License Plate Location And Character Segmentation Algorithm To Achieve

Along with the coming of the Information age, our life become more and more convenient, at the same time there is more and more vehicle. So it is urgent for us to study an efficient method to manage them. Intelligent Transportation System emerges in this condition, and the key technology of this system is License Plate Recognition System (LPRS). This article studies the algorithm of license plate location and the algorithm of license plate character segmentation in details. LPRS is widely used in the monitoring of vehicle on highway, the monitoring of vehicle at cross-road and vehicle management in district. The system can work in day and night efficiently, which saves labors and resources.LPRS is consisting of hardware and software, and this article studies the algorithm of license plate location and the algorithm of license plate character segmentation in LPRS. The algorithm is based on pictures obtained at one cross-road in Mianyang. For the reason that the tested pictures are obtained on roadway, the picture are accordant with what we will meet when the system works. So the algorithm is not based on ideal pictures but based on normal pictures. The license plates on roadway have different characteristics as the weather, illumination change, and they are dirty and old, all of this increase the difficulty of the system algorithm. The algorithm described in this article take all conditions into consideration. LPRS includes three steps: Locating License plate, Segmenting Characters, Recognizing Characters. And this article studies the algorithms of the former two steps.a. License plate location algorithm. License Plate location algorithm is the foundation of the whole system algorithm. Wavelet transform is adopted in our algorithm. Firstly we can obtain high frequency coefficient after the transformation and form the coefficient into image data. Then we dispose the image data with median filter, binarization and morphology operations. We label and record the area appears in the image. Then we can obtain license plate candidates from the area according to the area information.b. Character segmentation algorithm. Character segmentation requires us segment every character out from the license plate preparing for character recognition. When we get a license plate candidate we detect its lean angle using Hough transform and then we rotate it to horizontal. After that we remove the border of the license plate including the up-down border and right-left border. So far, we get an accurate candidate, then we segment characters based on characteristic of bimodal curve.The algorithm described in this article is programmed in C, and tested on pictures obtained at one cross-road in Mianyang. From the test results we know that the accuracy rate of license plate location is 96.9% in daytime and 97.5% at night, and the accuracy rate of license plate segmentation is 96.7% in daytime and 97.4% at night.