航空影像自动化镶嵌及几何纠正  

Aviation image mosaic and geometric correction based on SIFT algorithm

作　　者：张海荣[1] 刘燕华[1] 孙久运 

机构地区：[1]中国矿业大学,江苏徐州221116

出　　处：《测绘科学》2015年第8期72-76,共5页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41171343)

摘　　要：针对传统空三与正射纠正对航空摄影要求高,侧重于精度与质量,自动化程度低、处理速度慢,无法满足(无人机)航空影像应急、救灾应用问题,该文提出了一套基于尺度不变特征转换的航空影像镶嵌和纠正的自动化解决方案。应用尺度不变特征转换算法实现航空影像的特征自动提取,并通过K-D树最近邻搜索算法进行尺度不变特征转换特征的自动匹配,根据匹配的尺度不变特征转换计算出变换矩阵,通过矩阵变换和融合算法将相邻影像进行镶嵌,再用多项式模型对得到的镶嵌影像进行整体几何纠正,最终获取摄区正射影像。在Opencv和ArcGIS平台的基础上将整个正射影像制作过程开发成了一个正射影像制作系统,并且采用实际的航拍数据对算法实现效果进行了验证。实验表明,该方法可行,可提高航空正射影像的生产效率,具有生产实际意义。Traditional aerial triangulation and ortho-rectification have high requirements for aerial photog- raphy, especially for its precision and quality, as a consequence, they cannot satisfy the application of aerial images on emergency and disaster relief for their low degree of automation and slow speed of process. To solve the problem of image stitching and rectify automation, this paper put forward a set of aviation image mosaic and correct automation solutions based on SIFT（Scale Invariant Feature Transform）. Based on SIFT this pa- per realized the feature extraction from aerial photos and matched the SIFT feature by the K-D tree algorithm automatically, with the matched features the transform matrix was calculated and the mosaic image was com- pleted through matrix transformation and fusion algorithm. The polynomial model was used to correct the ob- tained images. After these steps, the digital orthophoto was completed. With the help of Opencv and ArcGIS, a system which could be used to make the DOM was produced, and the experiment proved that the system was feasible.

关 键 词：尺度不变特征转换 特征匹配 影像镶嵌 几何纠正 

分 类 号：P208[天文地球—地图制图学与地理信息工程]