PhD thesis.


The High resolution imaging Spectrometer (HIRIS) is designed to acquire images simultaneously in 192 spectral bands in the 0.4-2,5 μm wavelength region. It will make possible the collection of essentially continuous reflectance spectra at a spectral resolution sufficient to extract significantly enhanced amounts of information from return signals as compared to existing systems. By effectively utilizing these signals, direct identification of the parameters of species can be achieved and their subtle changes can also be observed and measured. The advantages of such high dimensional data come at a cost of increased system and data complexity. For example, since the finer the spectral resolution, the higher the data rate, it becomes impractical to design the sensor to be operated continuously. Even operating HIRIS in a request only mode, its 512 Mbps raw data rate still constitutes a serious communication challenge. In order to solve this problem, it is essential to find new ways to preprocess the data which reduce the data rate while at the same time maintaining the information content of the high dimensional signal produced. In this thesis, four spectral feature design techniques are developed from the Weighted Karhunen-Loeve Transforms, They are : non-overlapping band feature selection algorithm, overlapping band feature selection algorithm, Walsh function approach, and infinite clipped optimal function approach. From a simplicity and effectiveness point of view, the infinite clipped optimal function approach is chosen since the features are easiest to find and their classification performance is the best. This technique approximates the spectra) structure of the optimal features via infinite clipping and results in transform coefficients which are either +1, - 1 or 0. Therefore the necessary processing can be easily implemented on-board the spacecraft by using a set of programmable adders that operate on the grouping instructions received from the ground station. After the preprocessed data has been received at the ground station, canonical analysis is further used to find the best set of features under the criterion that maximal class separability is achieved. In this research, both 100 dimensional vegetation data and 200 dimensional soil data are used to test the spectral feature design system. It will be shown that the infinite clipped versions of the first 16 optimal features derived from the Weighted Karhunen-Loeve Transform have excellent classification performance. Further signal processing by canonical analysis increases the compression ratio and retains the classification accuracy. The overall probability of correct classification is over 90% while providing for a reduced downlink data rate by a factor of 10.

Date of this Version