Research on Animation Character Motion Capture System Based on Computer Vision

This research is dedicated to the development of an animation character motion capture system predicated on computer vision. In the initial stage of system construction, multiple cameras are employed synchronously to collect the motion frames of animation characters, ensuring the acquisition of omnidirectional information. During the feature extraction phase, the improved Harris corner detection algorithm is utilized. By constructing a 2×2 matrix to calculate the autocorrelation function, the feature points of key parts such as the character’s contours and joints can be precisely located. Compared with traditional algorithms, the precision of feature point extraction has been enhanced by approximately 15%. Based on the extracted feature points, the Lucas-Kanade optical flow algorithm is adopted to track their motions. By constructing the optical flow constraint equations and iteratively solving the velocity fields of the feature points, the restoration of motion trajectories is achieved. Verified by a plethora of experiments, for complex motion sequences, the capture accuracy of this system can reach as high as 90%, and the average processing time per frame has been shortened to within 30 milliseconds, effectively meeting the requirements of real-time animation production and providing solid technical support for the efficient creation of the animation industry.