The Algorithms Research On License Plate Character Segmentation And Recognition

The Age of Big Data is rife with all kinds of information. With its core technology: LPRS(License Plate Recognition System), the ITS(Intelligent Transportation System) is making transportation smoother and safer. With the improved living standard of the people and increased requirement for the resolution of video in the security field, the ITS has higher requirement for LPRS. Despite the successful application of some existing products, a lot of universities, institutes and enterprises are still working on the optimization of the algorithm.In this paper, Chapter 1 introduces the background and status of LPRS as well as some typical methods for license plate character recognition. Chapter 2 presents the main license plate pre-processing operations before character recognition and briefly introduces the license plate location, skew correction algorithms, which are the basis for character recognition. Chapter 3 creatively proposes the hierarchical segmentation algorithm based on the maximum license plate character spacing. This flexible algorithm can even perfectly segment highly degraded license plate. Chapter 4 introduces the LS-SVM algorithm based on the wavelet kernel function. Combining the extracted character features, character recognition can be realized via the classifier. In Chapter 5, with RBM as the feature extractor and combining the softmax classifier to form the character recognition system, algorithm comparison has been made through some experiments. The results demonstrate the advantages of the algorithm in letter and number recognition.Although the position of the processed license plate in the image is not limited for the algorithm introduced in this paper, the focus is still the single-row characters in the front and back of the car for practicability and marketing reasons. Test results have shown its strong anti-jamming ability against the distortion, deformation, tilt and contamination of the license plate and its great adaptability to different lighting conditions. Besides Chinese characters, it also has high recognition rate for the other two types of characters.