A stereo image matching method to improve the DSM accuracy inside building boundaries

• DOI: 10.5589/m13-039
• Publication Date: 2013-10-30T08:56:35Z
• Authors: Chuiqing Zeng; Jinfei Wang; Peijun Shi
• Keywords: Footprint; Smoothing; Elevation (ballistics); Orthophoto; Building model

In a digital surface model derived from high-resolution stereo images, the accuracies of building rooftop elevations are usually much lower than that open areas without tall objects. This problem makes height estimation difficult images. The inaccuracy elevation is caused by many factors including shadows, occlusions, smoothing constraints in matching algorithms, and mismatch on extremely high buildings. To improve accuracy elevations, this study enhances existing image methods adding footprint maps as constraint. proposed method consists three steps. First, was used to identify corresponding locations Second, left right images were matched according color shape. Third, sub-images further at pixel level generate detailed roof elevations. Validation using surveyed demonstrated can estimate with one-third error current commercial software. addition, errors for low-rise high-rise buildings consistent method.