In-field crop physiology sensing aided real-time apple fruit surface temperature monitoring for sunburn prediction

Abstract

Heat and light-instigated abiotic stresses during summer can cause several physiological disorders in perennial specialty crops. Such stressors increase the fruit surface temperature (FST) and prolonged exposure above a critical FST can result in sunburn. Sunburn in apple may cause considerable crop loss and reduce fresh produce marketability. Existing approaches for sunburn prediction and management, based on atmospheric temperature data, are often unreliable and inefficient for timely actuation of remedial measures. Therefore, this study focuses on the development of a non-invasive and real-time sunburn monitoring tool. We developed a crop physiology sensing (CPS) unit that uses visible-infrared imagery and in-field weather data for FST monitoring, the prime indicator of sunburn susceptibility. The CPS unit consists of a thermal-red-green-blue and all-in-one weather sensor integrated with a single-board computer. Acquired imagery data was analyzed in real-time using a custom-developed algorithm in python ‘OpenCV’ library to estimate imager-based FST. The algorithm was optimized for processing the data on a single-board computer with limited computational resources. Moreover, the processing unit was configured to acquire in-field weather data and to utilize a temperature dynamics weather model for weather-based FST estimation. Two automated CPS units were deployed in the commercial orchards of cv ‘Honeycrisp’ and ‘Cosmic Crisp™’ during the 2019 production season. For each cultivar, field data was collected for three days between 12 and 5 pm at 5-minute intervals. A contact type thermal probe of accuracy ±0.4 °C was also utilized for ground truth apple FST (FSTa) measurements. Furthermore, imagery data was analyzed to derive mean FST (FSTi), maximum FST (FSTi-max), and mean FST of the 10%, 15% and 20% hottest part of the fruit surface (i.e. FST10, FST15, and FST20, respectively). The results showed significant differences between FSTi, FSTi-max, FST10, FST15, and FST20 for Honeycrisp (F4,162 = 73.4, p < 0.0001) as well as Cosmic Crisp (F4,46 = 19.4, p < 0.0001) cultivars. Moreover, no significant difference was recorded between FSTi and FSTa. The weather model-predicted FST (FSTw) was found to be highly sensitive to fruit shading and a significant difference was recorded between FSTw and FSTi. Overall, the developed CPS unit demonstrates a promising potential for reliable FST monitoring that could aid growers in real-time apple sunburn susceptibility prediction and actuation of sunburn preventive strategies.