Research On The Safety Warning Method Of Driver's Blind Area Based On Machine Vision

Since the car was invented by Benz in 1886,it has become an important means of transportation in people's daily travel,but in the process of automobile popularization,driving safety has also become an important safety issue that people pay attention to.According to the survey of the accident investigation,the driver is prone to traffic accidents if he / she is absent-minded for 1.5 seconds during driving.If the time is extended to 3 seconds,the risk of traffic accidents will increase exponentially.In order to ensure the safety of drivers and pedestrians,basing on the concept of intelligent vehicle driver blind area safety warning method,this paper focuses on the acceleration of image processing and the algorithms and technologies involved in driver blind area pedestrian detection.Based on the driver's actual driving perspective,this paper designs a multi feature fusion blind area pedestrian detection method,and a monocular camera ranging scheme based on interpolation calibration.The research contents are as follows:(1)Through the research of GPU parallel computing and related framework,this paper has accelerates the system image processing and target detection by GPU,providing real-time guarantee for complex system image processing and computing.(2)After analyzing the characteristics of hog?Haar and LBP,this paper proposes an excellent feature fusion scheme from vehicle video,which is verified by experiments.Finally has got the histogram of fusion scheme and classify the final pedestrian positive or negative.(3)This paper puts forward a safety warning scheme for the blind area of driver's field of vision.Based on the interpolation measurement method and the basic principle of monocular distance measurement,some pixel points are calibrated in the laboratory,and different position points are calibrated in combination with the actual scene to improve the measurement accuracy and calculation speed in the fixed angle scene.Finally,according to the reaction time of driver and the braking distance of vehicle,the measurement accuracy and calculation speed are improved Calculate the optimal warning distance and appropriately remind the driver to reduce the probability of accidents or reduce the consequences of accidents.The experimental results show that the scheme proposed in this paper can effectively identify pedestrians and ensure the real-time warning to drivers in the case of system acceleration and optimization,and has practical application ability and market economic value.