Deep Learning-Based Nutrient Deficiency Analysis in Plant Leaves

Abstract

Abstract: Plants frequently face nutrient deficiencies, which can lead to reduced yields, stunted growth, and inferior crop quality. The proposed approach involves dividing an image of a leaf into smaller blocks and then processing each block through convolutional neural networks (CNNs). Each CNN is specifically trained to identify a particular nutrient deficiency and evaluate whether the corresponding symptoms are present in the block. It employs a technique to combine all CNN responses into a single block response. Ultimately, the proposed approach combines the responses from individual blocks by employing a multi-layer perceptron to generate a final response for the entire leaf. The study focused on various types of deficiencies, including calcium, iron, potassium, magnesium, and nitrogen deficiencies, as well as complete nutrition leaves. According to the experimental results, the method we proposed exhibited superior performance compared to that of humans trained to identify nutrient deficiencies.