Research On Object Tracking Algorithm In Human-Computer Interaction System

In recent years, with the development of virtual reality technology, the natural, portable and no invasive way has become the developmental direction of the human-computer interaction system. The aim of this paper is to provide a natural, portable and no invasive way of human-computer interaction based on the ordinary paper by object tracking technology. Because the ordinary paper lacked significant texture features, the traditional tracking algorithm is difficult to complete the robustness of tracking: inaccuracies or failure. It can greatly impact the user experience in the human-computer interaction. This paper mainly studies the real-time and robust tracking methods of the ordinary paper.Because the efficient and accurate paper detection is the foundation of real-time and robust tracking, this paper proposes a paper detection algorithm based on LSD algorithm. To different types of paper as well as in the case of paper with partial occlusion, robustness is very good in experiments. In comparison with the paper detection algorithm based on Hough, the proposed approach takes an average time of 24.1ms, while the algorithm based on Hough an average time of 35.7ms in our test cases. It shows that the proposed algorithm is efficient.This paper proposes an ordinary paper tracking algorithm by combining the paper detection algorithm based on LSD and compressive tracking algorithm. The algorithm first uses the positive and negative samples computed by the result of paper detection to update classifier for tracking, then detects the paper by tracking result. In large areas of occlusion, due to the context, the relationship between the frame and the frame can be used according to estimate the location of the paper. So the algorithm does not fail, which is that the robustness is very good. Because the proposed paper detection algorithm based on LSD and the compressive tracking algorithm both are very efficient. And the algorithm only do small range detection, it reduces the time complexity. So the efficiency of the ordinary paper tracking algorithm proposed in this paper is very high(in this paper, the test results show that can achieve 52 frames per second).In conclusion, the ordinary paper tracking algorithm proposed in this paper is not only robust but also efficient. Finally, through by the test of the human-computer interaction system composed of Kinect and mini projector, simulation result shows that the algorithm can satisfy the actual requirements of the human-computer interaction system.