Development of an integrated plant-based image sensing system for soil-water and nitrogen status estimation in cotton

Optimal crop yields require optimisation of both water and nitrogen application. Industry-standard soil-water sensors require contact with the soil and provide information for a single point in the field although there is often spatial variability in soil type and crop growth. Nitrogen content is typically determined using destructive manual soil coring followed by laboratory testing. It is often not practical to install multiple soil-water sensors in a commercial field situation or to conduct multiple soil cores throughout the cotton season. A non-contact soil-water and nitrogen estimation system offers growers potential savings by optimising water and fertiliser management and efficiency and crop productivity. Existing non-contact approaches typically have low spatial resolution and cannot discriminate plants from soil. An alternative approach is a camera-based sensing system that estimates soil-water and plant nitrogen status. This project has developed a proof-of-concept infield sensing system and crop model to determine current and predict future soil-water, nitrogen and fruit load of cotton plants based on day of the season, weather data and visual plant response captured using cameras. Artificial intelligence was used to analyse the data and determine the model. These models have potential to be used instead of industry-standard models APSIM and OZCOT to predict crop production throughout the season as part of automated control systems to optimise irrigation and fertiliser application. The procedure used to developed model could be applied to any crop.