在线阅读 --自然科学版 2020年2期《一种基于场景规则与深度卷积神经网络的行人检测方法》
一种基于场景规则与深度卷积神经网络的行人检测方法--[在线阅读]
单巍, 崔少华, 周正
淮北师范大学 物理与电子信息学院, 安徽 淮北 235000
起止页码: 130--135页
DOI: 10.13763/j.cnki.jhebnu.nse.2020.02.006
摘要
深度卷积网络是解决分类问题的一种有效手段,但行人检测任务并不能通过分类来直接实现.为了在行人检测问题中进一步发挥深度卷积网络的优越分类性能,在实拍场景下,针对平直道路的情况,提出了一种基于摄像机安装位置和摄像机参数的感兴趣区域分割方法,合理利用先验知识和规则,对行人在图像当中可能出现的位置,以及不同位置上行人的尺度大小给出限制,从而系统仅对可能发生危险的区域进行搜索,避免了传统方法中多尺度遍历搜索整副图像的弊端.在此基础上,将危险区域所得的候选目标窗口作为待检测样本传送到构建好的深度卷积网络中进行分类,完成行人检测任务.实验结果表明,所研究的算法在一定距离内达到了预期的检测效果.

A Pedestrian Detection Method Based on Scene Rules and Depth Convolutional Neural Network
SHAN Wei, CUI Shaohua, ZHOU Zheng
School of Physics and Electronic Information, Huaibei Normal University, Anhui Huaibei 235000, China
Abstract:
The depth convolutional neural network (DCNN) is an effective technique for the classification problem.However,it cannot be applied directly in the pedestrian detection.To investigate DCNN in pedestrian detection task,in this paper,we proposed a method based on the installation position and parameters of the camera to segment the region of interest (ROI) under the straight road in real scene.The method of scene rules limit the location and size of pedestrians in the image,and the system only searches the possible dangerous area,avoiding the disadvantages of traditional methods which need multistage traverse the whole image.Based on it,the samples collected by this method from the dangerous areas are sent to a depth network,to finish the pedestrian detection task.The experimental results show that the algorithm achieves desired detection results within a certain distance.

收稿日期: 2019-11-26
基金项目: 安徽省高等学校自然科学研究项目(KJ2018B10);安徽省高等学校质量工程项目(2017kfk043,2017kfk044,2018jyxm0530)

参考文献:
[1]BERTOZZI M.Pedestrian Detection in Infrared Images[C]//IEEE.IEEE Intelligent Vehicles Symposium.Columbus:IEEE,2003:662-667.doi:10.1109/IVS.2003.1212991
[2]KIM S,JI Y,LEE K B.An Effective Sign Language Learning with Object Detection Based ROI Segmentation[C]//IEEE.IEEE International Conference on Robotic Computing.Laguna Hills:IEEE,2018:17190142.doi:10.1109/IRC.2018.00069
[3]XU Y L,MA B P,HUANG R.Person Search in a Scene by Jointly Modeling People Commonness and Person Uniqueness[C]//IEEE.Proceedings of the 22nd ACM International Conference on Multimedia.New York:IEEE,2014:937-940.doi:10.1145/2647868.2654965
[4]XI L,SUN S Y,LI L N,et al.Depth Estimation of Monocular Infrared Image Based on SVM Model[J].Laser & Infrared,2012,42(11):1311-1315.doi:10.3969/j.issn.1001-5078.2012.11.024
[5]HINTON G E,SALAKHUTDINOV R R.Reducing the Dimensionality of Data with Neural Networks[J].Science,2006,313(5786):504-507.doi:10.1126/science.1127647
[6]CHEN E G,TANG X H,FU B W.Pedestrian Search Method Based on Faster R-CNN with the Integration of Pedestrian Detection and Re-identification[J].Journal of Computer Aided Design & Graphics,2019,31(2):332-339.doi:cnki.sun.jsjf.0.2019-02-016
[7]LECUN Y,BOTTOU L,BENGIO Y,et al.Gradient-based Learning Applied to Document Recognition[J].Proceedings of the IEEE,1998,86(11):2278-2324.doi:10.1109/5.726791
[8]WU X F,ZHANG J X,XU X C.Hand Gesture Recognition Algorithm Based on Faster R-CNN[J].Journal of Computer-aided Design & Computer Graphics (In Chinese),2018,30(3):468-476.doi:10.3724/sp.j.1089.2018.16435
[9]TIAN Yonglong,LUO Ping,WANG Xiaogang,et al.Pedestrian Detection Aided by Deep Learning Semantic Tasks[C]//IEEE.IEEE Conference on,Computer Vision and Pattern Recognition (CVPR).Boston:IEEE,2015:5079-5087.
[10]CUI Shaohua,LI Suwen,WANG Xude.De-noising Method of Seismic Data via Improved Convolution Neural Network[J].Laser & Optoelectronics Progress,2020,57(6):061501.