Multi-Camera 3D Mapping with Object Detection, Similarity Matching and Depth Estimation

The purpose of this research is to develop a room mapping and an animal and human tracking algorithm. We will use a 360-degree camera to get a broad point of view (POV). Then a simulated robotic agent will move throughout the room to show a different POV where the 360-degree camera cannot see. With our two cameras we can run deep learning object detection to locate animals and humans. We then use ORB points for feature matching across the two cameras. By using two POVs, we can build a 3-D model of the room using Simultaneous Location and Mapping (SLAM). This 3D model will be analyzed for changes over time comparing the actual room before and after. This room mapping will increase the awareness of the environment.