Chitkara University Researchers Achieve 99.99% Accuracy in Pest Classification Transforming Precision Agriculture

87

• Researchers at Chitkara University have developed an advanced AI model that achieves 99.99% accuracy in pest classification, marking a major leap in precision agriculture.
• The system enables farmers to quickly and accurately identify pests, reducing crop losses and minimising reliance on chemical pesticides.
• By integrating this technology, growers can adopt more targeted and eco-friendly pest management strategies, improving both yield and sustainability.
• The breakthrough positions India as a leader in agri-tech innovation, with potential for global adoption in smart farming solutions.

A pioneering study from Chitkara University in India has achieved a remarkable 99.99 per cent accuracy rate in pest classification, marking a significant leap forward in the field of precision agriculture. 

Published in the journal Scientific Reports, the research is led by Vikas Khullar of the Chitkara University Institute of Engineering and Technology and promises to revolutionise the way farmers monitor and protect their crops.

The study addresses a pressing challenge in agriculture: the time-consuming and error-prone nature of manual pest identification. By automating this process, Khullar and his team aim to provide farmers with faster, more reliable tools to safeguard yields. 

Their system integrates multiple pretrained deep learning models to extract visual features from pest images and applies Linear Discriminant Analysis (LDA) for feature selection. 

This hybrid approach significantly reduces computational and memory requirements, making it practical for remote and resource-limited settings.

“Our proposed method combines the strengths of several deep learning models, enabling us to manage a large number of pest classes effectively,” explained Khullar. “By selecting only the most relevant features, we reduce the computational load and deliver a lighter, more efficient solution for deployment in the field.”

The research drew on pest datasets combining nine and twelve classes, resulting in a comprehensive set of nineteen. Pretrained models—DenseNet201, EfficientNetB3, and InceptionResNetV2—were chosen for their efficiency. 

Features extracted from the second-to-last layers of these models were merged and refined through LDA, then classified using a lightweight dense neural network. 

(Read Also: Chongqing Scientists Unlock Potato Tuberisation Code for Sustainable Agriculture)

The outcome was striking: 99.99 per cent accuracy, 100 per cent validation, and 99.99 per cent recall with negligible loss. Comparative tests against benchmark approaches demonstrated the superiority of this hybrid system in both accuracy and efficiency.

The implications for farming are profound. Automated pest identification can cut labour costs, speed up pest management, and improve decision-making in real time. 

Such technology is particularly relevant for Smart Farming, where data-driven insights and automated systems underpin sustainable practices. By helping farmers protect crops more effectively, this research contributes to improved food security and reduced environmental impact.

Khullar emphasised the broader significance: “Our approach aligns with the goals of precision agriculture by delivering high accuracy while conserving resources.”

This breakthrough not only underscores the role of artificial intelligence in agriculture but also highlights how resource-efficient solutions are shaping the future of global food production.