Recent Advances in Plant Leaf Disease Detection: A Survey on Control Strategies

Agriculture has a large contribution in the Indian economy, and crop diseases are a usual occurrence due to change in climate. They reduce crop quality and quantity, and present a huge problem to farmers especially with the emerging weather trends. It has been established that image processing is effective in early plant disease diagnosis in terms of leaf texture, color and shape. Different techniques used for extraction of features include; the histograms, Sobel edge detectors, gray level co-occurrence matrices and the Gabor wavelets to establish the disease markers. SVM, ANN, CNN and others are applied once extracts to classify other diseases. The cause of these diseases varies and the appropriate treatment can therefore be applied. Despite the effectiveness demonstrated by the current image processing and classification methods, the process of feature extraction is lab-or-intense and time consuming and usually involves manual input. More importantly, these techniques are best applied in a small data environment. This paper has shown that CNNs, a promising deep learning technique, have found considerable application in the automated recognition of plant diseases. Several applications using CNNs have been designed to diagnose plant leaf diseases, providing hope for expert systems that improve sustainable agriculture and food security. This paper also presents various research works that have employed CNNs in this area, assesses the efficacy and execution of CNNs. The survey outlines some of the difficulties and restrictions with present research studies and offers potential future developments for disease identification and management in plants. The paper also looks into various plant diseases and their causes, manifestations and treatments as well as provides suggestions for enhancing current practices