Computer Vision and Agent-Based Modeling for managing Human-Elephant conflict
Team
Supervisors
Table of content
Project Summary
Human-Elephant conflict has caused damage to the well-being of life for many decades. The methods taken to avoid and mitigate these harmful interactions have not provided safe outcomes as expected. So the requirement of a long-term solution for these conflicts arises to reduce the impacts. We propose a method using computer vision and Agent-based Modelling (ABM) techniques together, to provide a set of optimal long- term policies to reduce the impacts significantly for conflict- bound regions. The initial stage of this work focuses on creating a segmented image from a given satellite/aerial image with different classes such as water, buildings, croplands, etc to represent the area of impact. Then in the second stage, these images should be converted to an ABM-friendly format. At the final stage, they can be used to simulate the real-world environment with the related agents using a suitable Agent-based model. This method is capable of providing policies with optimal parameters for population counts, and safe distances to human regions from elephant habitats for notable reductions in conflicts. The real environment is simulated through ABM frameworks and human-elephant behavioral patterns are provided to create heterogeneous agents with stochastic movements. The identified best-case policies for a region can aid the decision-makers of the governments, and wildlife authorities to move forward with safe construction planning and conservation of elephant habitats. Ultimately, the use of this prediction and simulation model will provide recommended strategies to improve the survival rate by reducing human-elephant conflicts in many regions of the world.
We believe this research expands the scope that can be explored with the use of combining computer vision with agent-based modelling for solving interaction-based problems that is a threat to the world with simulated environments. Further research and improvements with this existing model will generate more substantial results that would solve world crises with the use of modern technology capabilities for a better future.
Publications
- Semester 7 slides
- Semester 8 report
- Semester 8 slides
- Hans Thisanke, Chamli Deshan, Kavindu Chamith, Damayanthi Herath, Sachith Seneviratne and Rajith Vidanaarachchi “Semantic Segmentation using Vision Transformers: A survey” (2022)