中文摘要：

图像特征点中心定位是位姿测量中求解位姿的关键．针对同一图像有多个几何形状的问题，首先利用开源计算机视觉库OpenCV对原图像进行区域分割，得到目标区域的边缘，其次将每个区域进行像素填充；然后用Harris算法进行角点检测；最后根据角点坐标再计算中心点，并将文中算法就一幅图像中有单个几何形状的情况与原始Harris和改进的Harris算法从准确率和效率上进行了比较．实验证明，这种中心定位算法不但提高了准确率和效率，而且误差可以减小到1个像素以内．

英文摘要：

The determination of center is the key to solve attitude measurement problem To solve the problem that there are multiple geometrical figures in one image, the method segments regions in the original image to get the edge of the target area using Open Source Computer Vision （OpenCV） Library. Then pixels were filled in each region. After that corners were detected with Harris comer detector. At last central points were calculated according to the comer coordinate. The paper compared this method to original Harris and improved Harris in accuracy and efficiency when there is a single geometric figure in one image. Experimental results show that this determination of center algorithm advances in both accuracy and efficiency, and its error is within 1 pixel.