融合矢量数据的高分遥感影像建筑物轮廓优化方法

Optimization of building contours by fusing vector data with high-resolution remote sensing images

针对高分遥感影像中分类法提取边缘不完整、锯齿状等问题,该文提出一种融合矢量数据优化建筑物轮廓的方法。通过融合具有较高数据完整性和准确性的矢量数据,对轮廓的不规则、缺失部分进行重构,从而提高提取轮廓结果的准确性和完整性。首先对提取的初始建筑物轮廓进行多边形拟合,并利用Shi-Tomasi算法优化建筑物局部内凹轮廓,以实现轮廓的初步优化;接着计算初步优化轮廓与矢量数据之间的距离相似度,其中对于距离较小且满足融合条件的建筑物轮廓线段点,利用矢量数据中对应的线段点进行融合,以实现轮廓的深度优化。通过3幅遥感影像进行实验和分析,该文方法的轮廓准确度达到94.8%,相比于优化前提高了7%。此外,与最小外接矩形优化轮廓方法对比,该文方法在交并补(IOU)指标上表现更好。实验结果表明,该方法通过融合矢量数据进行轮廓优化,能够更加准确、完整地反映建筑物细节,在一定程度上提高了与实际建筑物轮廓的吻合度。

A method to enhance building contours in high-resolution remote sensing images was presented in this paper by addressing the issues of incomplete and jagged edges extracted through the classification method.The proposed approach involved the fusion of vector data with higher data integrity and accuracy to reconstruct the irregular and missing parts of the contour.This fusion process aimed to improve the accuracy and completeness of the extracted contour results.Firstly,the initial building contour lines were fitted using apolygonal method.Subsequently,the local concave contours of buildings was optimized by the Shi-Tomasi algorithm.This meticulous process were conducted with the objective of attaining the initial optimization of the contour lines,ensuring an initial improvement of the contour line.Secondly,the distance similarity between the initial optimized contour and vector data was calculated.Later,the corresponding line segment points from the vector data were used to fuse with building contour line segment points meeting fusion conditions that have smaller distances to achieve a depth optimization of the contour.Experimentation of three remote sensing images yielded a contour accuracy of 94.8%,marking a 7%improvement over the pre-optimization period.Compared to the minimum outer rectangle optimization method,the proposed method performed better in the intersection over union(IOU)index.The results suggested that contour optimization by fusing vector data could depict building details more precisely and comprehensively,yielding a better match with the actual building contour.