Variable-Length License Plate Character Segmentation

In recent years, intelligent transportation system (ITS) is one of the most compelling and the fastest developing speed areas. Vehicle plate recognition system is a developed high-tech one coming after ITS. Now vehicle license plates recognition system has been widely applied in bridge and intersection automatic unattended parking fees, automatic recording offending vehicles, etc. With the development of economy, more and more private cars appear in our life, accordingly more traffic accidents or illegal driving happen. Traffic safety and effective traffic management are indispensable. License plate recognition system will play a more and more important role in our life.License plate identification technique refers to some fields including mainly digital image processing, computer vision and pattern recognition. Overall, license plate identification technique is the combination of image processing technology and characteristics of the license itself and the effective use of neural network knowledge. The raw data of License plate identification technique is images with complex background from video tape. So the license plate recognition system mainly includes three parts-license plate location, license plate character segmentation and character recognition. License plate location part extracts license plate image from the background picture license plate images, and then license plate character segmentation part will divided them into single characters, finally, character recognition part will do the job of single characters'identification.License plate segmentation is a very important part of plate recognition system. Effective character identification is dependent on segmentation part. Based on the thorough research of traditional algorithms for license plate segmentation, I suggest an algorithm of length-variable length character segmentation which is used to split uncertain character number license plates. My main job is as follows:First:I do some research about license plate image preprocessing technologies such as binary, tilt correction, removing the borders. Second:I do a thorough research on license plate segmentation algorithms about Chinese license plate and do a comparison of these algorithms, their advantages and disadvantages. And I came up with an algorithm based on projection.Third:I came up with an algorithm about variable-length license plate segmentation based on the study of other algorithms'principles, advantages and disadvantages. Algorithm based on existing projection algorithm, and puts forward some improvement. I do some preprocessing to vehicle license plate before segmentation part such as binary, tilt correction, removing the borders. Then use connecting area algorithm to find the top and the bottom of the license plate image, and remove other useless parts leaving character part only. Finally use vertical projection algorithm to cut the license plate image into single character parts.