Logo Leibniz Universität Hannover
Logo: Institut für Photogrammetrie und Geoinformatik/Leibniz Universität Hannover
Logo Leibniz Universität Hannover
Logo: Institut für Photogrammetrie und Geoinformatik/Leibniz Universität Hannover
  • Zielgruppen
  • Suche

Road Junction Extraction from High Resolution Aerial Imagery Using an Existing Geospatial Database (2008)

Bearbeitung:Mehdi Ravanbakhsh

Dieser Artikel liegt nur in englischer Sprache vor!!

The need for accurate and detailed geospatial databases is growing rapidly. This requires a detailed modeling of topographic objects appearing in high resolution image data and effiecent use of image analysis tools in order to supply high quality of topographic information.

The focus of this work is on road junctions. Junctions are mainly extracted in the context of automatic road extraction in that they are modeled as point objects. This kind of modeling used so far, however, does not always reflect the required degree of detail.

In this research, junctions are modeled in detail as area objects. Furthermore, possible presence of islands is considered in our proposed junction model. We developed an approach that combines a road extraction method with a new snake model to capture the junction outline. Furthermore, a level set formulation in conjunction with a selection procedure is exploited to detect islands. Prior information derived from ATKIS is used to facilitate the extraction (Fig. 1).

Figure 1: Input and output of the approach

The approach analyses the geospatial database and derives geometric, radiometric, semantic and topological information on the junction. This information gives a rough idea of the junction and guides later processing steps. Scale space properties of roads are exploited to extract reliable road segments. Furthermore, road markings if present in the scene are detected in order to verify the obtained road segments. Road arms are obtained after road segments with similar properties are linked. The resulting road arms supply initial conditions for our snake model. We propose a novel snake model that employs ziplock snake concept and integrates it with a new external force field. This new field is a combination of the balloon force and the Gradient Vector Flow (GVF). Furthermore, the balloon force is associated with the junction shape features incorporated into our snake model implicitly. The GVF increase the capture range of snakes and provide a dense force field assuring a stable optimization. The balloon force, however, helps to overcome high variation of curvature in junctions and lack of sufficient contrast between the junction central area and the surrounding. Before snake optimization starts, initial snakes are modified based on the junction geometrical shape to assure a close global initialization. Due to the strong internal snake energy compared to the external force, the junction outline is delineated without being affected by various kinds of disturbances. Figure 2 illustrates one example of re

                                                                                                  (a)Vector data                                         (b) Initialization 


                   (c) Intermediate result                                  (d) Final result
Figure 2. Delineation of the junction outline. In (b) road segments in white provide initial boundary conditions for the snake.

 The obtained junction outline defines an area within which possibly islands exist. A level set approach is used to detect islands. Initialization of the level sets is carried out using the results obtained through a segmentation procedure. In order to select the evolved curves converged to the island boundaries, some geometric and topological constraints are introduced (Fig. 3).

(a) Junction outline                            (b) Extracted island                        (c) Final result
Figure 3. Extraction of the junction island

This type of evolution strategy, however, is not useful for roundabouts. Since the roundabout’s shape is heavily affected by the shape of its central island, we need initially to detect the central island followed by the modification of the snake external force field in order to handle cases with a high variation of curvature and all kinds of possible disturbances. The central island is detected using the used level sets and a hybrid evolution strategy to overcome various existing disturbances. This hybrid strategy includes two steps: shrinkage and iterative expansion curve evolution. Eventually, the central island is obtained when some post-processing steps are applied on the evolution results. The external snake force field is modified using a signed distance function so that initial snakes are pulled toward the outline regardless of where they are located (Fig. 4).

             (a) Initialization                    (b) Intermediate result                   (c) Final result
Figure 4. Reconstruction of the roundabout

The most outstanding feature of our approach is the integration of an existing geospatial database with active contours to extract simple and complex road junctions.
Extensive tests of the approach have been carried out using various datasets in different landscapes. The ground resolution of test gray value images is 0.1 m. The experiments show that 84% of road borders in rural and suburban areas are correctly and reliably extracted, and the achieved geometrical accuracy of better than 0.6 meter. The achieved geometrical accuracy for islands is 0.22 m with 87% of them been correctly extracted. A better assessment result is expected for central islands such that all of them are extracted with the geometrical accuracy of 0.35 m.




Ravanbakhsh, M.;Sadeghian, S. (2005): Comparative Study of Automatic and Analogue Interior Orientation for KFA-1000 Space Photo. In: the photogrammetric journal of Finland 19 (2005), Nr. 2, S. 54-63  | file |

Ravanbakhsh, M. (2005): Automatic Interior Orientation of KFA-1000 Space Photo: IntArchPhRS. Band XXXVI 1/W3. Hannover, 2005, 5 S., CD | file |

Ravanbakhsh, M. ; Heipke, C. ; Pakzad, K. (2007): Road Junction Extraction from High ResolutionAerial Images: IntArchPhRS XXXVI. Band 3/W49B. München, 2007, S. 131-138 | file |

Ravanbakhsh, M. ; Heipke, C. ; Pakzad, K. (2007): Knowledge-based Road Junction Extractionfrom High Resolution Aerial Images: 4th IEEE/GRSS/ISPRS Joint Workshop on “Remote Sensing and Data fusion over urban areas”. Paris, 2007, 8 S., CD | file |

Ravanbakhsh, M.; Heipke, C.; Pakzad, K. (2008): Extraction of road junction islands from high resolution aerial imagery using level sets: IntArchPhRS. Band XXXVII, Teil B1, Peking, 2008, S.209-214 | file |

Ravanbakhsh, M. (2008): Road Junction Extraction from High Resolution Aerial Images Assisted byTopographic Database Information : Dissertation : Deutsche Geodätische Kommission Reihe C, Nr.621 (ebenfalls in: Wissenschaftliche Arbeiten der Fachrichtung Geodäsie und Geoinformatik der Leibniz Universität Hannover, Nr. 273). Hannover, 2008, 90 S. | file |

back to list