Strategy for searching cigarette center based on template distance distribution
CAO Jichang, DENG Zhen(College of Computer Sci.& Tech., Huazhong Univ. of Sci.& Tech., Wuhan 430074, China)
Based on the thought of recognizing and counting cigarettes with template matching method, a method for optimizing the strategy searching for the cigarettes center in cigarette images is presented. Firstly, the distance distribution of the cigarette template center is analyzed based on the characteristics of cigarette images. Secondly, the possible boundary of cigarette is identified in a rough way and the distance between the non-boundary pixel and its closest boundary is obtained.It is combined with the template distance distribution and six distance distributions of pixels in the cigarette image are deduced. So the pixels of cigarette image can be classified to several parts and endowed with different identification priorities according to these six distance distribution. At last, the cigarette centers can be searched according to these priorities. The experiment shows that this method is efficient to locate both the center of isolated cigarette and the centers of connective cigarettes. Compared with the method of searching every pixel one by one without any condition, it can improve the accuracy of the cigarette recognition and counting.