Robust License Plate Recognition Algorithms

With the rapid development of intelligent transportation systems, license plate recognition. as one of the core technology, has attracting more and more attentions. Although there are a number of commercial vehicle license recognition system put into use. Most of them are subject to certain restrictions, such as limited scene, light and license plate standard. The research of robust license plate recognition algorithm has enormous scientific and practical value.This paper aims to research robust license plate recognition algorithms which could work well on diverse datasets. These datasets have different scenarios. different illumination conditions and different image quality. Our work has four main following contributions.1) To locate the license plate. We proposed a rule-based localization algorithm based on edge enhancement and local OTSU binarization, and a learning-based algorithm separately. For the two methods are highly complementary, we proposed a hybrid algorithm, which combines the advantages of the two methods, and achieve a high recall rate on the experiments.2) Traditional segmentation method is difficult to segment the license plate characters with the low image quality. In this paper, many digital image process technology are combined to achieve an algorithm to accurately split the low-quality license plate. The experimental results indicated that after applied this algorithm, the overall license plate segmentation accuracy increased from94.96%to98.63%.3)Drive a comprehensive analysis and discussion about LPR datasets. Figure out that simple dataset will cause the problem of dataset bias. We established an diverse dataset and hope to design robust algorithms on it.4) Build a complete license plate recognition system. The overall accuracy rate of experiments on two real datasets are92%and93%. The performance shows that our LPR system has high robustness and enormous practical value.