Image Processing And Recognition For Traffic Recorder Emergency Monitoring

Traffic recorder has high popularity and wide radiation range,which can effectively expand the intensity of traffic supervision.With the development of computer vision technology,assistant driving technology and chip technology,driving recorder has gradually changed from recording tool to intelligent assistant driving tool.In this paper,monocular camera and micro-electromechanical gyroscope are used to simulate the platform of driving recorder.There are three parts designed: image shake-proof,distance warning and lane deviation detection,which can realize the alarm of sudden driver fatigue,misjudgement,car-slipping and stable recording of road bumps.Real-time image stabilization are realized through gyroscope large-scale image correction and small-scale electronic image stabilization(based on feature point matching).So that a stable video stream output of the driving recorder can be get,which can reduce the difficulty of forensics and improve the recognition accuracy.The problems,include few feature points or mismatching of feature points in the traditional way,can be solved by a corrected based on angle or acceleration in advance.Front distance can be predicted by Kalman Filter and distance in the past,which can get by identification of license plate with invariable size and color.It makes up for the problem that the traditional distance measurement method based on shadow detection can not detect in close range.In addition,watershed algorithm are used to realize the identification of license plate on a blue car,which ensures the adaptability and reliability of distance detection.Early warning of lane deviation are realized by lane detection and car location based on image perspective transformation and fusion of color and gradient features.In addition,the vehicle deviation angle information can be used to judge the driver's diversion and deviation behavior,which solves the problem that the traditional vehicle deviation early warning system can not judge the active diversion and unconscious deviation.By the system integration test of the above three parts verifies the importance of image shake-proof algorithm in image identification technology,which gives a set of real-time image processing,identification and assistant early warning system which can be applied on driving recorder.