手势识别的自跟随智能车辆控制系统研究

Research on self-following intelligent vehicle control system based on gesture recognition

针对自跟随低速智能车辆视觉跟随效果不理想及车辆需要频繁启停的问题,提出一种手势识别的车辆控制方法。采用YCrCb颜色空间Cr分量的Otsu法对手部图像进行阈值分割,通过形态学处理方法提取手部轮廓,采用支持向量机训练模型对手势进行指令识别;将指令下发至运动模块对车辆进行控制,在直线型、“L”型和“U”型三种场景下对自跟随低速智能车辆的跟随精度进行测试。结果表明:基于OpenCV的手势识别方法可用于智能车辆的跟随控制;采用手势识别控制的平均相对误差为1.16%,比视觉自跟随的平均相对误差降低了71.85%。

Aiming at the problem that the visual following effect of self-following low-speed intelligent vehicles is not ideal and the vehicles need to start and stop frequently,a vehicle control method based on gesture recognition is proposed.The Otsu method with Cr component of YCrCb color space is used to segment the threshold of hand image,the hand contour is extracted by morphological processing method,and the instruction recognition is carried out by support vector machine training model.The command is sent to the motion module to control the vehicles,and the following accuracy of the self-following low-speed intelligent vehicles is tested in three scenarios,i.e.,linear type,“L”type and“U”type.The results show that the gesture recognition method based on OpenCV can be used for the following control of intelligent vehicles.The average relative error of gesture recognition control is 1.16%,which is 71.85%lower than that of visual self-following control.