License Plate Character Recognition Research Based on Shape Context

Classic shape context algorithm uses the correspondence points between contour points. Those points represent the outline of a shape characteristic. Select a different position and number of sampling points to produce different effect on the similar shape feature description. Uniform random sampling algorithm can not solve the problem of selective retention for classic shape context with similar characters discrimination of contour points. Improved Freeman chain code algorithm describe the edge profile, and the code value controls the location of sampling points. Experimental results show the proposed method has better effect on retaining the similar characters outline key points. Using shape context features can be accurate license plate character recognition.