Artificial intelligence model to assess root-lesion index of nematode infected banana roots Suryaprabha D.*, Seenivasan N.1 Department of Computer Applications, Nehru Arts and Science College, Coimbatore - 641105, Tamil Nadu, India 1Department of Nematology, Tamil Nadu Agricultural University, Coimbatore - 641003, Tamil Nadu, India. *Corresponding author; E-mail: nascdrdsuryaprabha@nehrucolleges.com
Online Published on 29 June, 2022. Abstract Lesion forming nematodes, Radopholus similis, Pratylenchus coffeae and Helicotylenchus multicinctus are a global threat for successful banana cultivation. They cause typical necrotic lesions and black colour discoloration in the root cortex. Root lesion index (RLI) is a parameter to measure the injury caused by these nematodes at 0–4 scale. RLI also serves as an important criterion to rate host susceptible or resistance reaction of banana against lesion forming nematodes. Manual assessment of RLI by visual observation is a difficult, time-consuming and error-prone task. Hence, an image processing-based Artificial Intelligence (AI) model has been developed to automate RLI assessment jobs. Images of banana roots in 0–4 scale RLI at 500 images for each category were collected. Four hundred images of each set were used for AI model development and the remaining 100 were used for testing or validating the developed model. Development of the RLI-AI model involved the following steps such as I) Collection of 0–4 scale RLI images, 2) Noise reduction and normalizing of image sets by fuzzy intensification operator and genetic algorithm, 3) Separation of roots from image background and identification of infected region by flower pollination based edge detection algorithm, 4) Extracting the required region of interest using the gray-level co-occurrence matrix, 5) Classification of 0–4 scale roots by SVM classifier, and 6) Testing the AI model using validation image sets.The overall accuracy of the RLI-AI model was 96.4%. This model is relatively more accurate than manual RLI assessment. To make the automation process easy, the RLI-AI model was embedded in Graphical User Interface Development Environment (GUIDE). Hence, the users can get the RLI scale of nematode infected roots by simply uploading the images of the root and clicking the input analysis button. The user-friendly RLI-AI model is useful for farmers, extension officials and researchers for nematode damage assessment as well as screening the banana for nematode resistance. Top Keywords Banana, Nematode, Root lesion index, Automation. Top |