文章引入动态感兴趣区域(Dynamic Region of Interest,DROI)策略,提高基于区域卷积神经网络的快速目标检测(Faster Region-based Convolutional Neural Networks,Faster R-CNN)模型在实时视频目标检测任务中的性能.首先,分析Faster R-CNN;其次,提出一种基于DROI的优化方法,通过动态调整感兴趣区域以适应目标的运动和变化;最后,在MOT17数据集上进行实验,验证该优化方法的有效性.
Optimization Method for Real-Time Video Target Detection Based on Convolutional Neural Network
This paper introduces the Dynamic Region of Interest(DROI)strategy to improve the performance of Faster Region-based Convolutional Neural Networks(Faster R-CNN)model in real-time video target detection.First,Faster R-CNN is analyzed.Then,an optimization method based on DROI is proposed to adapt to the movement and changes of the target by dynamically adjusting the area of interest.Finally,experiments are carried out on the MOT17 data set to verify the effectiveness of the proposed optimization method.
convolutional neural networkdynamic region of interesttarget detectionreal-time
NETL
NSTL
万方数据
