Moving object tracking using particle filter and observational model based on multi-feature composition

Moving object tracking in a sequence of an image is one of the favorable issues in machine vision. Recently, particle filter have based developed as a powerful method in this field. Particle filter is a following method which estimates a target route in video image sequences by probability approaches. In many cases, the follower encounters with problems such as: local lighting variations or abrupt movements. In most of these cases, target tracking is missing, so an appropriate filter coupled observational model is required to improve follower performance and increase the efficiency. An observational model is used in order to improve filter performance. However, this color feature based model contains less computational volume and is more rapid but does not provide good performance at the presence of background color or some objects with similar color. In this paper, an observational model is proposed that performs using particle filter accompanied by mean shift algorithm based on incorporated color and edge features. The results show that the introduced method is not sensitive to color and intensity change while it also has a good performance.