Target Detection Algorithms In Hyperspectral/Multispectral Remote Sensing And Their Applications In Rock Types Detection

With the development of sensors from single band to multiple and even ultra bands, the spectral resolution of remote sensing data has been improved quickly. The application of remote sensing becomes more and more wide, such as city planning, agriculture, forest industry, ocean, geology, etc. Targets detection is the main application of remote sensing. According to the operation mode, the current target detection algorithms can be divided into two systems: one is the target detection system using band intensity; the other is the system using spectral shape feature. Because algorithms using band intensity must perform statistics for all bands, their velocity is influenced seriously by the number of bands. Therefore, "low velocity" is their main defect. This defect causes that they can be mainly used by the multispectral remote sensing and very difficult widely used by hyperspectral remote sensing. Algorithms using spectral shape feature mainly use the spectral shape of the whole bands or only a part of bands to detect objects. They are mainly applicable to detecting objects whose spectral shape is typical in hyperspectral remote sensing, such as vegetation. So, "narrow scope" is their main defect. Therefore, an algorithm with high precision, fast velocity and wide application scope will be very useful. Based on the conclusion of the two typical algorithms, a new algorithm system which was the combination of band intensity and spectral shape feature was presented. The final outcome of the present study is as following:1. According to the operation mode, current algorithms were divided into two systems: one was using the band intensity and the other was using the spectral shape feature to detect objects. Based on the analysis of their principles and applications, the advantage and defect of each system were concluded. This was the base for the new algorithm.2. A new system was presented based on the combination of band intensity and spectral shape feature. Spectral Energy level Matching, Spectral Correlation Energy level Matching, Spectral Angle Cosine Energy level Matching were three members of this new system.3. Performing many experiments to simulate the boundary conditions of the new algorithms and assess their precision, velocity and application scope. Results of the experiments showed that they were high in precision, fast in velocity and wide in application scope.