Human Body Motion Tracking System For Natural Interaction Of Electronic Sand Table

Electronic sand table utilizes computer to simulate the real model.It is an important tool for commander to research terrain,enemy's situation and operational plan.The interaction mode of electronic sand table directly affects the efficiency of the command.The application of natural interaction techniques such as motion recognition and head tracking in the interaction of electronic sand table,which allows commanders to control the operable objects through simple body movements.This approach improves the naturalness of the interaction and the efficiency of command.In order to realize the interaction methods such as motion recognition and head tracking,this paper studies the optical motion capture technology and human motion recognition technology.The research mainly includes marker tracking,calculation of body motion parameters and motion recognition for natural interaction of electronic sand table.To solve the problem of occlusion and confusion in optical motion capture system,a continuous marker tracking method based on multiple constraint conditions is proposed.The method effectively eliminates the confusing points according to the spatiotemporal continuity,smoothness and local rigidity of markers on the same segment limb,which ensures the correctness of marker tracking.And the method can predict missing markers' location according to the visible marker on the same limb and the local rigidity and spatiotemporal continuity.In order to rediscover the missing markers,a method combining the structural characteristics of markers and physiological constrains of human skeleton is proposed.This improves the robustness of marker tracking.The experiment shows that this method can effectively track markers.Then,we propose a method of joint center calculation.We use multi group static posture to calculate accurate joint data,and establish a marker-joint relationship model based on the relative position invariance of markers and joints.According to the relationship model,we can calculate the joint data in real time.The experiment shows that this method can be used to calculate human motion parameters accurately.Finally,we collect data samples of interactive action of the electronic sand table through the human motion tracking module,and use multi-classification support vector machine algorithm to identify human behaviors.Then we combine the operator's action behavior with the region of interest to obtain the interactive semantics,and the system completes the corresponding task to realize the semantic interaction of the whole sand table.The research results of this paper are verified in the simulation environment of electronic sand table.