Research On Vehicle Detection And Ranging Technology Based On Kinect

With the increasing demand for life quality,the popularity of automobiles poses severe challenges to road traffic safety.In vehicle driving,vehicle assisted driving equipment is an effective method to prevent rear end accidents and improve traffic safety by detecting and ranging vehicles in front.With the rise of Kinect,its advantages of integrating color camera and depth camera make it a simpler,more integrated,more efficient and more popular method.This dissertation studies the algorithm of vehicle detection and ranging function with Kinect.As a role of getting image information,the software application program is developed to detect the vehicle target in the field of view,and obtain the depth distance information to the camera plane.Camera calibration is an important basis for vehicle detection and ranging.The first step is to calibrate camera internal parameters and the distortion coefficient of nonlinear distortion with Zhang Zhengyou's method.The second step is to optimize the inaccuracy of gaining Harries points in previous calibration with second order reprojection method.After collecting calibrated camera internal parameters,specific algorithm research can be performed.Vehicle detection is the prerequisite of distance detection.Two methods are proposed by studying linear filtering and morphological filtering.Linear filtering is used for noise reduction and morphological filtering is used to reduce the depth of image noise.The comparison among the Canny,Sobel and Laplacian image segmentation operator returns the result that Canny and Sobel operators show a better performance in the detection experiments.A specific model is proposed to fit vehicle tail shape with linear Hough transformation.With these methods,a new software is developed to do effective detection of vehicles in front.Depth ranging is the core function of the system.Research is conducted to show the depth distance acquisition method of Kinect sensor based on its ranging principle.Further ranging experiments under different light conditions show that Kinect ranging has good performance even without sufficient light.Moreover,in order to decrease th measurement error caused by increasing of the distance from the object to the sensor,further study is conducted.After analyzing the causes of errors,a depth error compensation model is established with pixel Euclidean distance and the corresponding depth compensation value.The measurement error is reduced to ±10mm,and the validity of error compensation is verified.Driving safety is a frequently mentioned topic.The application technology of Kinect in vehicle detection and ranging is studied,which provides an effective solution for it.At the same time,it provides a practical theory and technical support for the development of intelligent vehicles and driverless vehicles in the new era.It has significant social and economic benefits.