SOIL-MOISTURE SENSOR FOR PREDICTING SEED PLANTING DEPTH

A near-infrared (NIR) sensor was built to predict corn seed planting d epth based on moisture content and matric potential. The sensor uses t hree NIR wavelengths and a maximum likelihood classifier to predict a ''plant deeper'' or ''plant at current depth'' judgment. The sensor wa s trained on 29 different soils that varied in soil texture and organi c matter content and was able to predict the -10, -30, -50 kPa potenti als from :the -100 and -1500 kPa potentials with 82% accuracy. The sys tem was tested in the field on a silt loam soil and predicted a ''plan t at current depth'' for moisture contents above 19.44% and a ''plant deeper'' for moisture contents below 19.44% with 82% accuracy. The moi sture contents for the field rest ranged from 7 to 32%.