Search

We propose a method for automatic registration of high-resolution satellite imagery using LIDAR intensity data. We first generate a reference image using LIDAR intensity data as an alternative to ground control points, which allows us to conduct the GCP collection process automatically. Next, the proposed automatic matching is applied to the target and reference images. In the matching process, image chips are used as registration primitives and their edge information are employed as primary information for similarity measurements. Based on these principles, the overall registration procedure was developed to be automatic and straightforward. To test the feasibility of the developed method, we designed an experiment using multimodal and multitemporal real datasets: a stereo pair of IKONOS-2 images, one Quickbird image and LIDAR data. The experimental results show that the accuracy is acceptable for practical applications. The results of this study should pave the way for the development of an automatic and generally applicable registration method of high-resolution satellite imagery using LIDAR intensity data. [ABSTRACT FROM AUTHOR]

It is generally recognized that classification methods for remote sensing imagery show different performance levels depending on the type of dataset or classification scheme. Moreover, factors such as spectral bands, the use of ancillary data, and study area characteristics have an effect. Consequently, when users attempt land-cover classification, they encounter difficulties in selecting an appropriate image classification method. This study suggests a model Decision Support System (DSS) to assist in the selection of the optimal classification method or scheme. The DSS is established on case-based reasoning, which helps users to figure out solutions for new situations by comparison with successful experiences in the past. In this study, a hybrid method incorporating a feature-based and hierarchical structure is used to construct a case base. Four key features, dataset, location, climate, and class, are used. The DSS suggested in this study offers a graphical user interface that allows users to choose various conditions on-screen in a hierarchical structure. [ABSTRACT FROM AUTHOR]

Bezier spline fitting is a useful tool for generating smooth curves from a set of digitized points. These smooth curves may replace polylines for more accurate representation of natural linear entities. Maintaining its mathematical framework, the Bezier spline fitting can be improved so as to enhance its accuracy in representation. Key for improvement is developing methods to calculate tangent vectors and two coefficients for intermediate Bezier points, which allow higher flexibility of the control polygons than that of the Bezier spline fitting. Validity of the improved new spline fitting was tested by comparing accuracy of spline curves from the new spline fitting with spline curves from the Bezier spline fitting. Test was done using a set of mathematically simulated curves of the natural linear entities. [ABSTRACT FROM AUTHOR]