Quadratic discriminant feature selected broken stick regressive deep convolution neural learning classification for...

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Accurate turmeric crop disease classification and yield prediction are crucial for modern agriculture.
• Traditional methods often lack the precision required for optimal resource management and decision-making.

Purpose of the Study:

• To propose a novel technique, Quadratic Discriminant Feature Selected Broken Stick Regressive Deep Convolution Neural Learning Classification (QDFSBSRDCNLC), for turmeric crop disease classification and yield prediction.
• To evaluate the performance of different deep learning models for semantic segmentation of turmeric crop diseases.

Main Methods:

• Utilized Quadratic Discriminant Analysis (QDA) for feature selection and dimensionality reduction.
• Employed four semantic segmentation models: FCN8, PSP Net, MobileNetV3 (small), and Deep Lab V3.
• Collected and analyzed images of healthy and diseased turmeric crops from a dedicated research field.

Main Results:

• MobileNetV3 (small) achieved the highest performance among the tested models.
• Achieved an accuracy of 97.99%, Intersection over Union (IoU) of 96.82%, and a Coefficient of 97.80% for MobileNetV3 (small) after 50 epochs.
• The QDFSBSRDCNLC technique effectively classified diseases and predicted turmeric crop yields.

Conclusions:

• The proposed QDFSBSRDCNLC technique offers an effective approach for turmeric crop disease management and yield forecasting.
• MobileNetV3 (small) is a highly effective model for semantic segmentation of turmeric crop diseases, contributing to improved agricultural practices.