Entryway detection algorithm using Kinect's depth camera for UAV application

Small unmanned aerial vehicles (UAVs) are gaining popularity in aiding search and rescue teams in the wake of a disaster. When searching through ruins such as a collapsed building or a building under fire, it is almost impossible for the first rescue team to navigate inside the ruins in search for survivors. Small UAVs such as the quadcopter which is equipped with autonomous capabilities has the potential to navigate through the unknown ruins. One of the basic building blocks for any autonomous vehicle is a fast-detection sensor for detection and avoidance of obstacles. Payload and cost should also be considered when choosing the right sensor. In this study, a feature extraction algorithm using Microsoft Kinect depth camera is presented for application on a quadcopter operating in an indoor environment. The main objective of this project is to develop an algorithm that could detect entryway openings, based on the inputs from a Microsoft Kinect camera that will be mounted on a quadcopter. The algorithm is tested in a T-junction corridor of an office building, with objects such as walls, doors, glass, corridors, and fire extinguisher boxes occupying the space. The algorithm successfully detects all objects by using the depth information of each pixel in relative to other pixels. The ratio of each depth area is calculated to differentiate the entryway from the rest of the objects. The analysis reveals that the accepted ratio for entryway detection is 0.701 with +−5% error while values not within this range are considered as obstacles.