A deep learning-based model for plant lesion segmentation, subtype identification, and survival probability...

Area of Science:

• Agricultural Science
• Computer Vision
• Artificial Intelligence

Background:

• Manual plant disease monitoring is labor-intensive and error-prone, leading to significant crop yield losses.
• Early detection of plant diseases is critical for timely intervention and mitigation of economic impacts.

Purpose of the Study:

• To develop an automated system for plant disease detection using computer vision and AI.
• To achieve accurate segmentation of leaf lesions and identification of disease subtypes.
• To predict plant survival based on detected diseases.

Main Methods:

• A context-aware 3D Convolutional Neural Network (CNN) based on CANet architecture for lesion segmentation.
• A Deep CNN for leaf lesion subtype recognition.
• A hybrid model combining CNN and Linear Regression for plant survival prediction.

Main Results:

• The lesion segmentation model achieved 92% accuracy and 90% IoU.
• Disease subtype recognition models reached accuracies of 91.11% (pepper), 93.01% (potato), and 99.04% (tomato).
• The system demonstrates high efficacy in identifying plant diseases and predicting survival.

Conclusions:

• The developed AI system offers a robust solution for real-time, automated plant disease detection.
• This technology can be deployed on unmanned aerial vehicles for crop health monitoring.
• The system has the potential to significantly reduce crop yield loss and improve agricultural management.