Research On Moving People Segmentation In The Complex Situation

Segmentation of moving targets is applied widely in image processing, computer vision and so on. And the segmentation of moving people is the main base of recognizing and understanding people's behaviors by making use of computers. The accuracy of segmentation which would directly influence on the following process is very meaningful. Segmentation of moving people has been applied into many fields, such as video surveillance, virtual reality, video conference, etc. Researchers did much research on segmentation of moving people. But most of the algorithms are limited in particular situations, there is not a universal method that could be used to segment the moving people from the complex surroundings efficiently.In this paper, two methods are used to extract the moving target. First, when the background is already known, the method of background subtraction will be used, otherwise, the frame difference method will be used. An improved frame difference method is proposed. Based on local chroma coherency, a new extracting method of moving targets is presented. Otherwise, the background is established by an improved frame difference method: compared to traditional frame difference method, the change detection mask (CDM) is obtained by adopting auto threshold, and the moving target region is confirmed by making use of the external polygon of CDM. The proposed method in this paper improves the accuracy of the established background. In order to remove the shadow of the moving target, the paper combines the information of space position to improve the shadow detection based on intensity and chroma. In the end, an adaptive method is presented to segment the occluded people: by making use of forecasting based on the Kalman Filter, the extraction of face skin color based on chroma and head region matching based on the least mean-square error(MSE) function, the occluded people can be segmented adaptively.The algorithms proposed in this paper can segment the moving people from the complex situations in principle, it mainly resolve the segmentation of the moving target, background regeneration in the condition of abruptly change in illumination and so on,...