Research On Autonomous Navigation Algorithm Of Indoor MUAVs Based On Monocular Vision/INS

This paper based on the analysis of the current status and key technology of Unmanned AerialVehicle (UAV)’s visual navigation, considering the requirement of Micro UAVs for unknown indoorenvironment, the key technology of UAV’s autonomous navigation was researched. Based on theresearching on image matching from monocular vision, with the help of inertial sensors, the MUAV’sattitude and autonomous navigation algorithm was studied.The camera’s motion caused the perspective transformation can result in image rotation, sizetansformation, illumination tansformation and so on, which can easy to make the characteristics offeatures points unstable. To make the feature points more robust, the currently two main methods ofmulti-scale local invariant feature algorithm were researched: Scale-Invariant Feature Transform(SIFT) algorithm and Speeded Up Robust Feature (SURF) algorithm. And the difference ofperformance between both was analyzed through experiments in the paper.To increase the autonomy of UAV, the motion will be obtained from the matching feature points.Based on detailed analysising the camera projection model and the epipolar geometry relationship,according to the epiplolar constraint, random sample consensus (RANSAC) was be used to obtain thefundmental matrix. With the imformation contained in the essential matrix, the rotation and vectorscan finally be solved. Through repetitve experiments in purely visual navigtion algorithm, the possiblecauses result in error of algorithm was analyzed, and the improvement and error compensationmethod were researched.Pure visual navigation may failed due to the imformation of the images is not enough.To fixed it,inertial navigation will be introduced, not only for necessary imformation complement, but alsoprovide some intermediate variables to simplify the visual navigation algorithm. Based on that, thethree-dimensional reconstruction of feature points in world coordinate was researched, also theobstacle avoidance technique.To verify the feasibility of monocular vision/INS autonomous navigation algorithm, thefunctional modules were realized under the platform of VC6.0, monocular vision/INS algorithm,environment reconstruction, autonomous obstacle avoiding and so on. The theory and algorithmresearched in the paper built the foundation for further research, and it will be useful to realize theUAV prototype with indoor autonomous navigation functions.