Abstract: In the application scenes of online inspection of home appliances using machine vision,aiming at the problem that monocular vision cannot give satisfactory visual feedback,a dual-camera visual servoing method which performed a switching control scheme is proposed.The target was estimated by the global camera,and the robot was guided to move near the desired pose,assuring asymptotic stability of the visual servo.According to the image deviation and depth information provided by the eye-in-hand camera and the target velocity and orientation provided by the global camera,the dual-camera visual servo control was realized.The experimental results confirmed that the proposed method could complete positioning and tracking tasks and achieve greater tracking accuracy and stability when the velocity of target was below 0.17m/s,which had been applied toonline visual inspection of the air conditioning production line.

Key words: visual servoing, dual-camera, switching control, conveyor, visual tracking

摘要： 针对家电产品机器视觉在线检测应用场景中基于图像的视觉伺服不能保证全局收敛和单目视觉反馈不佳的问题,提出一种基于切换控制的双相机视觉伺服方法。首先使用全局相机估计目标位姿,引导机器人运动到期望位姿附近,确保视觉伺服方法能够收敛;然后根据机器人手上相机提供的图像偏差和深度信息,以及全局相机提供的目标速度和姿态实现双相机视觉伺服控制。相对于使用单应性矩阵的位置视觉伺服和混合视觉伺服方法,该方法能够定位与跟踪无标记目标,大幅提升跟踪的精度和稳定性。实验表明,基于切换控制的双相机视觉伺服方法可将控制周期缩短至33 ms,满足传送带速度在0.17 m/s范围内的机器人可获取清晰图像,实现实时跟踪检测。

关键词: 视觉伺服, 双相机, 切换控制, 传送带, 视觉跟踪