Reconstruction of Missing Markers in Motion Capture Based on Deep Learning

Abstract

With the great success of the movie Avatar, optical motion capture systems have been widely used in the fields of virtual reality, movie animation and robotics. However, the optical motion capture system is prone to noise due to the occlusion of the markers during the capture process. In order to remove the noise in the data, commercial methods are to provide manual repair for noisy data, and use interpolation to fill in the missing data, which is time-consuming and labour-intensive to process and the repaired data is not smooth enough to be jittery. In this paper, we propose a denoising method that can intelligently detect noise in the data and reconstruct missing markers data without manual intervention. This method uses a deep learning method, based on the temporal and spatial relationship of motion sequences, to learn the logical relationship between the data, to quickly find the lost data and reconstruct it. Simulation proves that our method is efficient in reconstruct missing markers.