Hyperspectral image analysis for water stress detection of apple trees

Plant stress significantly reduces plant productivity. Automated on-the-go mapping of plant stress would allow for a timely intervention and mitigation of the problem before critical thresholds were exceeded, thereby maximizing productivity. The spectral signature of plant leaves was analyzed by a hyperspectral camera to identify the onset and intensity of plant water stress. Five different levels of water treatment were created in young apple trees (cv. ‘Buckeye Gala’) in a greenhouse. The trees were periodically monitored with a hyperspectral camera along with an active-illuminated spectral vegetation sensor and a digital color camera. Individual spectral images over a 385–1000 nm wavelength range were extracted at a specific wavelength to estimate reflectance and generate spectral profiles for the five different water treatment levels. Various spectral indices were calculated and correlated to stress levels. The highest correlation was found with Red Edge NDVI at 705 and 750 nm in narrowband indices and NDVI at 680 and 800 nm in broadband indices. The experimental results indicated that intelligent optical sensors could deliver decision support for plant stress detection and management.

► Hyperspectral image is critically affected by illumination changes and must be carefully calibrated during image capture.
► In narrowband indices, Red Edge NDVI at 705 and 750 nm is most sensitive to plant water stress.
► In broadband indices, NDVI at 680 and 800 nm is most sensitive to plant water stress.
► Active spectral NDVI sensor and canopy coverage can be used to identify plant growth changes due to water stress.