Artificial neural networks based automatic feature extraction

Cadastral Surveying deals with ownership, boundaries, extent and value of land or property in a given area, mainly for the purpose of taxation. Traditional cadastral surveying approaches are always time consuming and require lot of efforts in field surveying especially in remote and mountainous areas. The photogrammetric technique is a possible solution for this. However, up-to-date aerial photographs are hard to obtain when compare to the high resolution satellite images. Therefore, in the study Worldview-02 satellite images were used for spatial data acquisition due to its resolution, coverage and continuous data availability. The implementation of automatic feature extraction from satellite images to collect some of the possible land features would definitely reduce the time and cost of completing cadastral mapping. Therefore, the study mainly focused on linear and area features to collect with automatic approaches. The results obtained show the possibility of using the proposed approach for automatic liner and area feature extraction from worldview-02 satellite images for cadaster data collection.