Dynamic Aerial Shadow Detection Based On Mini Unmanned Helicopter

Mini unmanned helicopter is widely applied in military and civilian for its lightness, lower cost and flight agility. In kinds of application tasks, the vision navigation system is a very important part of the whole mini unmanned helicopter system. Shadow is a familiar image degradation phenomenon which appears frequently in aerial images captured by mini unmanned helicopter vision system. Actualization of effective dynamic shadow detection for aerial images not only helps object recognition by eliminating disturbance cost by shadows, but opens out clues for 3D scene information as well. In this way, dynamic shadow detection assists mini unmanned helicopter system on environment apperception and vision navigation.To be applied on mini unmanned helicopter system, shadow detection must have low load and low cost, with robustness for changeful environment and dynamic camera. Aiming at these limitations and requirements, a dynamic shadow detection system based on mini unmanned helicopter platform is designed and actualized in this paper. The research work includes:1) Designing and building a hardware and software system frame for mini unmanned helicopter dynamic shadow detection, the system has low load and low cost.2) Designing a dynamic shadow detection algorithm based on semi-supervised learning strategy for aerial images captured by micro unmanned helicopter. Firstly two independent features are extracted based on the analysis of shadow feature set; then a modified Co-EM strategy is applied to shadow detection for separate frames; finally, according to the motional character of the helicopter, an online detecting model based on parameter transfer is proposed.3) Building a software platform for the running of the above dynamic shadow detection algorithm. 4) Experimenting on the designed dynamic shadow detection method on a real mini unmanned helicopter platform; Making contrastive analysis between results of the designed method and several typical shadow detection methods at present, demonstrating the effectiveness and advantages of the dynamic shadow detection system presented in this paper.