Crop surveys from multiband satellite photography using digital techniques

Multiband and Multibase photography of the Imperial Valley, California, taken during the Apollo 9 mission was analyzed using statistical multispectral pattern recognition techniques to achieve highly accurate detection of bare soil, salt flats, and water.

Abstract

Multiband and Multibase photography of the Imperial Valley, California, taken during the Apollo 9 mission was analyzed using statistical multispectral pattern recognition techniques. With a scanning microdensitometer, 70-mm frames were converted into digital form and stored on magnetic tape. All analysis was conducted using a digital computer which enabled rapid automatic processing of any of over five million film density measurements. Automatic recognition of fields of barley, sugar beets, and alfalfa was attempted, and the recognition accuracies were 75% for barley, 59% for sugar beets, and 27% for alfalfa. Highly accurate detection of bare soil, salt flats, and water was achieved. Low percentage ground cover crops such as lettuce, onions, carrots, and cabbage were indistinguishable from bare soil. Results using multiband photography are compared with results using color separations of color infrared photography which covered approximately the same bands.