Design And Research Of Vision-based Navigation System For Small-scale Unmanned Helicopter Autonomous Landing

Autonomous landing is an important component of autonomous flight for small-scale unmanned helicopter(SUH). During the spot landing stage, the flight control system requires real-time and high precision relative location information, especially the height information. However, INS/GPS integrated navigation with the sonar altimeter can hardly meet the requirements. Therefore, in this paper, a vision-based navigation system is researched and designed, which is responsible for the output of the position and attitude information required during the spot landing stage, and lays the research foundation for the implementation of autonomous landing on the mobile platform. The main contribution of this paper is as follows:According to the characteristics of SUH platform, and considering the relevant factors of cost, performance, load and seismic characteristics, we build up a navigation system hardware platform which is suitable for SUH autonomous landing. Then we details the software framework and results of autonomous landing navigation system in all stages.In order to simplify image processing and solve the problem of the restricted horizons and edge occlusion, we design a graded landmark which is applicable to the autonomous landing scene, and present an improved threshold segmentation algorithm combining Otsu and the most frequent value method, which help to improve the real-time nature and is applicable to the indefinite lighting conditions and target size application environment.After identifying the landing landmark and obtaining effective image information, we use pose estimation algorithm to calculate the relative position and orientation between SUH body and the landmark, and then elaborate the output strategy of the navigation system. In the course of the research, we summarize some possible sources of error which will affect the accuracy of pose estimation. On the basis of this analysis, we propose a pose estimation algorithm based on the characteristics of the parallel lines, making the algorithm more stability in the noise conditions, Then, in order to further reduce normalized coordinate error for pose estimation accuracy, a position estimation algorithm based on area characteristics is proposed.At the end of the paper, we carry out some ground-based and actual flight experiments, associated with the real-time nature of the autonomous landing navigation system, the effect of the grading landmark positioning and recognition algorithm and pose estimation algorithm error. It takes about150ms to process a single frame for the visual navigation system, and the relative error of SUH height measurement is about2%at the height of3meters. Additionally, with the declining of height above ground, the absolute error of the position and the attitude information gradually converge. The results show that the vision-based autonomous landing navigation system can meet SUH actual flight requirements.