Within the field of pattern recognition (PR) a very active area is the clustering and classification of multispectral data, which basically aims to allocate the right class of ground category to a reflectance or radiance signal. Generally, the problem complexity is related to the incorporation of spatial characteristics that are complementary to the nonlinearities of land surface process heterogeneity, remote sensing effects and multispectral features. The present research describes the application of learning machine methods to accomplish the above task by inducting a relationship between the spectral response of farms’ land cover, and their farming system typology from a representative set of instances. Such methodologies are not traditionally used in crop-livestock studies. Nevertheless, this study shows that its application leads to simple and theoretically robust classification models. The study has covered the following phases: a)geovisualization of crop-livestock systems; b)feature extraction of both multispectral and attributive data and; c)supervised farm classification. The first is a complementary methodology to represent the spatial feature intensity of farming systems in the geographical space. The second belongs to the unsupervised learning field, which mainly involves the appropriate description of input data in a lower dimensional space. The last is a method based on statistical learning theory, which has been successfully applied to supervised classification problems and to generate models described by implicit functions. In this research the performance of various kernel methods applied to the representation and classification of crop-livestock systems described by multispectral response is studied and compared. The data from those systems include linear and nonlinearly separable groups that were labelled using multidimensional attributive data. Geovisualization findings show the existence of two well-defined farm populations within the whole study area; and three subgroups in relation to the Guarico section. The existence of these groups was confirmed by both hierarchical and kernel clustering methods, and crop-livestock systems instances were segmented and labeled into farm typologies based on: a)milk and meat production; b)reproductive management; c)stocking rate; and d)crop-forage-forest land use. The minimum set of labeled examples to properly train the kernel machine was 20 instances. Models inducted by training data sets using kernel machines were in general terms better than those from hierarchical clustering methodologies. However, the size of the training data set represents one of the main difficulties to be overcome in permitting the more general application of this technique in farming system studies. These results attain important implications for large scale monitoring of crop-livestock system; particularly to the establishment of balanced policy decision, intervention plans formulation, and a proper description of target typologies to enable investment efforts to be more focused at local issues.

Glaciers and ice caps (GIC) are central parts of the hydrological cycle, are key to understanding regional and global climate change, and are important contributors to global sea level rise, regional water resources and local biodiversity. Multispectral (visible and near-infrared) remote sensing has been used for studying GIC and their changing characteristics for several decades. Glacier surfaces can be classified into a range of facies, or zones, which can be used as proxies for annual mass balance and also play a significant role in understanding glacier energy balance. However, multispectral sensors were not designed explicitly for snow and ice observation, so it is not self-evident that they should be optimal for remote sensing of glaciers. There are no universal techniques for glacier surface classification which have been optimized with in situ reflectance spectra. Therefore, the roles that the various spectral, spatial, and radiometric properties of each sensor play in the success and output of resulting classifications remain largely unknown. Therefore, this study approaches the problem from an inverse perspective. Starting with in situ reflectance spectra from the full range of surfaces measured on two glaciers at the end of the melt season in order to capture the largest range of facies (Midtre Lovénbreen, Svalbard & Langjökull, Iceland), optimal wavelengths for glacier facies identification are investigated with principal component analysis. Two linear combinations are produced which capture the vast majority of variance in the data; the first highlights broadband albedo while the second emphasizes the difference in reflectance between blue and near-infrared wavelengths for glacier surface classification. The results confirm previous work which limited distinction to snow, slush, and ice facies. Based on these in situ data, a simple, and more importantly completely transferrable, classification scheme for glacier surfaces is presented for a range of satellite multispectral sensors. Again starting with in situ data, application of relative response functions, scaling factors, and calibration coefficients shows that almost all simulated multispectral sensors (at certain gain settings) are qualified to classify glacier accumulation and ablation areas but confuse classification of partly ash-covered glacier surfaces. In order to consider the spatial as well as the spectral properties of multispectral sensors, airborne data are spatially degraded to emulate satellite imagery; while medium-resolution sensors (~20-60 m) successfully reproduce high-resolution (2 m) observations, low-resolution sensors (i.e. 250 m+) are unable to do so. These results give confidence in results from current sensors such as ASTER and Landsat ETM+ as well as ESA’s upcoming Sentinel-2 and NASA’s recently launched LDCM. In addition, images from the Landsat data archive are used to classify glacier facies and calculate the albedo of glaciers on the Brøgger Peninsula, Svalbard. The time series is used to observe seasonal and interannual trends and investigate the role of melt-albedo feedback in thinning of Svalbard glaciers. The dissertation concludes with recommendations for glacier surface classification over a range of current and future multispectral sensors. Application of the classification schemes suggested should help to improve the understanding of recent and continuing change to GIC around the world.

