Hyperspectral Imaging for Detecting Water Stress in Potatoes

Water stress causes significant reduction in plant productivity. Soil moisture content level must be monitored and maintained at adequate level for optimal productivity. Accuracy of moisture sensors used for monitoring soil moisture content depends on installation technique and proper contact between soil and sensor which is difficult to achieve. Non-contact sensing technique on the other hand, does not have the limitation of contact with soil and can be located in remote location to monitor plant status parameters continuously. In this study, hyperspectral imaging was used as a non-contact method of detecting changes in spectral reflectance of potato plants with different water soil moisture content levels. An experiment was carried out to apply various treatments of water to potato plants planted in a greenhouse. A hyperspectral camera with waveband range from 400nm to 1000nm was used to acquire spectral images of leaves and canopies of these plants. Various indices were evaluated using spectral reflectance data at different water stress levels. Spectral indices showing strong correlation with plant water stress were chosen and used for developing a regression model to predict soil moisture content level. Red Edge NDVI, Modified NDVI, Modified Red Edge SRI, Vogelmann Red Edge Index (VOG REI) 1, VOG REI 2 and VOG REI 3 were found to be strongly correlated with soil moisture content. Highest correlation coefficient was found to be -0.9 for VOG REI 1 followed by 0.886 for VOG REI 2. Based on a second order regression model between soil moisture content and dry tuber weight, the soil moisture content level for maximum yield was found to be 17%. These results showed promise for development of spectral sensor for non-contact soil-moisture content monitoring system, which may lead to automated irrigation system for maintaining an optimal soil moisture content level during potato growing season.