基于视差的高分辨率立体影像建筑区协同检测研究

High-accuracy built-up area information has important significance for city planning, construction and management. The within-class spectral variation and between-class spectral confusion in remotely sensed imagery degrade built-up detection performance, while the performance can be increased by using height information achieved from stereo imagery. However, the rule of height variation in built-up areas is unrevealed now. The difference and the relationship among the same built-up area in the images having different viewing angles are undisclosed. It is also unknown that where great performance gain can be achieved by stereo imagery during built-up area detection. To address these problems, built-up areas are collaboratively detected from high-resolution stereo imagery based on the stereo-extracted disparity in this research project. The key research questions are as follows. (1) The rule of disparity variation in built-up areas is revealed, and then is used for improving the stereo pair disparity index (SPDI) proposed by the project principal to indicate built-up areas in stereo imagery. (2) A high-accuracy method of collaboratively detecting built-up areas from stereo imagery is proposed, using SPDI and planar features extracted from the images having different viewing angles. The relationship among these images is established by using the stereo-extracted disparity. (3) The scene types for which stereo imagery has a distinct advantage in terms of improving built-up area detection accuracy are disclosed. This research project introduces the new notion and method of detecting built-up areas from stereo imagery, and will enhance the application level of stereo imagery for built-up area detection. This research project has important theoretical value and practical significance.

高精度的建筑区信息对于城市规划、建设与管理具有重要的应用价值。遥感影像的“同物异谱，异物同谱”现象降低了建筑区检测精度，而立体影像蕴含的高度信息能提高检测精度。但是，目前对建筑区的高度变化规律理解不足，不明确同一建筑区在不同观测视角影像之间的差异和联系，也不明确立体影像建筑区检测精度有显著优势的地区。针对这些问题，本项目开展基于视差的高分辨率立体影像建筑区协同检测研究，主要内容包括：(1)研究建筑区的视差变化规律，并优化申请人提出的立体像对视差指数(SPDI)以便较好地反映建筑区；(2)研究利用视差建立不同观测视角影像之间的联系，并提取它们的SPDI和平面特征进行立体影像建筑区协同检测的高精度方法；(3)研究并确定立体影像建筑区检测精度有显著优势的研究区场景类型。本项目研究成果将为立体影像建筑区检测提供新思路和方法，提高立体影像在建筑区检测方面的应用水平，具有重要的理论价值和实际意义。

高精度的建筑区信息对于城市规划、建设与管理具有重要的应用价值。遥感影像是获取建筑区信息的重要数据源，但遥感影像的“同物异谱，异物同谱”现象通常制约了建筑区检测精度。高分辨率立体测绘卫星和敏捷成像卫星采集了大量立体影像，它们所蕴含的高度信息有利于获取高精度建筑区信息。本项目开展了基于视差的高分辨率立体影像建筑区协同检测研究。分析了建筑区与非建筑区所处的不同地形和包含的不同地物，挖掘了建筑区的高度变化规律，优化了立体像对视差指数(SPDI)，能较好地反映高分辨率立体影像的建筑区。对于建筑区在立体影像的不同观测视角影像之间的视觉差异与不同的投影差，利用立体影像的匹配视差建立了它们之间的联系，提出了基于SPDI的高分辨率立体影像建筑区协同检测方法，以及提出了斜视投影下的融合高度与平面特征的立体影像建筑区检测方法，实现了高分辨率立体影像建筑区高精度检测。对比了立体影像和普通单视影像应用于代表性研究区的建筑区检测精度，明确了立体影像比普通影像在建筑区检测精度上有显著优势的研究区场景类型。对于包含大量纹理单一、尺寸高大建筑物的平坦地形研究区，立体影像通常比普通单视影像有显著的建筑区检测精度优势。本项目研究成果能为合理选择和高效利用高分辨率立体影像进行高精度建筑区检测提供技术支持，有利于提高立体影像在建筑区检测方面的应用水平。