The Research Of Human Pose Estimation About BPOF Feature And Optimization Random Forest

With increasingly development of information technology and image processing technology, human pose estimation has become an important research direction in the field of computer vision, it has a very wide range of applications, including virtual reality, intelligent monitoring and robot motion control, and other fields. Depth camera makes the depth information in the application of human pose estimation more and more be taken seriously. Random forests is a kind of classifier based on statistical learning theory, it is applied to the human body pose estimation research work because of its high predict rate, small excessive fitting and the advantages of good tolerance of unusually.For feature extraction in human pose estimation, we propose an improved method based on binary phase-only filter(BPOF) algorithm, first of all, we calculate the scanning line length value from eight directions of every pixel in the image, then put the eight value into the BPOF algorithm to get the feature of the pixel, we also optimize the selsction method of random decision tree which is used to classify,finally we can estimate the human pose. The improved method has made a very big enhancement in recognition rate and robustness, at the same time, the optimized selsction method of random decision tree reduce the algorithm system time consumption.At the same time,the human pose estimation system which using the random forest as classifier has a problem about taking up too big memory footprint, so this paper puts forward an optimization random forest model to solve the problem above. The new model introduces the Poisson process and combines it with the depth information to form a filter before Bootstrap sampling, and then filter the original training dataset, moving the pixel sample which not play a positive role away. After that the goal of refactor the training dataset is achieved. So the insufficient about repeated sampling and the weak representative of random forest can be improved. And the experimental results show this optimization is effective, reducing the time and space complexity of the system, and make the system more general.