Image Tracking Algorithm Of Human Motion Based On Inverse Kinematics

With the development of computer science and image processing technology, the visual tracking and analysis of human motion is gradually becoming a hot topic in the field of computer vision. Meanwhile the urgent need of related fields such as human-computer interaction and virtual reality also prompted its rapid development. The tracking algorithm of human motion based on Inverse Kinematics is a method of unmarked human motion tracking for monocular video based on the appearance model of human and the template matching technique. It calculates the appearance template of the major human joints on the2D image, and then calculates the3D coordinates of the corresponding joints by template matching. Finally we use the local optimization techniques to reconstruct the posture of human. It has the advantage of requiring no mark and easy-using, but it is time-consuming and not accurate. In this paper, we presented the corresponding method to improve the accuracy and speed of the tracking algorithm of human motion based on Inverse Kinematics.The main work of this dissertation is as follows:1. We improved the joint position prediction method and calculation method of the rotation angle of the tracking algorithm of human motion based on Inverse Kinematics. Pros and cons of the joint position prediction method will affect the efficiency and effectiveness of the tracking algorithm. The tracking algorithm of human motion based on Inverse Kinematics supposed that joints’moving mode is uniform. Meanwhile, the joint motion model is speed-varying and contains random in actual tracking process. It will lead to joints’local optimization slower. When predicted locations of joints are far from the correct values or joints correct position are not in optimizing region, which will lead that we can’t find the correct position of the joint. Therefore, we used the particle filter method to predict the joint position and improved the effectiveness of the algorithm.2. We improved the template matching method of the tracking algorithm of human motion based on Inverse Kinematics. Tracking algorithm of human motion based on Inverse Kinematics used histogram matching approach. When the joint is located at a few positions in the image, the gray histogram of the deformation appearance model of the joint is same. Because the histogram indicated statistics and it did not take into account the deformation of each pixel corresponding to the appearance of the model, the position of which determines the optimal matching may not be optimal matching position. We use the method of calculating the correlation coefficient of the two image blocks to assess their degree of similarity. It considered the similarity of two corresponding image blocks in pixels. Thereby it increased the effectiveness of the tracking algorithm.3. We presented a method to generate image sequence with the3dmax as the testing sample. The tracking results are compared with the real data obtained from the3dmax to illustrate the effectiveness of the proposed algorithm. The picture sequence generation process is based on moving target joint position to produce the body’s action sequences, thereby generating the desired video experimental data. Then, using the traditional algorithm with this algorithm for tracking moving targets in the generated video we estimate the effectiveness of our algorithm. Finally, by comparing the estimated value with the original real value and calculating the deviation of the pixel values of the image, we illustrate the effectiveness of the proposed algorithm.