Design and implementation of a computer vision-guided greenhouse crop diagnostics system

An autonomous computer vision-guided plant sensing and monitoring system was designed and constructed to continuously monitor temporal, morphological, and spectral features of lettuce crop growing in a nutrient film technique (NFT) hydroponics system. The system consisted of five main components including (1) a stepper motor-driven camera positioning system, (2) an image acquisition system, (3) a data logger monitoring root and aerial zone of the growing environment, (4) a dynamic SQL database module for data storage, and (5) a host computer running the collection, processing, storage, and analysis functions. Panoramic canopy images were dynamically created from the images collected by color, near-infrared (NIR) and thermal cameras. From these three images, the crop features were registered such that a single extracted crop (or a crop canopy) contained information from each layer. The extracted features were color (red–green–blue, hue-saturation-luminance, and color brightness), texture (entropy, energy, contrast, and homogeneity), Normalized Difference Vegetative Index (NDVI) (as well as other similar indices from the color and NIR channels), thermal (plant and canopy temperature), plant morphology (top projected plant and canopy area), and temporal changes of all these variables. The computer vision-guided system was able to extract these plant features and stored them into a database autonomously. This paper introduces the engineering design and system components in detail. The system’s capability is illustrated with a one-day sample of the lettuce plants growing in the NFT system, presenting the temporal changes of three key crop features extracted, and identification of a stress level and locality detection as example applications.