KNOWLEDGE SUPERVISED PERCEPTUAL GROUPING BASED QUALITATIVE BUILDING DETECTION FROM MONOCULAR AERIAL IMAGES

Abstract

This paper addresses an important and practical problem in computer vision and pattern recognition — qualitative target detection from aerial images. In particular, it discusses the problem of qualitative building detection based on a monocular aerial image. The approach proposed, due to its independence of site models or camera calibration information, complements the model based approaches developed in the rest of the research community of building detection from aerial images. Specifically, a knowledge supervised perceptual grouping (KSPG) system based on reinvestigation, and hypothesis generation and verification, is presented, and is shown to be reasonably robust in experiments using real data.