Using CNN to Identify NPK Deficiencies in Paddy Fields: An Advanced Detection Method

Abstract

Our study introduces a cutting-edge framework employing imaging technology to identify nutrient deficiencies in rice plants, focusing on nitrogen (N), phosphorus (P), and potassium (K). We utilize various image datasets of rice plant leaves showing symptoms of these deficiencies to train a Convolutional Neural Network (CNN). The CNN's capacity for learning from diverse data inputs makes it ideal for this complex task. A crucial aspect of our methodology is the use of a pre-trained CNN. The early layers of this network are particularly adept at extracting distinct features from the images of paddy leaves, enabling precise identification of specific nutrient deficiencies. When introduced to the trained CNN model, a new test image can accurately determine whether the leaf is deficient in nitrogen, phosphorus, or potassium. Implementing this image-based nutrient deficiency detection method has significant implications for agriculture. It provides farmers a non-invasive and efficient tool to identify crop nutrient imbalances. This empowerment enables better-informed fertilization decisions, potentially leading to improved crop yields and more sustainable farming practices. Our approach achieves an impressive accuracy of 96.67%, demonstrating its effectiveness. Therefore, our method marks a substantial advancement in agricultural technology, offering a valuable solution for enhancing plant health and agricultural productivity.