Authors:
S. Rubin Bose, J. Angelin Jeba, R. Regin, B. Judy Flavia, S. Suman Rajest, Rahul Panakkal

Addresses:
School of Computer Science and Engineering, SRM Institute of Science and Technology, Ramapuram, Chennai, Tamil Nadu, India. Department of Electronics and Communication Engineering, S.A. Engineering College, Chennai, Tamil Nadu, India. Department of Research and Development, Dhaanish Ahmed College of Engineering, Chennai, Tamil Nadu, India. Department of Computer Science and Engineering, University of Illinois at Urbana, Champaign, Illinois, United States of America.

Abstract:

The global environment suffers from deforestation. It involves cutting down trees in a forest. It requires real-time monitoring, but the forest is huge and hard to monitor. ResNet-50 is our CNN model. Although a sophisticated deep learning model, it can spot patterns in satellite imagery that humans miss. Comparing pre- and post-deforestation satellite pictures helps identify tree loss and deforestation. This model enables us to monitor and detect major forest tree loss in real time, helping conserve natural resources. Some methods detect only large groupings of trees, but deforestation begins with modest losses. ResNet-50 can detect subtle changes, assess patterns, and graph the deforestation rate. ResNet-50 can process 224X224 or 256X256 pixel images. The default forest resolution is 5 km, covering 5 sq km. Another model covers 100 sq m with low accuracy. ResNet-50 took longer to process but was more accurate. Combining ResNet-50 and U-Net yields black-and-white images of deforestation and forest regions, with corresponding percentages. The project successfully detected deforestation using deep learning and remote sensing. These deep learning models consistently segment and measure deforestation areas across geographic locations and time intervals, as shown by their high accuracy and IoU scores. The results show that the approach is suitable for in situ, automated, large-scale environmental monitoring and mapping.

Keywords: Deforestation Detection; Remote Sensing; Deep Learning; U-Net Architecture; Forest Monitoring; Convolutional Neural Network; Synthetic Aperture Radar; Automated Ecosystem Monitoring.

Received on: 22/11/2024, Revised on: 26/01/2025, Accepted on: 04/04/2025, Published on: 14/09/2025

DOI: 10.69888/FTSESS.2025.000543

FMDB Transactions on Sustainable Environmental Sciences, 2025 Vol. 2 No. 3, Pages: 171-182

• Views : 131
• Downloads : 18

Download PDF