Skip to content
Open access

A Deep Learning-Based Real-Time Surveillance Framework for Mitigating Herbivorous Animal Intrusions in Rice Nursery Fields

2026 · International Journal of Advanced Computer Science and Applications · 0 citations · 25 references

Abstract

The rice nursery stage is a very critical stage of the rice growing cycle, as intrusion of herbivorous animals can lead to substantial economic losses and yield losses. Manual guarding and physical fencing are inefficient and ineffective. Real-Time Animal Intrusion Detection Network (RAID-Net) is an end-to-end intelligent agricultural surveillance system that consists of an integration of YOLO and a multi-task CNN to detect and recognize herbivorous animal intrusion in open-field rice nursery fields in real time. To this end, a novel and perfectly balanced dataset of 14,000 images with five different herbivore species (Buffalo, Camel, Cow, Goat, and Sheep) was curated, and a variety of seven YOLO models and five CNN architectures were benchmarked to identify the optimal detection and classification backbones. The model design includes a lightweight YOLOv8 Nano object detector and a custom-designed Multi-task Convolutional Neural Network (CNN) for species classification and behavior detection grazing and moving. The results of the Customized Multi-task CNN on cropped animal regions showed 97.72% species classification accuracy and 98.14% behaviour recognition accuracy. The RAID-Net with YOLO detection-aware alert pipeline yielded 90.05% species accuracy, 94.31% behaviour accuracy, and 82.73% hybrid accuracy, with an end-to-end latency of 17.61 ms and an inference rate of 56.79 FPS on edge-compatible hardware with an alert precision of 98% and recall of 89%. The system differentiates between high-threat grazing and low-threat moving behaviors, automatically activating a siren deterrent and sending instant WhatsApp/SMS alerts to farmers, providing an automated, scalable, and very accurate solution for precision agriculture and crop protection.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.