基于深度学习的行人和骑行者目标检测及跟踪算法研究Research on Target Detection and Tracking of Pedestrian and Cyclist Based on Deep Learning
胡超超;刘军;张凯;高雪婷;
摘要(Abstract)：
以YOLOv2网络作为目标检测的基础模型,为提高模型检测群簇小目标的准确率,在YOLOv2中加入残差网络,构成YOLO-R网络,通过构建行人和骑行者样本库,以及修改anchor boxes尺寸等网络参数,训练出更适合检测行人和骑行者目标的网络模型,并通过匹配算法完成行人、骑行者分类,进一步运用Kalman滤波实现多目标跟踪。试验结果表明:在训练样本、网络参数相同的情况下,YOLO-R比YOLOv2网络的平均精度均值(mAP)提高了3.4%,在满足速度要求的前提下,YOLO-R网络检测效果更优。
关键词(KeyWords)： YOLO-R网络;卡尔曼滤波;目标检测;深度学习
基金项目(Foundation): 国家自然科学基金项目(51275212)
作者(Authors): 胡超超;刘军;张凯;高雪婷;
DOI: 10.19620/j.cnki.1000-3703.20180628
参考文献(References)：
- [1]Wang L,Li W,Zhang Y,et al.Pedestrian Detection Based on YOLOv2 with Skip Structure in Underground Coal Mine[C]//2017 IEEE 3rd Information Technology and Mechatronics Engineering Conference(ITOEC).IEEE,2017:1216-1220.
- [2]韩凌波.一种新的K-means最佳聚类数确定方法[J].现代计算机,2013(20):12-15.
- [3]He K,Zhang X,Ren S,et al.Deep Residual Lear-Ning for Image Recognition[C]//Computer Vision and Pattern Recognition.IEEE,2016:770-778.
- [4]Glorot X,Bordes A,Bengio Y,et al.Deep Sparse Rectifier Neural Networks[C]//International Conference on Artificial Intelligence and Statistics,2012:315-323.
- [5]Maas A L,Hannun A Y,Ng A Y.Rectifier Nonlinearities Improve Neural Network Acoustic Models[C]//ICMLWorkshop on Deep Learning for Audio,Speech,and Language Processing,2013.
- [6]Ioffe S,Szegedy C.Batch Normalization:Accelerating Deep Network Training by Redu-cing Internal Covariate Shift[C]//.the 32nd International Conference on Machine Learning,Lille,France,2015.2015:448-456.
- [7]Neubeck A,Gool L V.Efficient Non-Maximum Suppression[C]//International Conference on Pattern Recognition.IEEE,2006:850-855.