基于深度学习的交警指挥手势识别

A Study of Traffic Police Command Gesture Recognition Based on Deep Learning

为解决无人驾驶汽车快速准确地识别交警指挥手势的问题,本文提出一种基于深度学习的三通道输入交警指挥手势识别方法.仿真实验表明,利用深度学习优化模型参数后,采集的8种交警指挥手势数据集的平均识别准确率可达97.87%,识别率较高,具有一定的应用价值.

In order to achieve quick and accurate identification of the hand gesture of the traffic police, a three-channel-input traffic police command gesture recognition method based on depth learning is proposed in this paper. Simulation experiments show that the average recognition accuracy of the 8 traffic police command gestures in the data sets can reach 97.87%, and is therefore of application value.