Edge Direction Dispersion and Its Application in Automatic Suitable-matching Area Selection

【摘要】：Automatic suitable-matching area selection is an important research in visual guidance,scene matching and local non-texture target tracking.Meanwhile,it is the major research subject in computer vision and accurate guidance.In this paper,a new characteristic quantity,i.e.,the edge direction dispersion was defined.Then the computation method,parameters selection and real-time realization of this characteristic quantity were investigated.Furthermore,the application of its characteristic in automatic suitable-matching area selection was studied.With a reasonable design for calculation precision,solving method for arctan(x) function in [-,+ ] could be simplified to linearization calculations in [0,0.25],then a fast algorithm for arctan(x) function was presented.Experimental results showed that the sub-image extracted by our fast algorithm had higher matching level and was suitable to be a matching template.All the calculations involved in the proposed matching algorithm could be done in one filed period and the real-time request could be satisfied.Moreover,this characteristic of edge direction could also be used to measure the comparability of template and real-time images,which could be widely used in automatic target recognition and target tracking.