期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2003
卷号：XXXIV-5/W12
出版社：Copernicus Publications
摘要：The process for automatic DTM, DSM generation in close-range photogrammetric applications is always a need but under investigation for many years so far. It is well known that close range conditions are wide big different from the aerial ones, basically due to the three reasons: (a) the perspective in images appears to be larger; (b) the existence of repeated patterns at every turn and (c) the occlusion problem eventuates prevailing. In all cases, the need for accurate and reliable automatic surface extraction still remains. In order, image matching applied and produced accurate results the detection of well-defined interest points is a requirement. Nowadays, a variety of key point detectors exist but each fits to specific needs. A survey of the existing interest point detectors is described in a short review. The correspondence problem, what we use to call image-matching process in digital Photogrammetry, remains of important interest in image analysis field. Matching feature points between images comprises a fundamental process in Digital Photogrammetry applications such as image orientation or DTM, DSM generation. In Computer Vision applications they use a different terminology for these processes and they target to the recovery of 3D scene structures, the detection of moving objects or the synthesis of new camera views. Even though solutions in automatic image matching are being examined y et, especially for difficult cases, such as close-range ones, significant progresses have been occurred in interest point detection. The most common advances in this field are the, Moravec, Plessey, F.rstner and SUSAN interest point operator. It is well known that in digital Photogrammetry F.rstner operator is the most usable operator for the detection of characteristics points. A strategy in matching process is considered useful if it is able to filter out many of the mismatches found in an input matching set of points, while keeping in most of the good matches present. This is the fundamental motive for detecting crucial interest points that will support in an optimum manner the matching strategy . The paper reports the problem of detection and representation of interest points in photogrammetric images as a part of a study for automatic image matching in close-range cases. We propose a new digital interest point detector, starting the approach from the existing key point detectors and extending the algorithm so as to fit in close- range photogrammetric conditions and requirements. The final result is the development of the interest point detector, what we call Plessey-Grid operator. The operator is tested in real photogrammetric condi
