Fast and Effective Identification of Window and Door Openings for UAVs' Indoor Navigation

Abstract

As one of today's popular research field, autonomous unmanned systems are widely used in entertainment, search and rescue, health, military, agriculture and many other fields with the advantages of technological developments. Object detection is one of the methods used for autonomous vehicles to gather and report information about its environment during mission. With the ability to detect and classify objects, an unmanned vehicle can determine the type and number of objects around it and use this information in its autonomous motion and path planning or reporting the objects with the desired features. In an indoor environment, the data coming from an optic sensor provides enough capability to detect crossing points for an intelligent unmanned aerial vehicle such as doors, windows and stairs. However, sliding objects like windows and doors need to be classified as open or closed if this information will be used for path planning. Moreover, the crossing point's width needs to be calculated to ensure it is safe to go through it. So as to provide a robust navigation plan, LIDAR sensor data can be used to ensure object classification. In this study, a method to fuse LIDAR and optic sensors data has been developed and by the way a stronger object detection method is created. A crossing point is classified as an open/closed and determined whether to have enough space or not to pass through. This method was operated in a developer board mountable in a flying quadrotor which makes it possible to use it in a real-time scenario. The success of the approach has been shown with different scenarios.