基于深度学习的行人和骑行者目标检测及跟踪算法研究

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。

Research on Target Detection and Tracking of Pedestrian and Cyclist Based on Deep Learning

In this research,YOLOv2 network is used as the basic model of target detection.In order to improve the detection accuracy of cluster small targets,residual network is added to YOLOv2 to form a new model called YOLO-R.Then the sample database of pedestrian and cyclist is constructed.The size of anchor boxes and other network parameters are modified.After this,the network model which is more suitable for pedestrian and cyclist detection is trained.The matching algorithm is used to classify pedestrian and cyclist,and Kalman filter is utilized to achieve multi-target tracking.The experiment shows that,when the training samples and network parameters are the same,YOLO-R is 3.4% higher than the mean of average precision (mAP) of the YOLOv2 network,on the premise of meeting the speed requirement,the detection result of YOLO-R is better.