Identification of Water-Stressed Area in Maize Crop Using Uav Based Remote Sensing

Abstract

Agronomic inputs such as water , nutrients and fertilisers play a vital role in the health, growth and yield of crops. The lack of each of these inputs induces biotic and abiotic stress in the crop. Farmers are relying on groundwater because of decreased rainfall. The irrigation method can be improved by acquiring awareness of the health of crops and soils. In general, crop and soil quality is controlled by means of manual observation, which is time-consuming, labour-intensive and contributes to incorrect choices and substantial waste of resources. There is also an immediate need to automate the inspection process that will finally benefit farmers and agricultural scientists. In this paper, the identification of the water-stressed areas in the crop(maize) field has been studied, and an Unmanned Aerial Vehicle (UAV) based remote sensing is used to automate the crop health-monitoring process. We proposed a framework (model) based on Convolutional Neural Networks (CNN) to identify the stressed and normal/healthy areas in the maize crop field. The performance of the proposed framework has been compared with different models of CNN, such as ResNet50, VGG-19, and Inception-v3. The results show that the proposed model outperforms the baseline models and successfully classify stressed and normal areas with 95 % accuracy on train data and 93 % accuracy with 0.9370 precision and 0.9403 F1 score on test data.