SEARCH · Search NASA
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Examined are: (1) the sequential clustering technique for the unsupervised automatic classification and mapping of earth resources satellite data, (2) theoretical analysis of the tests which were used, and (3) derivation of an alternative set of tests and their necessary algorithm.
Changes in land use in the Phoenix (1:250,000 scale) Quadrangle in Arizona have been mapped using only the images from ERTS-1, tending to verify the utility of a land use classification system proposed for use with ERTS images. The period of change investigated was from November 1970 to late summer or early fall, 1972. Seasonal changes also were studied using successive ERTS images. Types of equipment used to aid interpretation included a color additive viewer, a twenty-power magnifier, a density slicer, and a diazo copy machine for making ERTS color composites in hard copy. Types of changes detected have been: (1) cropland or rangeland developed for new residential areas; (2) rangeland converted to new cropland; and (3) possibly new areas of industrial or commercial development. A map of land use previously compiled from air photos was updated in this manner.
Land use detail in the CARETS area obtainable from ERTS exceeds the expectations of the Interagency Steering Committee and the USGS proposed standardized classification, which presents Level 1 categories for ERTS and Level 2 for high altitude aircraft data. Some Levels 2 and 3, in addition to Level 1, categories were identified on ERTS data. Significant land use changes totaling 39.2 sq km in the Norfolk-Portsmouth SMSA were identified and mapped at Level 2 detail using a combination of procedures employing ERTS and high altitude aircraft data.
The objective of canonical analysis is to obtain the maximum separability among a number of catergories. The application of canonical analysis was investigated using the merged MSS ERTS-1 data for one area viewed on two dates. The effect of threshold values on classification regions and confusion regions was investigated.
The author has identified the following significant results. Results of the continuing analysis of ERTS-1 imagery covering Iowa during 1972 and periods during 1973 are covered. Emphasis is placed on the identification and classification of major crop types at two test sites in Iowa. Standard photointerpretive methods were used in this analysis including the direct enlargement of black and white single-band products and additive color multi-band procedures using a miniadcol system. The use of sequential coverage during the crop growing season is emphasized as a means to improve the effectiveness of ERTS-1 photointerpretations of crop land acreage estimates in Iowa. Illustrative black and white and color prints of both ERTS-1 and underflight imagery are included. In addition, forest land inventories at one test site are reported. A new method for the inventory of forest lands using ERTS-1 imagery is reported and compared with estimates obtained using earlier underflight imagery.
Statistical methods for pattern recognition and classification applications
The potential use is discussed of three remote sensors for developing a four level land use classification system. Three types of imagery for photointerpretation are presented: ERTS-1 satellite imagery, high altitude photography, and medium altitude photography. Suggestions are given as to which remote sensors and imagery scales may be most effectively employed to provide data on specific types of land use.
This report examines the spatial and spectral clustering technique for the unsupervised automatic classification and mapping of earth resources satellite data, and makes theoretical analysis of the decision rules and tests in order to suggest how the method might best be applied to other flight data such as Skylab and Spacelab.
The author has identified the following significant results. Recent signature identification and refinement techniques indicate that with automated classification of MSS CCT data commercial stands of cottonwood and white spruce can be identified with 80% accuracy in the Bonanza Creek experimental forest. Since that forest is representative of the vast interior Alaska forests, this finding has substantial economic importance to public and private forestry interests in Alaska.
The prime emphasis was on classification of pixels in field centers, away from boundary effects. Results were encouraging in both training and test field centers for wheat and other major types of vegetation present. However, the location of fields was found to be a serious problem and it was even more difficult to select field-center pixels for fields of sizes less than 20 acres (or even larger, depending upon field shape) for use in the field-center analysis. The majority of fields in the segment are less than 20 acres in size. ERTS-1 data were received on 12 September 1973. Ground truth information and aerial photography were received on 9 and 15 September. The data were analyzed and processed digitally using the ERIM multispectral software system.
A comprehensive study has been undertaken to determine the extent to which conventional image interpretation and computer-aided (spectral pattern recognition) analysis techniques using ERTS-1 data could be used to detect, identify (classify), locate, and measure current land use over large geographic areas. It can be concluded that most of the level 1 and 2 categories in the USGS Circular no. 671 can be detected in the Houston-Gulf Coast area using a combination of both techniques for analysis. These capabilities could be exercised over larger geographic areas, however, certain factors such as different vegetative cover, topography, etc. may have to be considered in other geographic regions. The best results in identification (classification), location, and measurement of level 1 and 2 type categories appear to be obtainable through automatic data processing of multispectral scanner computer compatible tapes.
A method is presented for using B-splines as potential functions in the estimation of likelihood functions (probability density functions conditioned on pattern classes), or the resulting discriminant functions. The consistency of this technique is discussed. Experimental results of using the likelihood functions in the classification of remotely sensed data are given.
The phenomenal increase in the amount of data and information being generated by remote sensing systems is stressed. A total system design approach as a solution to this problem is discussed with specific reference to the data and information system needs for Sortie Lab - a multiple use payload for the Shuttle. The development of a multispectral data processing system as a needed component of such a system is reviewed with emphasis on unsupervised multispectral classification techniques developed and presently in use at Marshall Space Flight Center.
Presented here is an algorithm that partitions a digitized multispectral image into parts that correspond to objects in the scene being sensed. The algorithm partitions an image into successively smaller rectangles and produces a partition that tends to minimize a criterion function. Supervised and unsupervised classification techniques can be applied to partitioned images. This partition-then-classify approach is used to process images sensed from aircraft and the ERTS-1 satellite, and the method is shown to give relatively accurate results in classifying agricultural areas and extracting urban areas.
A method based on the use of spectral variations in combination with spatial variations is developed for automatic boundary finding and sample classification of remotely sensed multispectral data. Preliminary applications of the method to agricultural data show significant improvements in accuracy as compared to the use of spectral data alone.
Description of some easily computable textural features based on gray-tone spatial dependances, and illustration of their application in category-identification tasks of three different kinds of image data - namely, photomicrographs of five kinds of sandstones, 1:20,000 panchromatic aerial photographs of eight land-use categories, and ERTS multispectral imagery containing several land-use categories. Two kinds of decision rules are used - one for which the decision regions are convex polyhedra (a piecewise-linear decision rule), and one for which the decision regions are rectangular parallelpipeds (a min-max decision rule). In each experiment the data set was divided into two parts, a training set and a test set. Test set identification accuracy is 89% for the photomicrographs, 82% for the aerial photographic imagery, and 83% for the satellite imagery. These results indicate that the easily computable textural features probably have a general applicability for a wide variety of image-classification applications.
A possible solution to the analysis of the massive amounts of multi-spectral scanner data from the Earth Resource Technical Satellite (ERTS) program is proposed. This solution is offered as an adaptive on-line classification scheme. The classifier is described as well as its controller which is based on ground truth data. Cluster analysis is presented as an alternative approach to the ground truth data. Adaptive feature selection is discussed and possible mini-computer implementations are offered.
A new clustering algorithm is presented that is based on dimensional information. The algorithm includes an inherent feature selection criterion, which is discussed. Further, a heuristic method for choosing the proper number of intervals for a frequency distribution histogram, a feature necessary for the algorithm, is presented. The algorithm, although usable as a stand-alone clustering technique, is then utilized as a global approximator. Local clustering techniques and configuration of a global-local scheme are discussed, and finally the complete global-local and feature selector configuration is shown in application to a real-time adaptive classification scheme for the analysis of remote sensed multispectral scanner data.