Tracking Technology Research, Based On The Vehicle License Plate Recognition Video

Intelligent Transportation System is a leading-edge research of the transportation in the world. One of its key technologies is vehicle tracking. As a sign of vehicle's identity, license plate is unique. In this paper, license plate recognition technology is applied to vehicle tracking. And the video tracking of vehicles based on license plate recognition is studied. Filming in certain circumstances, vehicle tracking is achieved by tracking license plate of the vehicle. The video tracking technology of vehicles based on license plate recognition has wide application prospect in highway toll collection system, road traffic monitoring systems and statistics, video speed measuring, car automatic driving, traffic control and other fields.Firstly, this paper studies the process of license plate location of static car images. Through improving the algorithms of the existing license plate location, sciagraphy based on edge detection combined with a priori knowledge of the license plate is used to locate the license plate.Secondly, considering the disadvantages of traditional vertical segmention methods, an improved vertical segmention method is used to segment the character of license plate in this paper. This method can reduce the effect of the license plate frame, space mark and noises in the images while segmenting the characters. In order to facilitate the subsequent recognition, the characters were normalized after segmention.Because the BP algorithm for training multilayered feedforward neural network always suffers from slow convergent rate and local convergence, a learning algorithm based on the extended Kalman filter for training a neural network is proposed to apply to the character recognition of license plate. The simulation results show the effectiveness of this method. While training the neural network of characters, the letters and numbers are divided into two parts by the idea of closed loop. Through the trained neural network, the normalized characters of the license plate can be recognized correctly.Finally, the technology of license plate recognition based on the static images is applied to the dynamic video images. Kalman filter theory is used to predict the position of the license plate. So the search window is the neighborhood of the predicted position. The license plate is located and recognized by the algorithm which is used in the static car images. Thus, tracking the license plate of the moving vehicles in complex dynamic background is achieved.