3-D Reconstruction of Urban Scenes from Aerial Stereo Imagery: A Focusing Strategy

Baillard, Caroline; Maître, H.

doi:10.1006/cviu.1999.0793

Cited by 73 publications

(43 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Using these data sources as input information, the classification of the segmented regions leads to sufficient results in our test site. In more complex situations it can be useful to formulate the segmentation as a statistical classification problem; for example the use of a Markov random field is proposed by [2]. The detection of trees is performed together with the detection of buildings, in some cases as a pre-processing step for the building reconstruction, as described in [10].…”

Section: Interpretation Processmentioning

confidence: 99%

Automatic Mapping of Settlement Areas Using a Knowledge-Based Image Interpretation System

Straub¹,

Gerke²,

Pahl

2003

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. We introduce the knowledge-based image interpretation system GeoAIDA and give examples for an image operator, extracting trees from aerial imagery. Moreover we present a generic grouping approach, based on the Relative Neighborhood Graph. The application of the tree operator to a test site shows that the introduced approach for the delineation of trees using Active Contour Models leads to good results. The grouping algorithm is used in order to identify building rows. In the paper we shortly describe the theory of the image operator, and the performance of the whole system is demonstrated by means of examples.Results from a test area show that the information about building rows can be used for the enhancement of the building reconstruction.

show abstract

Section: Interpretation Processmentioning

confidence: 99%

Automatic Mapping of Settlement Areas Using a Knowledge-Based Image Interpretation System

Straub¹,

Gerke²,

Pahl

2003

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…The general procedure of building modeling includes two steps: the detection of building boundaries and the reconstruction of building models [13,[15][16][17][18][19]. The building detection studies [10][11][12]15,[17][18][19] separated building roofs from other features by using a series of hypotheses, which are based on the spatial, spectral and texture characteristics of buildings.…”

Section: Introductionmentioning

confidence: 99%

“…Existing studies of the reconstruction of 3D city models from LiDAR-derived nDSM or high-density point cloud data with or without additional data sources included three major categories of strategies: data-driven (bottom-up) [5,7,8,[20][21][22][23][24][25][26][27][28][29][30][31][32], model-driven (top-down) [1,2,6,9,13,14,16,[33][34][35][36][37][38][39][40] and hybrid approaches [41][42][43][44][45][46].…”

Section: Introductionmentioning

confidence: 99%

A Hybrid Approach for Three-Dimensional Building Reconstruction in Indianapolis from LiDAR Data

2017

View full text Add to dashboard Cite

3D building models with prototypical roofs are more valuable in many applications than 2D building footprints. This research proposes a hybrid approach, combining the data-and model-driven approaches for generating LoD2-level building models by using medium resolution (0.91 m) LiDAR nDSM, the 2D building footprint and the high resolution orthophoto for the City of Indianapolis, USA. The main objective is to develop a GIS-based procedure for automatic reconstruction of complex building roof structures in a large area with high accuracy, but without requiring high-density point data clouds and computationally-intensive algorithms. A multi-stage strategy, which combined step-edge detection, roof model selection and ridge detection techniques, was adopted to extract key features and to obtain prior knowledge for 3D building reconstruction. The entire roof can be reconstructed successfully by assembling basic models after their shapes were reconstructed. This research finally created a 3D city model at the Level of Detail 2 (LoD2) according to the CityGML standard for the downtown area of Indianapolis (included 519 buildings).The reconstruction achieved 90.6% completeness and 96% correctness for seven tested buildings whose roofs were mixed by different shapes of structures. Moreover, 86.3% of completeness and 90.9% of correctness were achieved for 38 commercial buildings with complex roof structures in the downtown area, which indicated that the proposed method had the ability for large-area building reconstruction. The major contribution of this paper lies in designing an efficient method to reconstruct complex buildings, such as those with irregular footprints and roof structures with flat, shed and tiled sub-structures mixed together. It overcomes the limitation that building reconstruction using coarse resolution LiDAR nDSM cannot be based on precise horizontal ridge locations, by adopting a novel ridge detection method.

show abstract

“…This is mainly due to the image matching difficulties. Image matching in urban areas are extremely difficult for two main reasons (Baillard and Maı̂tre, 1999;Marinov, 2007): 1) Complexity: the 3D model of the urban area is very complex, with many height discontinuities and large differences in elevation; 2) Density: the high density of buildings often adjacent to each other, leads to many occlusion and shadows.…”

Section: Introductionmentioning

confidence: 99%

Photogrammetric Point Clouds Generation in Urban Areas From Integrated Image Matching and Segmentation

2017

ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

ABSTRACT:High-resolution imagery is an attractive option for surveying and mapping applications due to the advantages of high quality imaging, short revisit time, and lower cost. Automated reliable and dense image matching is essential for photogrammetric 3D data derivation. Such matching, in urban areas, however, is extremely difficult, owing to the complexity of urban textures and severe occlusion problems on the images caused by tall buildings. Aimed at exploiting high-resolution imagery for 3D urban modelling applications, this paper presents an integrated image matching and segmentation approach for reliable dense matching of high-resolution imagery in urban areas. The approach is based on the framework of our existing self-adaptive triangulation constrained image matching (SATM), but incorporates three novel aspects to tackle the image matching difficulties in urban areas: 1) occlusion filtering based on image segmentation, 2) segment-adaptive similarity correlation to reduce the similarity ambiguity, 3) improved dense matching propagation to provide more reliable matches in urban areas. Experimental analyses were conducted using aerial images of Vaihingen, Germany and high-resolution satellite images in Hong Kong. The photogrammetric point clouds were generated, from which digital surface models (DSMs) were derived. They were compared with the corresponding airborne laser scanning data and the DSMs generated from the Semi-Global matching (SGM) method. The experimental results show that the proposed approach is able to produce dense and reliable matches comparable to SGM in flat areas, while for densely built-up areas, the proposed method performs better than SGM. The proposed method offers an alternative solution for 3D surface reconstruction in urban areas.

show abstract

3-D Reconstruction of Urban Scenes from Aerial Stereo Imagery: A Focusing Strategy

Cited by 73 publications

References 26 publications

Automatic Mapping of Settlement Areas Using a Knowledge-Based Image Interpretation System

Automatic Mapping of Settlement Areas Using a Knowledge-Based Image Interpretation System

A Hybrid Approach for Three-Dimensional Building Reconstruction in Indianapolis from LiDAR Data

Photogrammetric Point Clouds Generation in Urban Areas From Integrated Image Matching and Segmentation

Contact Info

Product

Resources

About