公路汽车视觉导航中运动目标的检测与跟踪方法的研究  

Moving Target Detection and Tracking in Vision Navigation of the Vehicle on the Road

作　　者：苏开娜[1] 丰丽军[1] 刘忠云[1] 金蕾[1] 

机构地区：[1]北京工业大学计算机学院,北京100044

出　　处：《计算机工程》2001年第4期46-48,81,共4页Computer Engineering

基　　金：北京市自然科学奖金资助项目(4992001);北京市教委资助项目(99KJ06)

摘　　要：通过对室外行驶车辆上所采集的长序列图象的处理与分析，研究公路汽车计算机辅助视觉导航中运动目标的检测与跟踪问题。算法的核心是根据图象中的水平、垂直边缘及目标尾部的局部对称性来检测和确认图象中目标的存在。进而建立描述目标与摄象机之间的相对运动的状态模型和表示目标特征的透视投影模型，应用扩展的卡尔曼滤波器进行3维运动参数估计。Motion this paper researches on moving target detection and tracking in automatic visual road vehicle guidance system by processing and analysing a long sequence images of real road which are gathered on a vehicle driving outside. The core of our algorithm is to detect a target by finding the horizontal and vertical edges, and to verify whether a target exists or not by computing the local rear symmetry of the target in the image. And building the state model that can represent the relative motion between the target and the camera, the perspective model that can represent the features of th the target, then estimate the 3D-motion parameters by generalized nonlinear Kalman filter.

关 键 词：汽车目标 视觉导航 运动目标 检测 长序列图象 运动参数估计 

分 类 号：U492.22[交通运输工程—交通运输规划与管理]