Research And Implementation Of License Plate Recognition System In Video Stream Environment

In recent years,with the increasing number of vehicles in China,the application demand of license plate recognition technology in the field of intelligent transportation is also increasing.The traditional license plate recognition uses software and hardware assistance.The vehicle is detected in advance by laying hardware such as vehicle sensors on the recognition section,and when it passes,it takes a picture to obtain a vehicle image and identify the image.Due to the problems of light,front vehicles blocking the rear vehicles,etc.,the vehicle information obtained by taking pictures has incompleteness and distortion,and it is necessary to preset the hardware to sense the vehicle cost.Therefore,this thesis designs a set of license plate recognition system based on video streaming.The system detects and recognizes the vehicle license plate information in the road monitoring video in real time.Compared with the traditional camera recognition system,users can efficiently obtain the target vehicle license plate information and intuitively obtain the target vehicle driving trajectory.The main research work of this is as follows:(1)This thesis analyzes the advantages and disadvantages of traditional inductive camera license plate recognition and license plate recognition in the adverse environment of video streaming,introduces the characters included in license plate in China,analyzes the characteristics of license plate in China,and introduces the process of license plate recognition.Based on the characteristics of China's license plate and the process of license plate recognition,the related technologies used in the license plate recognition system in the video streaming environment designed in this thesis are introduced.(2)Weighing the calculation amount and recognition rate in video stream processing,this thesis improves the key frame extraction algorithm based on SIFT features to extract key frames containing complete information of vehicle license plates.This algorithm draws on the idea of matching the background difference and SIFT features,and divides the video frames into two types: mutation frames and non-mutation frames.It determines whether it is a key frame,and finally determines its similarity based on the hash algorithm to remove redundancy frame.The improved SIFT feature key frame extraction algorithm is compared with the traditional background difference method and the SIFT motion feature algorithm.The experimental results show that the improved SIFT feature key frame extraction algorithm has lower missed frames in key frame extraction.Rate,real-time better.(3)On the basis of traditional induction camera license plate recognition,combining the characteristics of video stream processing and the needs of this subject,the overall system architecture and functional module division of the thesis are proposed.According to the functional requirements,the license plate recognition system in the video stream environment is divided into three parts: single road section license plate recognition,multiple road sections license plate recognition and vehicle information statistics.Each part is introduced in detail.The core function of the system's single road section license plate recognition is analyzed in detail and outline design.(4)The improved algorithm is applied to the license plate recognition of video stream to realize the license plate recognition function of a single section,and it is tested with surveillance video to finally complete the core recognition function of the system and verify the system feasibility.