一种新的车辆辅助驾驶动态障碍物检测与分类方法

摘要

To solve the problem of low accuracy in detection and classification of moving objects in road video,this paper proposed a detection and classification method based on Kalman filter and naive Bayesian network.Firstly,it detected moving obstacles in the video using Kalman filter algorithm and extracted obstacle features.Then,it established naive Bayesian network to classify the obstacles in front of the vehicle using symmetry and edge linear horizontal degree features.Experimental results show that accuracy of moving obstacles detection is 95％,and the accuracy of obstacles recognition for motorcycle or bicycle,car front,car side and pedestrian is 98.5％.%为了解决道路视频中的运动障碍物检测和分类准确率低的问题,提出了一种基于卡尔曼滤波和朴素贝叶斯网络结合的检测与分类方法.首先采用卡尔曼滤波算法检测视频中的障碍物,并将检测到的障碍物进行特征提取.采用障碍物对称性与边缘直线水平度等特征,建立朴素贝叶斯网络对车辆前方的障碍物进行分类.实验结果表明,障碍物检测的准确率达到95％,对摩托车或自行车、汽车正面、汽车侧面和行人等障碍物识别准确率达到98.75％.