A New Method to Detect the License Plate in Dynamic Scene

License plate detection includes license plate positioning, segmentation characters, character recognition. The recognition rate of license plates under dynamic scenes is affected by many factors. Each process deviation may affect the overall system recognition rate, and the accuracy of each part is affected by many factors, in order to reduce this error, we combine the advantages of a variety of algorithms to propose a comprehensive detection model. In the license plate positioning phase, we propose HSV space and morphological methods; in the segmentation character phase, we propose the maximum adjacent character horizontal center distance segmentation method; in the character recognition stage, we choose to use the CNN algorithm. In the final simulation test, there are a set of 1 errors in the 30 groups of license plate recognition, the accuracy is higher.