Study The Method Of Typical Blind Spot Accident Warning Of The Truck

There is a typical accident in diverse these. Due to rearview mirror and A-pillar blindspot and difference of radius between inner wheels(DRIW) during turn right driving of thetruck, the driver can not judge the state between pedestrians and the vehicle accuratelyresulting in traffic accidents. Consequently, it is important meaning for improving vehicletravel safety that study on warning of the truck typical blind spot accident．This paperbanalyzed accident reasons from the steering blind spot, DRIW and the driver vision so on.Based on studying on pedestrian detection algorithm and monocular vision distancemeasurement, the warning method is chosen. The main works contains:(1)Adopt pedestrian detection based on Haar-like feature using Adaboost algorithmand HOG feature using SVM, achieving the off-line training and on-line detection. Duringthe training, accelerate feature by the calculation of integral image. During the detection,set the Area of Interest to reduce the time of detection and reduce the false detection rate.Ultimately the pedestrian detection based on Haar-like feature using Adaboost algorithm ischosen through experiment.(2)Through of internal parameters obtained by the camera calibration, perspectiveprojection transformation to get frame coordinate and geodetic coordinate transformation,a monocular visual measuring distance model is established. It achieves estimate thedistance between pedestrians and the vehicle. The experiment resuits are presented.(3)Coalesing DRIW, vehicle blind spot and vehicles steering steady-state response, aVCW method is chosen. Vehicle collision warning(VCW)using the fuzzy patternrecognition(FPR) is a new attempt. Clustering analysis was carried out on the distance andtime information to get the standard model, further use FPR of principle of selecting nearto realize recognize samples online. The experimental results show the method has goodreal-time and robustness.