critical data comprises: Acquiring Euclidean distance of the joint key points in each frame of action in the aligned key data; solving the average value of Euclidean distances of all joint key points to obtain the average joint position of each frame of motion; and calculating the average value of the average joint positions of all frames to obtain the joint position error.
9. The method of claim 1 or 2, wherein the calculating a test index based on the post-alignment critical data comprises: acquiring a target distance of a predicted value and a reference value of each joint key point in the aligned key data; Confirming a target key point of which the target distance is smaller than or equal to a distance threshold value; and calculating the percentage between the number of the target key points and the total number of the joint key points to obtain the key point detection accuracy.
10. A device for testing a markerless visual dynamic capture system, comprising: The system comprises an acquisition module, a marking module and a marking module, wherein the acquisition module is used for acquiring first dynamic capturing data under an infrared dynamic capturing system and second dynamic capturing data under a non-marking visual dynamic capturing system in the same dynamic capturing area, wherein the first dynamic capturing data comprises first rigid body data and first action data; the alignment module is used for aligning the joint key points in the key point corresponding table of the first dynamic capturing data and the second dynamic capturing data to obtain aligned key data; The system comprises a test module, a test index and a detection module, wherein the test module is used for calculating a test index based on the aligned key data, the test index is used for representing the precision of the unmarked vision dynamic capture system, the test index comprises joint position errors and/or key point detection accuracy, the precision is in negative correlation with the joint position errors, and the precision is in positive correlation with the key point detection accuracy.
11. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program, wherein the computer program when run by a processor performs the method according to any of claims 1-9.
12. An electronic device comprising a memory, a processor, and a computer program stored on the memory and running on the processor, wherein the computer program when run by the processor performs the method of any one of claims 1-9.

Description

Method, device, medium and equipment for testing unmarked visual dynamic capture system Technical Field The application relates to the technical field of computer vision, in particular to a method, a device, a medium and equipment for testing a non-marking visual dynamic capture system. Background The motion capture (simply called "motion capture") technology is widely applied to the fields of movie and television animation, game development, virtual reality, sports analysis and the like, and is mainly divided into an infrared motion capture type and a visual motion capture type. The OptiTrack infrared dynamic capturing system is often used as an industry performance benchmark by virtue of the advantages of high precision and low delay, and the LydCap unmarked visual dynamic capturing system does not need to wear marked points, has the characteristics of convenience in use and strong scene adaptability, and cannot be compared with a benchmark system in standardized precision. Currently, in order to realize the precision test of LydCap unmarked vision dynamic capture system, a manual alignment mode is generally adopted, but the efficiency of the mode is lower, and the error fluctuation of the precision test is larger. Therefore, how to provide a method for testing an efficient label-free visual dynamic capture system becomes a technical problem to be solved. Disclosure of Invention The application aims to provide a method, a device, a medium and equipment for testing a non-marking visual dynamic capture system, and the technical scheme of the embodiment of the application can realize effective test of the precision of the non-marking visual dynamic capture system, and has high accuracy of a precision test result and higher test efficiency. In a first aspect, some embodiments of the present application provide a method for testing a non-marking visual dynamic capturing system, which includes acquiring first dynamic capturing data under an infrared dynamic capturing system and acquiring second dynamic capturing data under the non-marking visual dynamic capturing system in a same dynamic capturing area, wherein the first dynamic capturing data includes first rigid body data and first action data, the second dynamic capturing data includes second rigid body data and second action data, aligning joint key points in a key point corresponding table of the first dynamic capturing data and the second dynamic capturing data to obtain aligned key data, and calculating a test index based on the aligned key data, wherein the test index is used for representing accuracy of the non-marking visual dynamic capturing system, the test index includes joint position error and/or key point detection accuracy, the accuracy is inversely related to the joint position error, and the accuracy is inversely related to the key point detection accuracy. According to some embodiments of the application, first dynamic capturing data and second dynamic capturing data in an infrared dynamic capturing system and a non-marked visual dynamic capturing system are collected in the same dynamic capturing area, then joint key points in a key point corresponding table extracted from the two dynamic capturing data and the second dynamic capturing data are aligned to obtain aligned key data, and finally a test index is calculated based on the aligned key data so as to reflect the accuracy of the non-marked visual dynamic capturing system through the test index. The embodiment of the application can realize the effective test of the precision of the unmarked visual dynamic capture system, and has high accuracy of the precision test result and higher test efficiency. In some embodiments, the acquiring first motion capture data under the infrared motion capture system comprises acquiring the first rigid body data of a rigid body created under a coordinate system corresponding to the infrared motion capture system after the calibration of the infrared motion capture system is completed, and deriving the first motion data of the motion capture person from video data of the motion capture person. According to some embodiments of the application, after the infrared dynamic capture system is calibrated, the first rigid body data and the first action data under the corresponding coordinate system are collected, so that accurate collection of the dynamic capture data is realized. In some embodiments, the method for acquiring second motion capture data under the non-marked visual motion capture system comprises the steps of determining a ground calibration area by reading a calibration file of the non-marked visual motion capture system, acquiring the second rigid body data of a plurality of rigid body mark points under the non-marked visual motion capture system in the ground calibration area, and extracting the second motion data from video data of a motion capture person, wherein the first motion data and the second motion data are acquired based on the video