The Research And Implement Of License Plate Recognition System Based On Android Platform

As the core of the intelligent transportation system, license plate recognition system has become an important issue in the field of the traffic system presently, which also has a broad prospect of application. With the official launch of wireless city construction, the wireless intelligent transportation system provides a new opportunity for the governance of traffic jams. Under the background, the license plate recognition mobility is also put on the agenda.Through the analysis of the traditional license plate recognition algorithm advantages and limitations, the thesis propose a license plate recognition solution which is suitable for the mobile platform, and make it to be implemented on the Android platform. Meanwhile, the thesis has made a thorough study on the following problems, such as blur, uneven illumination and low-resolution etc.The main points are as follows:(1)Camera shake during exposure leads to image blur which belongs to motion blur. At first, the thesis studies the degradation model of the motion blur image, makes analysis of the limitation that the wiener filtering is sensitive to parameters in the process of image deblur. Through the experiment, founding that there is a mapping between the blur length of Point Spread Function and the fuzzy degree, the thesis puts forward a new adaptive iteration blind deconvolution deblur algorithm.(2)The illumination variation as well as shooting angle and other factors, which often cause the uneven illumination of license plate image, even large areas of shadow, makes a lot of interference to image binarization, and causes the fracture or adhesion of the license plate character after binarization. Therefore, the thesis uses DCT transform to achieve illumination compensation of license plate gray image, which can overcome the impact of uneven illumination effectively.(3)The license plate images that captured by mobile phone are always with low resolution, which makes the problem of the high misidentification rate in ambiguous characters recognition. The thesis summarizes and classifies the ambiguous characters which found in experiment, proposes the ideas of recognising characters secondary by using discriminative local features.