Research On License Plate Location And Character Segmentation Based On Fuzzy C-means Clustering

With the development of economy, a significant increase in demand and use of civilian car, which caused the problems of traffic congestion, has attracted people’s attention gradually. As an important means of monitoring traffic, car license plate recognition system has gradually developed into research focus. Thesis research on the basis of currently available, the relevant algorithms about license plate location and character segmentation are further research and improvement. The main research work is as follows:(1) License plate location algorithm based on mathematical morphology and characteristics of fusion has been researched. Firstly, Experimental analysis and comparison of several common edge detection operators and operators based on morphological. For these operators are more sensitive to noise, so that the edge portion of the plate image information can not be extracted effectively. An improved multi-structure elements edge detection operator based on morphological is proposed, which based on the anti-noise type edge detection operator. Compared to a single structural element, it can be a good noise removal, detect more edge detail, avoid jagged edges and discontinuous phenomenon. After continuous operation of mathematical morphology, we will get better connected regions, then mark these regions. Secondly, combined with the inherent characteristics of the license plate image to remove the false license plate regions and complete coarse positioning plate. Finally, remove the borders and rivets, then achieve precise positioning plate region.(2) Research on the distance formula of fuzzy C means clustering algorithm. An improved distance algorithm based on spatial information is proposed to replace the traditional Euclidean distance. The improved algorithm not only reduces the effect of the isolated noise points, but also takes into account the effect of the location information of the sample points on the clustering results. In the new distance formula, the distance between the sample points and the cluster centers is adjusted by the weight factor. Experimental verification results show that, The improved method is used as the similarity measure of clustering algorithm, which improves the accuracy of the clustering algorithm’s noise immunity and segmentation results.(3) Using the improved fuzzy C means clustering algorithm to study the license plate character segmentation. As the Chinese character is not connected, prone to adhesion and fracture, so that it can lead to the wrong segmentation of license plate characters. The fuzzy clustering algorithm is used to segment the low quality license plate image, which has good anti noise performance and robustness. So an improved fuzzy C means clustering algorithm based on membership matrix smoothing is proposed. Firstly, the clustering center is initialized by using the fast fuzzy C means clustering algorithm. And then, the improved algorithm is applied to vehicle license plate character segmentation. The experimental results show that the algorithm can improve the precision and accuracy of license plate character segmentation.