Research On Taxi Occlusion Behavior Determination And Image Forensics Technology

Building a smart green transportation city and opening up a new situation in traffic management is an urgent need of China under the development trend of "Internet +".With the acceleration of the digital modern city construction process,while providing people with more convenient travel conditions,the number of motor vehicles and the number of taxis are also increasing day by day,resulting in frequent increases in illegal and illegal operation,and urgently need to improve the level of off-site law enforcement supervision to avoid human feelings,to improve law enforcement efficiency.Based on the Beijing transportation industry science and technology project,based on the investigation of the off-site law enforcement status and application requirements of taxi operation management in Beijing,the existing license plate recognition technology is generally applied to the fixed height,angle and distance of the camera,which is difficult to apply.The drawback of identifying license plate occlusion behavior in complex environments.The specific work is as follows:(1)A taxi detection algorithm based on Haar-like+Adaboost and color features is proposed.Firstly,based on the Haar-like+Adaboost algorithm,a taxi detection model is constructed to realize real-time detection of taxis.Further,the unique color of the body of Beijing taxis is used to improve the accurate detection of taxis.The test results show that under the premise of ensuring high timeliness,the algorithm proposed in this paper has a recall rate of 95.8% for taxis.(2)A license plate occlusion decision algorithm based on the number of character projection hops and pixel ratio is proposed.Firstly,based on the Adaboost classifier,the license plate location model is constructed and the license plate location is carried out.The test results show that the accuracy of license plate location is above 93%.Then,for the positionable license plate and the non-positionable license plate,the method of using the number of character projection jumps and the proportion of the blue pixel points respectively is used to determine whether the license plate is occluded.Finally,the above algorithm is used to effectively determine the occlusion behavior of the license plate.(3)Analyze and give the requirements for off-site law enforcement for license plate occlusion behavior.Firstly,it analyzes the evidence collection standards of other traffic violations in the domestic transportation industry.Secondly,it combines the actual needs of the site and the detection environment to analyze the evidence-collecting elements of the license plate occlusion behavior.Finally,the evidence for the off-site law enforcement evidence of license plate occlusion behavior was formulated,which laid the foundation for the subsequent development of industry standards.(4)A set of license plate occlusion detection system was constructed.The system is based on C/S(Client/Server)architecture development,and has designed and implemented four main functional modules: taxi detection,license plate location,license plate occlusion identification and forensic storage,and deployed on the image information management platform of Beijing Traffic Law Enforcement Corps.The test results show that the system realizes the automatic detection of the occlusion behavior of the license plate.