机动车辆辅助驾驶动态障碍物检测方法  

Detection of Moving Obstacles in Auxiliary Driving System

作　　者：董大明[1] DONG Da-ming(Shanxi Traffic Vocational and Technical College,Taiyuan 030031,China)

机构地区：[1]山西交通职业技术学院,太原030031

出　　处：《火力与指挥控制》2018年第4期9-12,17,共5页Fire Control & Command Control

摘　　要：针对背景复杂多变和摄像机的随车运动对机动车辆辅助驾驶系统动态目标检测算法提出的鲁棒性和实时性要求,提出了一种基于快速全局背景运动补偿和改进三帧差分法的快速目标检测算法,基于影像金字塔两步法估计改进SIFT方法,实现快速全局背景运动补偿参数估计,在三帧差分图像上,通过改进假设检验及边缘检测与目标区域相"与"最终实现动态目标的准确检测。实验表明,算法极大地提高了特征配准和动目标检测速度,提高了全局运动补偿参数估计和目标检测的精确性,满足系统的实时性要求。In view of the robustness and real-time requirements of the dynamic target detection algorithm resulting from the complex background and the moving motion of the camera,for the vehicle assisted driving system,a fast target detection algorithm based on fast global background motion compensation and improved three frame difference method is proposed.Firstly,the image pyramid is used to improve the SIFT method by two stages to achieve fast parameter estimation for global background motion compensation.And then,by improving hypothesis testing method in the three frame difference image,the moving targets can be detected by conducting‘and’operate between edge detection and target area.Experimental results demonstrate that the accuracy of parameter estimation is improved by the proposed algorithm,and the speed of feature registration and moving target detection is greatly improved,meeting the real-time requirements of the system.

关 键 词：全局背景运动补偿 改进SIFT算法 假设检验 帧间位置估计 三帧差分法 

分 类 号：U463.6[机械工程—车辆工程]