JoWUA

Abstract

Throughout history, farmers and agricultural engineers have focused on the issue of increasing the yield of crops using different farming methods. In today’s digitalized world, these techniques have been combined with IoT technology and machine learning algorithms, which have given rise to smart agriculture systems. However, farmers who live in developing countries hesitate to use such systems because of their hardware and maintenance costs. To address this issue, this paper proposes a lowcost farmland digital twin framework called AgriLoRa for smart agriculture. AgriLoRa consists of a wireless sensor network established in the farmland and cloud servers that run computer vision algorithms to detect plant diseases, weed clusters and plant nutrient deficiencies. In order to assess the feasibility of accurate plant disease detection, we have performed initial experiments with agricultural vision datasets using two different algorithms, the MobileNet and UNet models, and achieved successful results. AgriLoRa is promising to achieve a low-cost, high-precision smart agriculture solution to address the growing high-yield production needs of farmers worldwide.