Research On Segmentation Of Cigarettes Image Based On Mathematical Morphology

Object recognition and counting according to image is a research direction in the field of image processing and recognition. In medium-sized or pint-sized cigarette factory, it is a necessary procedure between the making and packaging workshop to count cigarettes. So it is a feasible solution to adopt the image processing and recognition technology.Image segmentation is one of the most important steps before cigarette image recognition and counting. Its main target is to partition the image to several components which are highly related to the objects. The methods of segmentation can be classified to three classes according to the features: the first one is the global knowledge of related image or parts of image, this knowledge is generally expressed by the histogram of image features; the second one is edge-based segmentation; and the third one is region-based segmentation. The segmentation meaning to cigarette image is to distill the section of single cigarette from image, and to make it easy for counting and recognition. There are two steps of cigarette image segmentation. The fist step is preprocessing segmentation which contains eliminate the frame of cigarettes container, color image threshold and noise filtering. Color image threshold makes use of the color difference between cigarette section and background to differentiate the objects and the background. Noise filtering designs the morphological structure elements to filter cigarette holes and isolated points in the binary image.The second step is cigarette section segmentation which is aim at overlapping cigarettes. Just because of cigarette image's obvious geometrical features and morphology pays attention to the features of basic graph in the image, three algorithms are proposed. The first one is the algorithm of glide window whose essence is a method of matching. The second one is the algorithm of shape-classfy. The third one is the algorithm of watershed. Accordingly the result of three algorithm's experiment is given. In the whole course of morphological processing, a concept named feature code is introduced.On the one hand, feature code converts the morphological operations to the bit operation of bytes which can quicken the speed of algorithm. On the other hand, it can describe the characteristic of boundary and some geometry feature of the region, so this new region descriptor value is worthy to further research.In order to evaluate the veracity of segmentation, the counting results obtained by the algorithm are compared with the real counting results of cigarettes.