Research Of Character Recognition Technology Based On The Thought Of Sequential Similarity Detection Algorithm

Character recognition technology has been applied widely in recent years. There have been many kinds of recognition methods. Among them, Shape Context algorithm has efficient recognition rate. It has been widely used in image recognition technology since it was been proposed in2000. However Shape Context algorithm requires a lot of character information, therefore, it has low the processing speed. On this, this thesis makes some improvements on the processing speed of Shape Context and introduces Sequential Similarity Detection Algorithm which reduces processing time evidently on the unmatched characters. First, it needs to choose an experiential threshold in training, and then a character matches a sample in a sample library, if the cumulative error is greater than this experiential threshold, the algorithm throws that sample away and calculates the next sample. If the experiential threshold is not suitable, it needs to train again to find a new threshold. The recognition rate being kept constant, it reduces to processing time on the unmatched characters to improve the processing speed in the whole process. According to processing points in training, We calculated that the overall processing time reduced by around65%.