Search

Vanishing point detection algorithms based on a Gaussian sphere representation have been employed in a variety of computer vision systems, for extracting 3D line orientations as a first step towards object detection. Typically, these algorithms have been applied to imagery with strong perspective effects and with little noise or texture, resulting in good solutions for line orientations. However, these algorithms can fail if perspective effects are weak, or if texture edges are predominant; they also fail to take advantage of knowledge about the objects to be detected. In this paper, two new techniques for robust vanishing point detection on the Gaussian sphere are presented; primitive-based vanishing point analysis and interpretation plane error modeling. The performance of these methods, along with two other existing methods from the literature, are quantitatively evaluated and compared for the task of building detection in complex aerial imagery. Index Terms - Object recognition, shape representation, vanishing points, photogrammetry, building detection, cartographic feature extraction, performance evaluation.

Research in monocular building extraction from aerial imagery has neglected performance evaluation in three areas: unbiased metrics for quantifying detection and delineation performance, an evaluation methodology for applying these metrics to a representative body of test imagery, and an approach for understanding the impact of image and scene content on building extraction algorithms. This paper addresses these areas with an end-to-end performance evaluation of four existing monocular building extraction systems, using image space and object space-based metrics on 83 test images of 18 sites. This analysis is supplemented by an examination of the effects of image obliquity and object complexity on system performance, as well as a case study on the effects Of edge fragmentation. This widely applicable performance evaluation approach highlights the consequences of various traditional assumptions about camera geometry, image content, and scene structure, and demonstrates the utility of rigorous photogrammetric object space modeling and primitive-based representations for building extraction. Index Terms - Monocular building extraction, performance evaluation, object detection, feature delineation, photogrammetry, computer vision, metrics.