PURPOSE. To image the fundus non-invasively at two different penetration depths using a multispectral imaging system. Monochromatic images at visible spectrum wavelengths and near-infrared wavelengths were qualitatively assessed for choroidal visibility. These images were used calculate oxygen saturation in retinal tissue, optic nerve head tissue, vein, and choroidal tissue in healthy controls and glaucoma patients. METHODS. A fundus camera-based multispectral snapshot oximeter imaged the fundus of healthy subjects and patients with varying ophthalmological pathology. The images of healthy controls and glaucoma patients were analyzed to determine oxygen saturation in the optic nerve head cup and rim, superficial and deep vein, macula and choroidal tissue. RESULTS. Visible: Average oxygen saturation for the ONH cup was 65 ± 6 percent for healthy controls and 61 ± 10 percent for glaucoma patients. For the ONH rim, it was 67 ± 3 percent for healthy controls and 64 ± 17 percent for glaucoma patients. For the vein, it was 67 ± 15 percent for healthy controls and 56 ± 22 percent for glaucoma patients. For the macula, it was 87 ± 10 percent for healthy controls and 93 ± 1 percent for glaucoma patients. NIR: The average oxygen saturation for the vein was 66 ± 20 percent for healthy controls, 58 ± 0.4 percent for glaucoma suspects and 54 ± 17 percent for glaucoma patients. For the choroidal tissue below the macula, it was 99 ± 5 percent in healthy controls and 81 ± 8 percent in glaucoma patients. CONCLUSIONS. Choroidal visibility is enhanced in near infrared monochromatic images from visible spectrum monochromatic images. Oxygen saturation results were lower in glaucoma patients for all anatomical areas analyzed except the avascular macula. / acase@tulane.edu

Multispectral imaging (MSI) provides a wealth of imagery data that, together with modern signal processing techniques, facilitates the enhancement of document images. In this thesis, four topic areas are reviewed and applied to ancient documents. They are image fusion, matched filters, bleed-through removal, and shadow removal. These four areas of focus provide useful tools for papyrologists studying the digital imagery of documents. The results presented form a strong case for the utility of MSI data over the use of a single image captured at any given wavelength of light.

Multispectral aerial videography was used to reproduce the jurisdictional delineation of wetland area of approximately 50 hectares in Davis County, Utah Imagery from the system consisted of three-band composite with wavelengths covering 550 nm (±10 nm), 650 nm (±10 nm), and 850 nm (±10 nm). The site was overflown at three different flight dates during the 1992 growing season (June 2, July 22, October 1). Imagery resolution varied from 0.56 m to 0.81 m. Mosaiced images were analyzed with a Supervised clustering/maximum likelihood classifier, ISODATA clustering/Euclidan classifier, statistical clustering/maximum likelihood classifier, and fuzzy c-means clustering. Overall accuracies for wetland/upland designations as compared to ground truth data varied from 60% to 75%. The ISODATA method was the poorest performer for all dates and both of two accuracy testing techniques. Supervised clustering and statistical clustering were comparable with a slight edge in accuracy to the supervised clustering. The best all-round performer was the fuzzy c-means algorithm in terms of time spent and accuracy.

Cotton crop condition was assessed from an analysis of multispectral aerial video imagery. Visible-near infrared imagery of two cotton fields was collected towards the end of the 1990 crop. The digital analysis was based on image classification, and the accuracies were assessed using the Kappa coefficient of agreement. The earliest of three images proved to be best for distinguishing plant variety. Vegetation index images were better for estimating potential yield than the original multispectral image; so too were multi-channel images that were transformed using vegetation indices or principal component analysis. The seedbed preparation rig used, the nitrogen application rate and three plant varieties, a weed species and two cotton cultivars, could all be discriminated from the imagery. Accuracies were moderate for the discrimination of plant variety, tillage treatment and nitrogen treatment, and low for the estimation of potential yield.

