License plate recognition technology plays a very important role in the field of intelligent transportation,but for different types of license plates,the recognition method and effect are also different.License plate recognition technology mainly includes three parts: license plate detection,license plate segmentation and license plate recognition.In the part of license plate detection,this thesis proposes a license plate location method based on dual probability density function.Firstly,the edge texture feature and the double color feature extraction method are used to extract the similar regions of the license plate,and then the tilt angle of all the similar regions of the license plate is corrected.Finally,according to the linear and global probability density function characteristics of the two line license plate,the double probability density function value is used as the threshold value to accurately locate the license plate from the similar area of the license plate.In the part of license plate segmentation,this thesis proposes a license plate segmentation method based on adaptive threshold.Firstly,the license plate image is transformed into the HSV color space,and the image is binarized according to the characteristics of the image in the HSV space.Then,the characters on the license plate number are segmented by the adaptive projection method proposed in this thesis.In the part of license plate recognition,according to the characteristics of double line license plate,the character recognition method based on convolution neural network is used to recognize the characters obtained from segmentation.Experimental results show that,according to the characteristics of non motor vehicle license plate,the proposed license plate recognition method can accurately recognize the license plate number in the natural state,which is suitable for the license plate recognition of all kinds of vehicles with double row license plate number. |