Hybrid Metaheuristic Method for Vegetation-Rich Outdoor 3D Point Cloud Image Registration

Abstract

This paper presents a hybrid metaheuristic point cloud registration procedure combining Differential Evolution (DE) and Particle Swarm Optimization (PSO), referred to as DE-PSO. The method pairs the broad search capabilities of DE with the rapid convergence of PSO, avoiding the local minima traps that degrade standard Iterative Closest Point (ICP) implementations. Vegetation-rich outdoor environments present significant challenges due to dense foliage, irregular geometries, and severe point cloud noise. The proposed DE-PSO strategy addresses these complex scenes, achieving high registration accuracy and avoiding traditional ICP misalignments. Implemented using open-source C++ tools and tested on a custom dataset, the approach yields superior alignment with the lowest root mean square error and fastest execution time compared to standalone configurations. Overall, the method shows the feasibility of combining population-based metaheuristics for complex data, providing a practical tool for outdoor applications where conventional alignment struggles.