One computer disk in pocket inside back cover. / System requirements for accompanying computer disk: Macintosh computer. / Bibliography: leaves 148-160. / xii, 161 leaves : ill. (some col.) ; 30 cm. + 1 computer disk (3.5 in. DD) / Title page, contents and abstract only. The complete thesis in print form is available from the University Library. / This thesis presents two new families of classification algorithms for pixel classification based on multi-spectral texture. The research demonstrates that algorithms making use of multispectral texture can be constructed that produce better classifications than standard algorithms at comparable computational cost. / Thesis (Ph.D.)--University of Adelaide, Dept. of Electrical and Electronic Engineering, 1994?

The main focus of this dissertation is to develop a multispectral imaging system with spectral zooming capability and also successfully demonstrate its promising medical applications through combining this technique with microscope system. The realization of the multispectral imaging method in this dissertation is based on the 4-f spatial filtering principle. When a collimated light is dispersed by the grating, there exists a clear linear distribution spectral line or spectrum at the Fourier plane of the Fourier transform lens group base on the Abbe imaging theory and optics Fourier Transform principle. The optical images, not the collimated light, are applied into this setup and the spectrum distribution still keeps linear relationship with the spatial positions at Fourier plane, even through there exists additional spectral crosstalk or overlap. The spatial filter or dynamic electrical filters used at the Fourier plane will facilitate randomly access the desired spectral waveband and agilely adjust the passband width. It offers the multispectral imaging functionality with spectral zooming capability. The system is flexible and efficiency. A dual-channel spectral imaging system based on the multispectral imaging method and acousto-optical tunable filter (AOTF) is proposed in the dissertation. The multispectral imaging method and the AOTF will form spate imaging channels and the two spectral channels work together to enhance the system efficiency. The AOTF retro reflection design is explored in the dissertation and experimental results demonstrate this design could effectively improve the spectral resolution of the passband. Moreover, a field lens is introduced into the multispectral imaging system to enhance the field of view of the system detection range. The application of field lens also improves the system spectral resolution, image quality and minimizes the system size. This spectral imaging system can be used for many applications. The compact prototype multispectral imaging system has been built and many outdoor remote spectral imaging tests have been performed. The spectral imaging design has also been successfully applied into microscope imaging. The prototype multispectral microscopy system shows excellent capability for normal optical detection of medical specimen and fluorescent emission imaging/diagnosis. Experiment results have demonstrated this design could realize both spectral zoom and optical zoom at the same time. This design facilitates fast spectral waveband adjustment as well as increasing speed, flexibility, and reduced cost.

This thesis investigated multispectral detection of the invasive floating macrophyte, European Frog-bit, using Quickbird imagery and fuzzy image classification. To determine if the spectral signature of European Frog-bit were separable from other wetland vegetation, a species level land cover classification was conducted on a 6km section of the South Nation River in Ontario, Canada. Supervised and unsupervised imagery classification approaches were evaluated using the fuzzy classifiers, Fuzzy Segmentation for Object Based Image Classification (FS) and Fuzzy C-Means (FCM). Both approaches were sufficiently robust to detect European Frog-bit. User’s and producer’s accuracies for the European Frog-bit class were 81.0% and 77.9% for the FS classifier and 63.5% and 73.0% for the FCM classifier. These accuracies indicated that the spectral signature of EFB was sufficiently different to permit detection and separation from other wetland vegetation and fuzzy image classifiers were capable of detecting EFB in Quickbird imagery.

This thesis investigated multispectral detection of the invasive floating macrophyte, European Frog-bit, using Quickbird imagery and fuzzy image classification. To determine if the spectral signature of European Frog-bit were separable from other wetland vegetation, a species level land cover classification was conducted on a 6km section of the South Nation River in Ontario, Canada. Supervised and unsupervised imagery classification approaches were evaluated using the fuzzy classifiers, Fuzzy Segmentation for Object Based Image Classification (FS) and Fuzzy C-Means (FCM). Both approaches were sufficiently robust to detect European Frog-bit. User’s and producer’s accuracies for the European Frog-bit class were 81.0% and 77.9% for the FS classifier and 63.5% and 73.0% for the FCM classifier. These accuracies indicated that the spectral signature of EFB was sufficiently different to permit detection and separation from other wetland vegetation and fuzzy image classifiers were capable of detecting EFB in Quickbird imagery.