LARS Tech Report Number
Multispectral imagery collected over southeastern Pennsylvania by Willow Run Laboratories, University of Michigan, was analyzed at the Laboratory for Applications of Remote Sensing (LARS) using its current capabilities for automatic classification. The project involved the evaluation of imagery in thirteen discrete bands of the spectrum as a source of data on engineering soils. Detailed computer classifications of two 4-mile segments of the Pennsylvania flightline, each containing predominantly one parent soil material, were obtained. Classification accuracy as measured by training field and test field performance was in excess of 90%. Using these two segments as a basis, a computer-implemented map was obtained for 12 miles of the flightline.
Soil pattern recognition using the LARS approach predates this study but most previous work differed from this in its objectives and analysis procedures. Agricultural soil studies have involved detailed mapping of relatively small areas (about 100 acres) located in the glaciated areas of the midwestern corn belt where topographic relief is low. A previous engineering soils study (Tanguay and Miles), in which the LARS and other remote sensing techniques were considered, was made over the same general terrain as the agricultural studies using the LARS techniques available at that time (1969). The Pennsylvania flightline, over a region of bedrock-derived residual soils having moderate relief, presented a new challenge to the continually evolving LARS techniques.
Date of this Version