Remote Assessment of Size, Ripeness, and Count of Fruits in Orchard Greenhouse

We developed and validated a remote fruit recognition system designed for greenhouse orchards, integrating image processing with sensor-based monitoring and long-range (LoRa) wireless communication.Utilizing the open source computer vision library (OpenCV) and Jetson Nano, the system identifies fruit quantity, ripeness, and size from high-resolution images captured by the Pi Camera V2.1.The recognition algorithm incorporates hue-saturation-value color space conversion, morphological operations, and advanced parameters such as bounding box area, aspect ratio, and intersection over union to enhance accuracy.Validation was conducted using a scaled-down greenhouse simulation and kumquat tree images.The modified system achieved recognition accuracies of 90% for fruit count, 100% for unripe and half ripe fruits, 89% for fully ripe fruits, and 81 and 88% for large and small fruits, respectively.These results demonstrate significant improvements over the initial system and confirm the feasibility of integrating sensor technologies with image recognition for real-time agricultural monitoring.The system's compatibility with LoRa communication enables deployment in network-deficient environments, offering a scalable solution for precision agriculture.