Research On Visual Servo Algorithm Based On Monocular Camera

With the development of computer technology and image processing technology, image-based visual servo algorithm caused widespread concern because of low noise sensitivity and conducive to intelligent control. This paper focusing on image-based visual servo algorithm does the following work:Firstly, traditional3D pose-estimation based on single calibrated camera cannot identify whether the object in image is target, to improve it, we proposed a method by detecting the match degree between the outline of image and3D model of target. Based on this, the framework for multi-target3D pose-estimation is proposed by filling the detected outline of the image as the background to eliminate the interference to other contours.Secondly, to solve the problems that conventional cameras suffer from limited vision and panoramic cameras suffer from costly image processing, an image-based visual servo method with an angle adaptive control camera was proposed. A rotation matrix can be calculated by the transformation between the different camera coordinates. By expressing the rotation matrix in polar coordinates, the corresponding angle can be found out. This approach reduces the input image information, thus can reduce the data in image processing in visual servo and improve real-time performance of the system.