摘要
行人检测是使用计算机视觉技术来判断图像或序列当中有没有行人经过,并同时对行人进行定位的技术,这项技术是无人驾驶领域中至关重要的一个研究方向。受行人个体差异、尺度姿态变化大、鲁棒性要求高等影响,使其具有挑战性,并得到了广泛关注。通过对建立在电脑视觉中的行人检测提箱进行目标分辨和分割的研究情况进行梳理和总结。首先,从阈值法、边缘检测法以及语义分割三个方面,对图像分割技术进行阐述,然后将特征提取的方法分为底层特征和机器学习的两种不同类别作出介绍,之后对分类器的构造方法进行归纳,最后通过目前存在的问题对未来的行人检测技术研究进行展望。
Pedestrian detection means using computer vision technology to determine whether there is a pedestrian in an image or a sequence and give accurate positioning.It is an important research direction in the field of unmanned vehicles.Affected by individual pedestrian differences,large changes in postures and scales,as well as high robustness requirements,pedestrian detection is challenging and has received extensive attention.The research status of image segmentation and target recognition in pedestrian detection based on computer vision was summarized and analyzed.First,the image segmentation technology was explained from the three aspects of threshold method,edge detection method and semantic segmentation,and the methods of feature extraction were divided into two different categories of low-level features and machine learning to introduce,then the construction methods of the classifier were summarized,and finally the future research on pedestrian detection technology was prospected through the existing problems.
作者
耿艺宁
刘帅师
刘泰廷
严文阳
廉宇峰
GENG Yining;LIU Shuaishi;LIU Taiting;YAN Wenyang;LIAN Yufeng(College of Electrical and Electronic Engineering,Changchun University of Technology,Changchun Jilin 130000,China)
出处
《计算机应用》
CSCD
北大核心
2021年第S01期43-50,共8页
journal of Computer Applications
基金
吉林省教育厅“十三五”科学技术项目(JJKH20200671KJ)
吉林省科技厅项目(20200703016ZP)。
关键词
计算机视觉
行人检测
图像分割
语义分割
目标识别
特征提取
深度学习
分类器
computer vision
pedestrian detection
image segmentation
semantic segmentation
target recognition
feature extraction
deep learning
classifier