Research Overview: Real-Time Wildlife Monitoring Using Edge AI and Geospatial Analysis
Author-presented overview of my research paper on real-time wildlife monitoring using computer vision, edge devices, and GIS.
I’m Mohit Patel, an Electrical Engineering undergraduate with a strong interest in computer vision, edge AI, and conservation technology.
This page presents an author-level overview of my research paper
“Real-Time Wildlife Monitoring on Edge Devices Using Computer Vision and Geospatial Analysis.”
Instead of repeating the paper verbatim, this post explains what I built, why it matters, and how it reflects my technical and research profile — similar to how a research project is presented on an academic CV or portfolio.
📄 Featured Research Paper
Author: Mohit Patel
Affiliation: Electrical Engineering, Jabalpur Engineering College
Keywords: Wildlife Monitoring, Computer Vision, YOLOv8, Deep SORT, Edge AI, GIS, Conservation Technology
Research focus and motivation
This research addresses a core problem in wildlife conservation:
How can animal movement, behavior, and resource usage be monitored at scale without invasive methods or expensive infrastructure?
Traditional monitoring approaches such as manual surveys, tagging, and static camera traps are:
- Labor-intensive
- Limited in coverage
- Often intrusive
- Poorly integrated with spatial analytics
The objective of this work was to design a low-cost, real-time, automated wildlife monitoring system deployable in forest environments using edge AI and open-source tools.
What I built (system overview)
I designed and implemented a unified monitoring framework that integrates:
- Real-time animal detection
- Species classification
- Multi-object tracking
- GPS-based geospatial logging
- Interactive spatial visualization
The entire system is optimized for edge deployment, enabling long-term, scalable monitoring.
Key technical contributions
Real-time detection using YOLOv8
- Implemented YOLOv8 for live wildlife detection
- Fine-tuned pretrained models on wildlife datasets
- Optimized inference for Raspberry Pi-class hardware
Species classification
- Built a lightweight CNN for species recognition
- Operates on cropped detections
- Achieved >90% accuracy for common species
Multi-object tracking (Deep SORT)
- Enabled persistent identity tracking
- Extracted trajectories and movement patterns
- Identified recurring activity and congregation zones
Edge-based deployment
- Raspberry Pi 4 for continuous video inference
- ESP32-CAM for low-power image-based monitoring
- No reliance on cloud connectivity
Geospatial mapping and analysis
- GPS-tagged detections mapped in WGS84 format
- Integrated Folium and QGIS for spatial visualization
- Correlated wildlife activity with proximity to water sources
Research outcomes
- Robust detection in dense forest and variable lighting
- Stable tracking with minimal ID switching
- Clear spatial clustering around water sources
- Actionable insights for habitat and resource planning
Practical and conservation impact
This work enables:
- Data-driven placement of artificial water sources
- Identification of high-risk and high-activity zones
- Support for anti-poaching strategies
- Long-term ecological monitoring
The system shifts conservation efforts from manual observation to intelligent, real-time analytics.
Future scope
Planned extensions include:
- Thermal and infrared imaging
- LoRa / mesh networking for remote nodes
- Anomaly detection for intrusion events
- Multi-site longitudinal ecological studies
Why this research matters for my profile
This project demonstrates my ability to:
- Design end-to-end AI systems
- Work across AI, hardware, and geospatial domains
- Apply research to real-world environmental problems
- Build deployable, production-ready systems
Final note
This research represents a practical application of Edge AI for conservation technology.
If you’re interested in computer vision, edge computing, or environmental informatics, I’m always open to collaboration and discussion.