Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2024
- Goals
- Record
- f7b56efa-5454-49d1-88e4-f715efef6ed1
- Published
- 2026-10-30
Universitas Dian Nuswantoro (UDINUS), through its researchers, conducted a study on vegetation density classification using convolutional neural networks (CNN) and a proposed modified activation function. The activity addressed the need for accurate land mapping in Indonesia. Forest cover reached 95.97 million hectares, or 51.2% of the total land area, while fluctuations from 2014 to 2022 highlighted the need to understand factors influencing landscape patterns. The source states that decreases in forest cover increase vulnerability to natural disasters, particularly floods, causing economic and environmental losses.
The study focused on land cover mapping strategies, especially in under-monitored regions, and considered recent satellite and drone data for rapid and precise mapping. The proposed activation function added a linear component to improve CNN accuracy and effectiveness. It was evaluated using the Vegetation Density of Peatland Drone dataset. During testing, the proposed function maintained an accuracy of 90.07%. Statistical analysis showed high data distribution and significant performance variation. Although computation times were slower, the proposed function demonstrated convergence speed and effectiveness compared with other activation functions. ANOVA and Friedman analyses confirmed significant differences in performance and computation time among activation methods.
This UDINUS research activity applies deep learning to vegetation density and land cover classification. It provides evidence of an institutional research approach focused on improving mapping accuracy and effectiveness in contexts involving landscape patterns, forest cover, and flood vulnerability.