基于改进Tiny-YOLOv3的人数统计方法

卷积神经网络已经成为了计算机视觉处理最为广泛的技术方法,基于卷积神经网络的目标检测技术也是一个热门的研究话题。本文通过引入通道注意力机制,对目标检测算法Tiny-YOLOv3进行改进,训练人体头部的目标检测模型,从而统计当前监控环境下的人数。实验结果表明该方法能取得较好的头部检测效果,人数统计准确率高。

Convolutional neural networks have become the most widespread technical method for computer vision processing.Object detection technology based on convolutional neural networks is also a hot research topic.This paper introduces the channel attention mechanism to improve the object detection algorithm Tiny-YOLOv3,and trains head detection model to count the number of people in the current monitoring environment.The experimental results show that this method can achieve a good head detection effect,and the people counting accuracy is high.