A Software Scheme for UAV's Safe Landing Area Discovery

This paper proposes a software processing scheme for small-sized UAV in its landing area discovery using its single onboard camera and machine learning algorithms. The two-stage processing procedure was proposed. In first stage a similarity based textured area identification method was adopted to find the possible landing areas. Afterwards, in second stage, these results were refined and evaluated by using some machine learning algorithms. The UAV can then take use of these results as its emergency landing target options. The software scheme we designed implemented the whole process but still allow the developers to embed their own algorithms to make better results. Our preliminary research has disclosed that this software and application are useful and can provide great convenience and efficiency.