Study On Monocular Vision-based Objective Vehicle Detection Algorithm

With the improvement of people's life level,the vehicle is increasingly to be a necessity in people's daily life.However,in recent years,the increasing number of nation-wide car ownership has gradually been a potential obstacle to decrease the traffic jams.Therefore,what measures should be taken to improve the vehicle safety driving has been attached to importance in whole society.At the same time,it has a fast development in the vehicle and real-time Inter-vehicle distance detection,which is able to offer an excellent solution to assist the vehicle safety driving.This paper will conduct a deep study on real-time Inter-vehicle distance and vehicle type algorithm by adopting the method of feature extraction based on monocular-vision platform.The paper proposed a divergent-expression scanning principle based adaptive threshold-segment algorithm,which is able to raise the accuracy,real time and robustness of the vehicular bottom shadow and Inter-vehicle distance detection.Firstly the ROI(Region of Interest)is determined according to the area,where the front vehicles include.The gray-scale information of the points in the scanning lines is acquired by adopting the divergent-expression scanning principle;Secondly,considering the gray-scale feather of shadow,the image segmentation is completed by using sectional-type adaptive threshold computational approach,and then the area of the vehicle shadow is determined;thirdly,the vehicle type is detected by using vehicle-contour feather,rectangularity and vehicle symmetrical characteristic;finally,the inter-vehicle relative distance is estimated by using IPM(Inverse Perspective Mapping).The results indicate that the adaptive threshold computation,which uses divergent-expression scanning method in sectional-type-ROI,is able to decrease the impact on extract the vehicle bottom shadow in near and far area and complex road condition,and then the real time and robustness is improved;The scanning area can be restricted based the ROI certificated by the three lane.and then the computation is put down and the real time is improved by using the point gray-scale information in scanning lines,which is used to achieve the threshold segmentation and ensure the enough information.