利用自由形状线特征的遥感影像分级匹配方法

发布时间：2018-04-29 05:03

本文选题：free-form + linear　；参考：《测绘学报》2017年02期

【摘要】：为较好地解决自由形状线特征匹配研究中线特征丰富信息的充分利用与有效描述这对矛盾,提出一种利用分级匹配策略的遥感影像匹配方法。为保证方法具有较高的匹配精度,从影像中检测亚像素边缘并进行有效的跟踪以得到连续性较好的自由形状线特征;再从中提取闭合线特征、线特征交点和角点等较稳定的特征作为共轭实体进行粗匹配,在确定各类特征的待选同名特征后,根据面积、角度等几何信息以及模型参数的分布特点逐步剔除误匹配,进而利用同名特征确定精匹配的初始参数;精匹配时对线特征中的亚像素边缘点加以利用,采用多层次二维迭代最邻近点(ICP)法依次利用由低到高采样率的边缘点进行匹配。试验结果表明,粗匹配选取的特征性能稳定,具有较高的匹配正确率和精度,能为精匹配提供较准确的初始参数,精匹配能达到与最小二乘影像匹配相当的亚像素级匹配精度,并且对具有较小仿射变形的影像也能实现稳定、准确的匹配。
[Abstract]:In order to solve the contradiction between the abundant information of line feature and the effective description in the research of free shape line feature matching, a method of remote sensing image matching based on hierarchical matching strategy is proposed. In order to ensure that the method has high matching accuracy, the sub-pixel edge is detected from the image and effectively tracked to obtain the free shape feature with good continuity, and then the closed line feature is extracted from it. The stable features such as the intersection of line features and corner points are used as rough matching of conjugate entities. After determining the homonymy features of various features, the mismatch is eliminated step by step according to the geometric information such as area, angle and distribution of model parameters. Then the initial parameters of fine matching are determined by using the same name feature, and the sub-pixel edge points in line feature are used in fine matching, and the edge points from low to high sampling rate are successively used to match by multi-level two-dimensional iterative nearest neighbor point (ICPP) method. The experimental results show that the feature performance of rough matching is stable, with high accuracy and accuracy, and can provide accurate initial parameters for precision matching. The precision matching can achieve sub-pixel matching accuracy equivalent to that of least square image matching. And the image with small affine deformation can also achieve stable and accurate matching.
【作者单位】：信息工程大学地理空间信息学院;地理信息工程国家重点实验室;
【基金】：国家自然科学基金(41601507) 地理信息工程国家重点实验室开放基金(SKLGIE2015-M-3-3)~~
【分类号】：TP751

【相似文献】