本文提出一种基于YoloV3的视频检测算法,使用Mobile-Net的思想优化模型,并采用基于三帧差分法和粒子滤波的视频运动自适应推理算法,充分利用视频前后帧之间的目标关联性,将其部署在VisionSeed上。实验结果表明,在COCO数据集上实现了0.331 s的单帧检测速率,速度提升了31.9%,满足了嵌入式平台的运行要求。
Abstract
In the paper,a video detection algorithm based on YoloV3 is proposed,which uses the idea of mobile net to optimize the model,and adopts the video motion adaptive reasoning algorithm based on three frame difference method and particle filter,makes full use of the target correlation between the front and back frames of the video,and finally deploys it on visionfeed.The experimental results show that the single frame detection rate of 0.331 s is realized on the COCO data set,and the speed is increased by 31.9%,which meets the operation requirements of the embedded platform.
关键词
深度学习 /
目标检测 /
YoloV3算法 /
VisionSeed /
Mobile-Net
Key words
deep learning /
target detection /
YoloV3 algorithm /
VisionSeed /
Mobile-Net
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] LIN T Y,DOLLAR P,GIRSHICKR,et al.Feature pyramid networks for object detection[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Washington D.C.,USA:IEEE Press,2017:936-944.
[2] REDMON J,DIVVALA S,GIRSHICKR,et al.You only look once:unified,real-time object detection[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2016:779-788.
[3] 靳明浩.基于面部特征的嵌入式驾驶防疲劳系统设计[D].青岛:山东科技大学,2020.
[4] 李珣,刘瑶,李鹏飞.基于DarkNet框架下的YOLOV2算法的车辆多目标检测方法[J].交通运输工程学报,2018(6):143-146.
[5] 刘彦,秦品乐,曾建朝.基于YOLOV3与分层数据关联的多目标跟踪算法[J].计算机科学,2021,48(S2):370-375.
[6] 谢红,原博,解武.LK光流法和三帧差分法的运动目标检测算法[J].应用科技,2016,43(3):23-27.
PDF(3970 KB)
