基于深度图像的动态指尖检测  

Dynamic Fingertip Detection Based on Depth Image

作　　者：敖良忠[1] 畅浩天 丁坚 贾文彬 AO Liangzhong;CHANG Haotian;DING Jian;JIA Wenbin(Aero Engine Maintenance Training Center,Civil Aviation Flight College of China,Guanghan 618307,China)

机构地区：[1]中国民用航空飞行学院航空发动机维修培训中心,四川广汉618307

出　　处：《传感器世界》2023年第6期31-37,共7页Sensor World

摘　　要：机器人的广泛应用使如何更方便地使用机器人成为研究热点,机器人演示编程是一种快捷的机器人编程方式。如何通过人的演示引导与控制机器人,视觉交互是一种主流的方法。文章提出一种基于深度图像的动态指尖检测,首先,通过深度相机采集手部演示的过程;然后,分析场景中不同物体的深度值,将手部活动区域从复杂的背景中提取出来;接下来,对手部活动区域使用帧差法,提取手部图像,在获取手部轮廓的基础上提取轮廓凸包,对轮廓凸包点使用曲率检测与重心距离相结合的判定方法,检测出单指尖点的位置。将单指尖点在相机坐标系下的三维坐标作为引导机器人运动的路径点。实验结果表明,该方法具有较高的准确性,并且跟踪速度快,符合实时性的要求。With the wide application of robots,how to use robots more conveniently becomes a research hotspot.Robot demonstration programming is just a fast way of robot programming.How to guide and control robots through human demonstration,visual interaction is a mainstream method.In the paper,a dynamic fingertip detection based on depth image is proposed.Firstly,the process of hand demonstration is captured by the depth camera;Then,the depth value of different objects in the scene is analyzed to extract the active area of the hand from the complex background;Next,the frame difference method is used to extract the hand image from the active area of the hand,and the contour convex envelope is extracted on the basis of obtaining the contour of the hand.The location of the single fingertip point is detected by the judgment method combining curvature detection and center of gravity distance for the contour convex envelope point.The three-dimensional coordinates of a single fingertip point in the camera coordinate system are used as the path points to guide the robot.The experimental results show that the method has high accuracy and fast tracking speed,and meets the requirement of real-time.

关 键 词：深度图像 指尖检测 人机交互 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]