Effective planning and management of informal settlements requires adequate and up-to-date geospatial data. Existing geospatial data acquisition techniques based on the photogrammetric processing of large-format aerial photographs are expensive and time-consuming when used for small area informal settlement mapping and geographic information system (GIS) data updating. Automated extraction of buildings from aerial images has been an active research topic in both digital photogrammetry and computer vision cpmmunities for a number of years. As fully automated methods are still unattainable, semi-automated approaches with user interaction are considered to be a good compromise. Moreover, the large variation in most shack properties implies that many of the vision cues used in formal building reconstruction are less reliable for use in shack reconstruction. Nevertheless, the simple geometry (e.g., flat-roofed, 4-sided, similar heights) of most shacks implies that the shack reconstruction task can be simplified into the reconstruction of a combination of the two-dimensional (2D) roofs with a representative roof height to form the two and a half dimensional (2.5D) shack models.
This dissertation deals with the development and implementation aspects of a semi- automated shack reconstruction strategy for informal settlement GIS (ISGIS) data collection. The main features of this strategy include: (1) the use of color orthoimagery and digital surface model (DSM) generated from high-resoltion small-format digital aerial imagery, (2) the separation of the shack reconstruction task into automated shack detection and interactive shack delineation, (3) the integration of multiple cues in a desktop GIS environment, and (4) the 2.5D shack modeling and the three-dimensional (3D) realistic visualization.
In accordance with this strategy an informal settlement modeling system (IS-Modeler) prototype has been developed. IS-Modeler is comprised of three software components: digital photogrammetry, image analysis and GIS. The former two are used to generate multiple cues, primarily DSM blobs, shadows and attributed edge contours, the latter is used to implement shack detection, delineation and visualization.
Firstly, a DSM-based approach to shack detection is proposed on the basis of assuming the existence of a detailed digital terrain model (DTM) and the presence of shadows. This approach fuses the shadow data derived from statistical classification with the 2.5D blobs derived from the segmentation of a normalized DSM to automatically detect shacks automatical within a desktop GIS environment. The shadow-refined blobs suffice for a coarse description of the individual shacks, as an end product for many informal settlement applications, such as shack counting.
Secondly, an interactive approach operated in the destop GIS environment is presented for the applications requiring more precise shack delineation. The approach employs the cue-guided edges based on photometric and chromatic attributes to form the shack-roof parallelogram hypotheses and to finally refine them by means of user interactions.
Thirdly, the ArcView 3D Analyst and the Virtual Reality Modeling Language (VRML) are employed to implement 3D representation of an informal settlement scene and to create a vitual reality model through integrating the generated 2.5D shack models, DTM and color orthoimagery, enabling the remote users to explore a virtual informal settlement world via Internet.
The novel and the most important feature of IS-Modeler is the integration of multiple cues in a desktop GIS environment. This implies performing shack reconstruction in object space, the advantage of which is that all georeferenced sources of information can be directly exploited. Experimental results show that IS-Modeler is low-cost, feasible and relatively reliable, and simple-to-use by non-expert users.
The successful development of IS-Modeler suggests that integration in a desktop GIS environment potentially offers an effective solution to simple and fast informal settlement GIS data updating. Some recommendations and suggestions are finally drawn up for further research to be conducted to ensure that this progress goes beyond the prototype stage and towards everyday use.