目标检测算法在嵌入式平台中的应用

Application of Target Detection Algorithm in Embedded Platform

目前深度学习中的目标检测算法,经过几年的发展已经比较完善,在一些公共数据集上,如COCO2017、Pascal VOC2010-12等数据集上已经有较好的识别效果。在一些实际应用场景和移动平台应用中,依然具有一定的局限性。以交通道路车辆检测为研究背景,阐述利用现有汽车数据集对目标检测框架进行训练,在得到训练好模型后将其移植到嵌入式平台进行性能参数优化。在嵌入式平台加装工业相机作为图像输入现场测试,使深度学习目标检测算法能够在嵌入式平台上得到很好应用。

At present,the target detection algorithm in deep learning has been relatively perfect after several years of development,and has a good recognition effect on some public data sets,such as coco2017,Pascal voc2010-12 and so on.There are still some limitations in some practical application scenarios and mobile platform applications.Taking the traffic vehicle detection as the research background,this paper expounds that the target detection framework is trained by using the existing vehicle data set,and after the trained model is obtained,it is transplanted to the embedded platform for performance parameter optimization.An industrial camera is installed on the embedded platform as image input for field test,so that the deep learning target detection algorithm can be well applied on the embedded platform.