视觉测量中基于单应性矩阵的平面靶标图像特征提取

Extraction of plane target image features based on homography matrix in visual measurement

针对平面圆特征靶标和平面方格特征靶标图像,研究一种基于映射矩阵实现特征提取的新方法。根据靶标图像上至少4个角点位置的特征图像坐标及其对应的三维空间坐标建立靶标平面与像平面的单应性关系矩阵,结合已知的特征点空间坐标得到与已知空间坐标特征点相对应像面上特征点的粗略图像坐标,进而通过最小二乘椭圆拟合(圆特征靶标)或角点提取算法(方格特征靶标)得到精确的特征点图像坐标。相关实验表明,本文方法可以实现平面靶标图像特征的精确提取及与空间特征点的准确对应。

In this paper, a method of extracting plane target image features based on homography matrix is proposed. With four groups of corresponding coordinates of features, the homography matrix of target plane and image plane can be established. For known spatial features, their corresponding rough image coordinates can be computed using the homography matrix With least square ellipse fitting or accurate corner pick-up algorithm,we can get the exact image coordinates of features further. Experimental results show that this method can achieve accurate features extraction and exact corresponding relationship between featuresr image coordinates and their space coordinates.