智能车中基于单目视觉的前车检测和跟踪.pdfVIP

第 25 卷第 1 期 计算机应用 Vol . 25 No . 1 2005 年 1 月 　 Computer Applications 　 J an . 2005 ( ) 文章编号 :1001 - 9081 2005 01 - 0220 - 04 智能车中基于单目视觉的前车检测和跟踪 皮燕妮 ,史忠科 ,黄 　金 (西北工业大学 自动化学院 ,陕西 西安 710072) (piyanni @sina . com) 摘 　要 :提出了一个改进的单 目视觉方法 ,用于智能车在结构化公路环境下准确检测和跟踪前方 车辆 。该方法先利用图像灰度梯度检测前车 ,剔除可能的虚检测 ,建立新 目标的二维模型 ;然后用卡 尔曼滤波方法预测下一帧的目标位置 ,在预测位置附近用边缘投影方法定位 目标 ;设计了一种新的四 因素似然度函数 ,验证跟踪结果与检测结果的匹配度 ,当跟踪失败时 ,重新检测前车 。利用长图像序 列 PETS2001 进行实验 ,结果表明该方法可以有效的检测和跟踪本车车道前方视野中的车辆障碍物 , 为智能车的防撞预警和控制系统提供可靠信息 。 关键词 :智能车 ;视觉导航 ;前车检测 ;卡尔曼滤波 中图分类号 : TP274 　　文献标识码 :A Preceding car detection and tracking based on the monocular vision PI Yanni , SHI Zhongke , HUAN G Jin (College of A utomation , N orthwestern Poly technical University , X ian S haanxi 710072 , China) Abstract : An improved monocular vision method was proposed for intelligent vehicle to detect the preceding car in structural road environment . This method firstly detected the preceding car by identifying the edges of the obj ect and then the false obj ect was eliminated. The eligible obj ect expressed as a 2D model was acquired. Secondly the location of the obj ect in the next frame was predicted by Kalman filter , and the obj ect was detected near that location by means of edge proj ection . Finally a novel likelihood function was desired to verify the tracking results. If the likelihood was too low , the obj ect was detected over again , or else the 2D model would be updated. Experiment results of the image sequence