Vision-Based Landing Site Detection for Unmanned Aerial Vehicle: A Review

Autonomous landing technique for Unmanned Aerial Vehicle (UAV) is a well-studied problem, for most flight accidents happened during this stage. This survey aims to provide an extensive overview for a guide for vision-based autonomous landing site detection development. According to whether an auxiliary marker is set, the detection tactics can be categorized as marker-aided and markerless ones. Marker-aided tactics usually employ an elaborate and distinctive geometric design to achieve robust detection efficiently. While the markerless tactics can be further decomposed into two main steps: pattern recognition and flat site detection. Moreover, the computational optics theory and deep learning are now showing extraordinary talents in mobile platforms including UAV, thus we elaborate monocular depth estimation with different supervision methods for flatness analysis in markless tactics. We hope our comprehensive overview may possibly be helpful in the analysis and improving the vision-based autonomous landing technique for UAV.