摘要
障碍物分类是智能交通应用的核心功能。本文提出了一种探测和识别道路场景下车辆与行人的算法。首先,建立棒状像素(Stixel)并由图像坐标转换至世界坐标,投影生成鸟瞰图;然后,借助聚类算法,得到原图像中待识别区域;之后,提取灰度稠密尺度不变特征变换(gray-DSIFT)特征并利用费歇尔向量(FV)编码;最后,选用随机森林(RF)分类器区分前向车辆与行人。实验结果表明:提出的方法能够有效地区分前景障碍物中出车辆与行人,提高了前期探测障碍物的效率。
Classification of obstacles is a core function of intelligent transportation applications.An algorithm for detecting and identifying vehicles and pedestrians in road scenarios is proposed.Firstly,Stixels is established and image coordinates are converted to world coordinates and a bird’s-eye view is generated by projection.With the help of the clustering algorithm,the region to be identified in the original image is obtained.Then,gray-dense scale invariant feature transform(gray-DSIFT)features are extracted and Fisher vector(FV)is used for coding.A random forest(RF)classifier is used to distinguish forward vehicles and pedestrians.The experimental results show that the proposed method can effectively distinguish vehicles and pedestrians in the foreground obstacles,the efficiency of early detection of obstacle is improved.
作者
胡福志
权悦
国海
张平娟
HU Fuzhi;QUAN Yue;GUO Hai;ZHANG Pingjuan(School of Electrical and Electronic Engineering,Anhui Science and Technology University,Bengbu 233000,China)
出处
《传感器与微系统》
CSCD
北大核心
2023年第8期161-164,共4页
Transducer and Microsystem Technologies
基金
安徽省高等学校科研项目(2022AH040234)。
关键词
棒状像素
障碍物分类
灰度稠密尺度不变特征变换
费歇尔向量
Stixel
obstacle classification
gray-dense scale invariant feature transform(gray-DSIFT)
Fisher vector(FV)
