Vehicle license plate character segmentation method based on machine learning and templates

Abstract

The invention provides a vehicle license plate character segmentation method based on machine learning and templates. According to the method, machine learning is used as a breakthrough point, the multi-ratio templates are used to slide on a to-be-segmented image, then a support vector machine is used to score each template on each sliding position, and finally, a template and a sliding position corresponding to a highest score are obtained for completing segmentation. The method completely discards image processing methods with the need for setting thresholds in subsequent steps, no longer needs to binarize the image, also does not need to set a threshold in an intermediate process, is easy to realize, has very high robustness at the same time, can complete segmentation on vehicle licenseplates under simple scenes, and can also complete segmentation on blurry vehicle license plates with sticky characters under complex scenes.

Description

technical field [0001] The invention belongs to image processing and pattern recognition technology. Background technique [0002] Intelligent Transportation System (ITS) is a frontier topic in the field of transportation research. It utilizes advanced wireless communication technology, navigation and positioning technology, information processing technology, automatic control technology, computer network technology, image processing and pattern recognition technology to strengthen roads, vehicles, and driving. It can realize the automation of road traffic management and the intelligentization of vehicle driving, ensure traffic safety, reduce traffic congestion, improve transportation efficiency, reduce environmental pollution, save energy, and improve economic vitality. As an important branch of the intelligent transportation system, the license plate recognition system aims to detect and identify the license plate information of passing vehicles in real time at the entranc...

AI Technical Summary

Problems solved by technology

[0003] The existing license plate character segmentation method with good effect generally uses the template of the standard license plate to guide the character segmentation. Both methods use thresholds in the implementation process, and use thresholds to binarize the vertical projection results or to binarize the image to be segmented. These methods are better for clear and high-quality license plate characters, but due to the existence of thresholds, this kind of Traditional character segmentation methods based on image processing and templates are extremely ineffective for license plates with blurred and conglutinated characters, and the process is complicated. It is necessary to judge the color of the license plate in advance and make special considerations for conglutinated characters, etc.

Images