Plant Leaf Disease Prediction Based On Deep Learning Using R2NN-WRS: Resnet Recurrent Neural Network and Watershed Region Segmentation Techniques

  • Unique Paper ID: 159185
  • Volume: 9
  • Issue: 11
  • PageNo: 936-941
  • Abstract:
  • Plants are prone to various diseases during the growing season. It has a practical impact on global food security and the agricultural economy. Early diagnosis of plant diseases is one of the most challenging problems in agriculture. Not detecting the condition early can affect the overall yield and reduce farmers' profitability. However, agronomists and plant pathologists have traditionally used the naked eye test to detect leaf diseases. This traditional method of plant foliar disease detection is subjective, time-consuming, expensive, and requires a large number of personnel and a lot of information about plant disease. To tackle this problem, in this project we design Resnet Recurrent Neural Network (R2NN) algorithm is used to find plant disease. This first step is pre-processing using the Gaussian filter to enhance image quality. Then we apply Contrastive Limited Adaptive Equalization (CLAE) algorithm to improve image contrast. Furthermore, we use Watershed Region Segmentation (WRS) technique to segregate the affected parts. Later, the R2NN algorithm effectively classifies the plant disease. We show experimentally that our R2NN approach is more robust and extraordinary to generalize to unseen infected plant disease domain images than classical techniques. We also analyze the focus of attention as learned by our R2NN and show that our approach is capable of accurately locating infectious diseases in plants. Our approach has been tested on many plant species, so thus, the proposed method contributes to a more effective means of detecting and classifying plant disease.

Related Articles