1977 Machine Processing of Remotely Sensed Data Symposium. The work reported in this paper was sponsored by NASA under contract No. NAS9-14016.


Aircraft and Landsat data were used with computer-aided techniques to delineate soils patterns of a field of 40 ha in a transition zone between soils developed under deciduous forest and those developed under prairie vegetation. Two computer-aided classification techniques, supervised and nonsupervised, were employed in classifying soils of the study area. The means and covariance matrix statistics were obtained for every cluster or soil class through the statistics algorithm. Each cluster of aircraft and Landsat data was identified and assigned to a specific soil type by correlating the cluster soil patterns with a standard soils map of the test site which was prepared as a part of the ground observation task. A sampling grid plan was used to select a training set for a supervised classification of the aircraft MSS data. The spectral soil patterns revealed in the classifications from aircraft and satellite MSS data resembled the general patterns of the soils of the conventionally prepared soil map. The spatial resolution of the aircraft scanner was adequate to recognize each soil type boundary in the test site. However, the limited spatial resolution of the satellite scanner made it difficult to delineate those soil features with widths less than the spatial resolution of the scanner. On the contrary those soil patterns which were broad enough to exceed the spatial resolution of the Landsat scanner were delineated very well.

Date of this Version