Research Of Counting Method For Card Based On Digital Image Process

With the development of society and improvement of information technology, thedigital card has gained widespread popularity and application, so the problem of cardcounting is becoming more and more important. Most of the works have been done bymanual mode in present stage with cumbersome procedure time consuming, and therecognition results maybe not accurate by the reason of lone time repetitive work thatwill cause visual fatigue and distraction. Taking this as the background, this paper hasstudied card counting methods based on digital image processing techniques, andintroduced two kinds of different counting methods according to the feature of image.The realization of algorithm in this paper is based on the edge of card, in thebasement of completing image preprocessing, taking the texture feature of image intoaccount, we use the sobel arithmetic operators to extract the edge information. Then westudied the counting methods based on the edge information, introduced segmentationcounting method and projection transformation counting method. The segmentationmethod convert gray image to binary image first,then use mathematical morphologymeans to get rid of interfere information to make the count results accurate. How ever,the projection transformation method do not need to convert gray image to binary image,the Radon transformation is performed directly and the optimal projection vector issingled out, so we convert the edge counting problem to the problem of countingnumbers of pulse in projection histogram, combine the periodicity of pulse, we can getthe result by seeking local maximum value.Different kinds of samples are input to the computer to processing and therecognition result is output. The segmentation method can acquire high accuracy rateform the analysis of test result, but can easily affected by factors such as illumination、material quality of card、noise and so on. The projection transformation method showsstronger robustness and counting accuracy, not only can get higher accuracy rate andmore stable result than segmentation method, but also shows significant advantage intime complexity.