Hyperspectral Data Processing: Algorithm Design and Analysis

30

Applications of Target Detection

Since there is no such a uniformly optimal technique that works for all applications, when it comes to hyperspectral data exploitation users should ask themselves a simple question, “what are the applications in which they are interested?” and “what are the benefits and advantages that a hyperspectral imaging sensor can provide for a specific application?” In fact, an application generally determines what techniques should be used, but not the other around. This chapter presents two interesting applications as examples, which require specific techniques to accomplish what they are designed for. One is size estimation of a subpixel target embedded in a single pixel vector due to its size smaller than pixel resolution determined by ground sampling distance (GSD). Under such circumstances, the target can only be detected spectrally at subpixel level, not spatially as ordinarily conducted by classical spatial domain-based image processing techniques. Besides, subpixel detection generally does not estimate the subpixel target size. In other words, a subpixel target detector may not be able to estimate the size of the subpixel targets that it detects. Therefore, new approaches must be sought for subpixel size estimation. Another interesting application is concealed target detection where targets to be detected are concealed by either natural background variations or hidden underneath man-made objects. As a result, many algorithms developed for detecting ...