Visual tracking of animals in three-dimensions using mobile handheld independent GoPro cameras and VSLAM software

I present the system PATMOS (Paths And Tessellated Meshes from ORB_SLAM2) for measuring three-dimensional paths of animals in situ using two handheld GoPro cameras and a small spatial reference object. Animal paths were triangulated from mobile camera positions obtained from a modified version of ORB_SLAM2, an open-source visual simultaneous localization and mapping software package. In addition to path calculation, this process provided a virtual three-dimensional surface approximation to the environment from which path to environment distances can be quantified. PATMOS can also fit a tranquil water's surface to an analytic plane if there are a sufficient number of visible objects intersecting the water's surface and can track objects over the water's surfaces with a single camera by measuring the object with its reflection. This technology was highly portable, could follow moving animals, and gave comparable spatial and temporal resolutions to fixed camera systems that use commercial cameras. An investigation of falling objects yielded a gravitational constant measurement of 978±40 cm/s2. I demonstrated PATMOS's utility in terrestrial and aquatic environments by quantifying dragonfly flight characteristics and the inter-spatial distances between substrate and damselfish.

本研究提出PATMOS系统（全称为Paths And Tessellated Meshes from ORB_SLAM2，源自ORB_SLAM2的路径与细分网格），该系统可通过两台手持GoPro相机与小型空间参考物体，原位测量动物的三维运动路径。动物运动路径通过基于修改版ORB_SLAM2获取的移动相机位姿进行三角化求解得到；ORB_SLAM2是一款开源视觉同时定位与地图构建（SLAM）软件包。除路径计算外，该流程还可生成研究环境的虚拟三维表面近似模型，借此可量化运动路径与环境要素间的距离。当存在足够多与水面相交的可见物体时，PATMOS还可将平静水面拟合为解析平面；此外，该系统可通过单相机结合目标物体与其倒影的测量结果，追踪水面上方的目标物体。该技术具备优异的便携性，可追踪移动中的动物，其空间与时间分辨率可媲美采用商用相机的固定摄像系统。通过对下落物体的测试实验，该系统测得万有引力常数为978±40 cm/s²。本研究通过量化蜻蜓的飞行特征以及栖息基底与雀鲷间的空间间距，验证了PATMOS系统在陆地与水生两类环境中的实用性。