火车票识别是实现自动检票的关键.采用最大方差阈值法对图像二值化求最佳阈值时,单向逐级遍历灰度级值,耗时较多.提出双向分组并行遍历的方法,从灰度级别值两端同时向中间,双向并行遍历,比较并计算出最合适的阈值.随机选取200幅火车票图像,分别使用指定阈值法、交互式阈值法和双向遍历最大方差法阈值法二值化处理,由BP神经网络识别,双向遍历最大方差阈值法比其他两种方法识别率分别提高了12.5％和5.5％,同时采用双向遍历比单向遍历求最佳阈值时,前者比后者在时间上缩短了一半,实验证明双向遍历最大方差阈值法是有效的.%Train tickets identification is the key to achieve automatic tickets checking. When seeking the optimal threshold of the image binarisation with maximum variance threshold algorithm, it is really time-consuming to undergo the unidirectional traversal on gray-scale values step by step. The thesis presents a bidirectional-grouping parallel traversing approach, it makes bidirectional-parallel traversal from two ends of gray-scale grade value to the middle simultaneously, and compares and calculates the threshold fitting best. 200 images of train tickets are selected in random, and are processed respectively by the specified threshold method, the interactive threshold method and the bidirectional traversal maximum variance threshold method. The processed images are identified with BP neural network, and the third method has higher recognition rate than the other two by increasing 12.5% and 5.5% respectively. Meanwhile, the method of bidirectional-parallel spends less time than the unidirectional one in seeking optimal threshold. Experimental results show that the proposed method is the most efficient one.
展开▼
