基于单目视觉和固定靶标的位姿测量系统

利用计算机视觉进行位姿测量的方法广泛应用于机器人系统、运动体控制系统和精密检测系统。研究和设计了一种基于固定靶标的单目视觉定位系统和方法，用最少硬件资源实现精密定位。首先，利用图像匹配的方法检测出平面靶标在图像中的坐标，图像匹配采用SIFT算法和映射匹配方法，之后利用固定靶标的特性求取中心点。实验利用多幅图像样本验证了图像匹配的准确性和鲁棒性。然后，针对呈矩形分布的PnP问题，提出了一种新的求解方法，以靶标控制点的图像坐标和空间坐标作为输入，得到了移动物体与摄像机的三维相对位姿。实验利用五维精密位移台移动目标物体并拍摄多副图像，结果表明位姿测量系统在800 mm范围内达到mm级精度，可以满足应用需求。

Position and orientation measurement system based on monocular vision and fixed target

The position and orientation measurement system based on computer vision is extensively applied on robotics, motion control and precision detection systems. Using the minimum hardware resource, a localization system based on mono-vision and manual planar target was designed and the method of image matching and position resolving was also studied. Firstly, the object detection based on image matching was used to get the coordinate of the planar target in image. The matching was based on SIFT features and projection estimation to detect the target in image, and then the accurate coordinate of the target's center was calculated by some inherent shape information. A number of image samples were used to validate the accuracy and robustness of the image matching algorithm. Secondly, a new method to solve the PnP problem based on the rectangular distribution was proposed. The method uses the locations of target control points in the image coordinate system and the object coordinate system was used to get the relative position and orientation between the moving object and the camera. The experiment was conducted on the 5D precision displacement stage and results show that the system can achieve the accuracy level of mm in the range of 800 mm, which meets the project requirement.