Characterizing Water Deficiency induced stress in Plants using Gabor filter based CNN

With the growing population, it has become necessary to increase the agriculture yield too. Chickpea is one of the most important crops in developing countries like India and is an important source of proteins and carbohydrates for a significant part of the population. Abiotic stresses like that produced due to lack of water greatly effect the health of the chickpea plants and the respective yields. As such, being able to monitor the plants for any water related stress can play a major role in enhancing their health and the overall yield. Plant phenotyping approaches that combine non-invasive image analysis and machine learning have been effectively employed to identify and quantify plant health and illnesses in the last decade. In this work, we propose a Gabor filter based CNN model to classify the chickpea plants into different water stress conditions and compare the results with the previous proposed methods. From the experimental results, we conclude that a 12-layer Gabor based CNN model have accuracies of 79% and 77% for the JG and Pusa varieties of chickpea, respectively, that is better than the 23-layer CNN model proposed in a recent work.