The method based on boundary chain-code for objects recognition and gesture analysis

摘要

Chain-code has been shown to be efficient and effective for describing objects in image. Based on the boundary chain-code, we present a method for object recognition. Firstly, we propose a novel method to form a closed contour and convert the chain-code into an ordered array and then transfer it to the frequency domain. Changing the start point of the chain-code, it shows that the amplitude distributions in the frequency domain is kept invariant. Secondly, the features of the amplitude distribution with the changing boundary scale and the adding noises are introduced. Finally, the experiment results have been given to illustrate the validity on the object recognition with the character of the amplitude distributions, while altering the start point of the chain-code and changing the object contour slightly. In addition, the gesture of human motion is analyzed.