Research Of License Plate Recognition System On Vehicle PTZ Camera

The Vehicle License Plate Recognition System (LPR) is one of the basic contents of the Intelligent Transport System(ITS). The aim of the thesis is to set up a mobile license plate recognition system by adding the function of license plate recognition to the vehicle PTZ camera which is researched and developed by ourselves. Compared with the traditional fixed license plate recognition system, It is more flexible, It can be a complement to the fixed license plate recognition system which has been widely used.Based on the mobility of the vehicle PTZ camera, this thesis studies the following aspects:vehicle detection, license plate location, license plate slant correction and character segmentation, license plate character recognition, etc. Specific works of this thesis are as follows:(1) The hardware of the license plate recognition system is introduced and the function of the vehicle PTZ camera and other hardware of this system are analyzed. It lays the foundation for the software algorithms of this system.(2) Several common methods of moving objects detection are studied and analyzed. This thesis discusses two working modes of this system, the camera is still and the camera is moving. A method of frame image difference of the set dummy loop is put forward when the camera is still. Experiment results show that this method is simple and effective.(3) Several common methods of license plate location are studied and analyzed. Then this thesis presents an effective license plate location algorithm based on improved haar-like features and Adaboost classifier. New haar-like features are designed according to the features of the license plate area. An Adaboost classifier is trained by using large numbers of positive and negative training samples. Using integral figure to accelerate the feature extraction.(4) Several common methods of license plate slant correction and character segmentation technology are studied and analyzed. Then this thesis puts forward a new method of license plate slant correction based on PCA of vertical edges of the characters and a new method of character segmentation based on the features of vehicle license plate and connected component analysis.(5) Several common methods of license plate character recognition are studied and analyzed. Three BP neural network Classifiers are trained to recognize characters of the vehicle plate by extracting Gabor feature and rough grid of the characters.(6) Ultimately, by qualitative analysis of the experiment results, this thesis points out some gaps and areas for improvement and makes sure the next work.