Multiple Vehicle License Plates Locating Techniques In Complex Backgrounds

With economic development and rising standard of living of our people, urban road traffic problem has become increasingly serious, urgent need to use modern technology to achieve intelligent management of road traffic. In this situation, the intelligent traffic system has been an unprecedented development, but is still a need for more in-depth study. In this paper, a modified threshold segmentation algorithm and a more systematic approach are proposed to settle the problems of the threshold segmentation and removing pseudo-license plates in multi-plates location.Through in-depth analysis of current vehicle license plate location, this paper uses of existing efficient methods in consideration of the time and consequents. First of all, by analyzing the RGB, HSI, gray space on the pros and cons of treatments, the gray space is a better choice to deal with the original image of the license plates, which is not only in time to meet the needs of real-time traffic, but is also able to meet the exact location of the multi-plates in the final Results. In the edge extraction, on the analysis of existing edge detection algorithms, this paper selected vertical edge detection method, which has advantages not only in time, and better reflects the characteristics of license plate characters. In the problems of threshold segmentation, this paper presents an improved gradient segment threshold algorithm by analyzing the Otsu, the mean method, the iterative algorithm. Input images are divided into two parts; the regions with multiple plates are binarized with the average of Otsu threshold and the interval gradient mean. This approach is robust to plate contamination and different location.In the multi-plates location, the problem of the residual pseudo-license plates is quite conspicuous. it not only increases the complexity of post-processing algorithms, and increased time cost. In order to remove the pseudo-license plates from candidates, first with the characteristics of the color and aspect ratio to remove part of the pseudo-plates, and then made use of the new algorithm by analyzing connected component analysis to remove the absolute pseudo-plates of the blue. Main connected components analysis is to make full use of gradient binary images of characters, that is not easy of cross and spacing abnormal in the connected domain, to remove blue pseudo-plates.Experimental results show that the proposed method is able to achieve recall rate of 97.3 % and precision rate of 88.5% after removing pseudo-plates of 92.5% which is comparable to existed techniques. In the Matlab environment, with the grammar of this article, the average processing time in the 0.8s / pieces around the plate to ensure a better real-time.