Image Processing Algorithms For Automated Cadastral Feature Extraction And Delineation Using Satellite Imagery

Computer vision and Image processing paradigms deal with edges and forms. Specifically edges are points where there is a boundary between two image regions. Visually, they form the outline of a feature. An edge is the boundary between a feature and its background. Technically, if edges of features are algorithmically identified and detected accurately, then all features could be located and their basic geometric properties could be determined. Similarly, these edges and forms of features are part and parcel of Cadastral Surveying and they are used to obtain the geometric values of the features, hence delineating and mapping cadastral features for the production of Cadastral plans using field oriented surveying methods. Besides cadastral plans, high resolution satellite imageries have potential cadastral values as they are used to depict cadastral features. In computer vision they are used to extract linear features. The extraction of linear features or boundaries defining the extents of lands, land covers or other topographic features are very important and useful in Cadastral Surveying.Cadastral Surveying is the cornerstone of any Cadastral System. From this type of Surveying, measurements are taken for boundary definition, to ascertain the extents of land for statutory cadastral functions. Any Cadastral surveying, among other methods, uses the principle of Plane Surveying, which in itself is a2D plane technique, to achieve its statutory goals. In developing countries, various kinds of survey methods have been employed to accomplish these goals. The most commonly used method is Plane Table Surveying for Compensation Surveys. A two dimensional cadastral map or plan is a model which represents both the cadastral and geometrical information of a2D labeled image. Cadastral Survey including plane surveying is the focus of this work vis-a-vis using image processing algorithms to process high resolution Satellite images.Though with the availability of higher spatial resolution satellite images, it is possible to prepare base maps larger than1:10,000scale using Geospatial technology, but more research using satellite imageries to delimit parcel boundaries are still needed to extensively exploit the potentials using other technologies and to respond to the United Nation Cadastral Vision call. So the research problem of this work is to provide a more effective alternative solution for classification, extraction and delineation of satellite imagery cadastral features using some integrated image processing algorithms instead of the field oriented techniques.The aims of the research are to employ the concepts of computer vision using Satellite imagery data to detect, extract, classify and delineate representations of cadastral feature boundaries with minimum human interventions. To achieve the objectives, some novel integrated algorithms have been proposed. The proposed algorithms have been used for Cadastral features detection, extraction, classification and delineation. Specifically, Canny edge detection and morphological closing algorithms have been proposed for the feature edge detections, segmentations and boundary extractions. On the resultant segmented image, Hough Transform has been proposed for the boundaries to be vectorized and delineated as straight line boundaries. Similarly, to obtain the thin boundaries, Canny edge detection algorithm has been proposed again. For cadastral features classification, discrete wavelet transform and color k-means clustering algorithms have been proposed. To preprocess the imagery, georectification algorithms for it to assume a planar surface, followed by image quality evaluation algorithms, for quantitative cadastral analysis have been proposed. Hence, novel integrated image processing algorithms have been proposed and employed to achieve the objectives, instead of the conventional field survey-oriented technique.The experimental results show the performance of the segmentation algorithm using Error Matrix for Farmland Detection Percentage and Quality Percentage parameters as87.18%and73.91%respectively. Furthermore, the Hough transform has successfully delineated the boundaries with an overall accuracy rate of73.33%. The imagery is as well classified into four distinct colored classes with an overall accuracy of88.89%using the precision and recall values. To assess the general performance and usefulness of the research, it is compared with another research whose aim was to test feature extraction and classification capability of high resolution satellite images using region growing based segmentation to extract cadastral features and object based method to classify the features. On comparison, it has been was discovered that our novel method has been very effective and its accuracy range has been within acceptable limit both in terms of Cadastral features extractions, classifications and delineations.Conclusively, this method has been proven to be effective and faster as it has minimizes the demerits associated with the conventional field survey oriented method, hence providing another perspective of achieving cadastral goals, as has been emphasized by the United Nation. As Cadastral Science evolves, this research has aimed at conducting2D plane surveys coupled with emphasizing the potential roles of image processing algorithms using high resolution satellite imagery for an effective and more viable digital Cadastre. This henceforth, would provide value-added socio economic development in the near future and beyond.