To overcome the limitations of single unmanned aerial vehicle (UAV) systems, research has increasingly focused on the coordination of multiple UAV platforms. These platforms are often deployed for tasks requiring advanced environmental perception, where the ability to autonomously generate and fuse dense maps is essential. However, existing algorithms for visual mapping with multiple UAVs exhibit significant shortcomings, including issues with map density, fusion accuracy, and comprehensive system testing. This article addresses these challenges by introducing techniques for dense mapping and map fusion in multiple UAVs systems. We propose a visual dense point cloud mapping algorithm that integrates generalized nearest neighbor iteration with voxel filtering. This method not only reduces redundant map points but also enhances mapping accuracy by mitigating sensor errors and drift that often generate invalid map points and overlaps. Additionally, we introduce a dense map fusion algorithm tailored for overlapping regions. This algorithm establishes precise criteria for map fusion and efficiently achieves dense map fusion by identifying overlapping areas and accurately matching point pairs. The efficacy of the proposed algorithms is demonstrated through experimental validation.