A new object detection method in color image processing
Yanjie Zhang, Xinghua Li · 2011
In this paper, a new object detection method is proposed for complex environment, and it is very important for target tracking, activity recognition and behavior understanding. The key idea of this research is selecting competent color features, and combining them with background subtraction method. First of all, the features used to segment the image should have these characteristics that specifically, integrity and sensitivity. After multilateral research, the hue-intensity feature is extracted from HSI color space, and the regular which the color feature changes with time and the lighting density is researched. It solves the problem that reflecting correctly the difference of the similar objects, which is a challenge for the past method. For another thing, object detection approach should be robust extremely, in order to adapt to the situations that illumination changes, moving disturbances and so on. The background subtraction method is adopted to overcome these problems after selecting appropriate background model and the strategy that used to update the model. The experiment results demonstrated that the proposed method constructed dependable background model, and obtained the accurate target area no matter the difference between foreground and background is remarkable or unobvious.