Map Building And Feature Matching Technology For Autonomous Navigation Of Underground Unmanned Aerial Vehicle

UAV will play an important role in the future of unmanned mining,and the map building of underground coal mine is the key to achieving autonomous navigation for underground drone.Aiming at part of area of underground coal mine,with the help of artificial landmark,the reliable and effective methods of map building and feature matching based on priori map are proposed,which is beneficial to more efficient autonomous navigation of the drone in the underground coal mine.Aiming at the rest of unknown area of underground coal mine,a reliable and effective line segment fitting algorithm is proposed,which is beneficial to the drone in the underground coal mine to build map.The researching results are as follows:(1)The map building method for the autonomous navigation of underground drone is studied.A geometric-topological hybrid map building method based on guided reflective tags is proposed.Reflective tags with the same shape which are parallelly deployed in pairs along underground coal mine wall are proposed as the artificial land1marks.By extracting the artificial landmarks and the natural landmarks within a certain range and correlating them,a local geometrie map based on the feature points can be built,so that each reflective tag can uniquely identify the specific location of underground coal mine.At the same time,the point in the middle of each pair of artificial landmarks is used as a topology node to building a global topology map.The simulating results show that the proposed algorithn can quickly and accurately identify landmarks even in the weak light.The hybrid map can ensure local accuracy and global consistency,and the storage capacity is small,which can applied to the structured underground coal mine.(2)The feature matching method based on priori map for the autonomous location of underground drone is studied.A Grid-FLANN-RANSAC feature matching method based on hybrid map is proposed.The method firstly divides the image to be matched and the reference image into a plurality of grids,and adopts the FLANN algorithm to find matching point pairs in units of grids,which can effectively reduce the calculation when the features are coarsely matched.Secondly,the RANSAC algorithm is adopted to eliminate pairs of mismatched points,so that the correct matching rate is close to 100%.The simulating results show that the proposed algorithm not only has good illumination invariance,rotation invariance and scale invariance,but also has a high correct matching rate.The current position information of the drone can be obtained in real time based on the feature image and the environment map.(3)The line segment fitting method of unknown environment for the autonomous navigation of underground drone is studied.A line segment fitting method that combining image enhancing,Canny edge detecting,Hough transforming and in-line merging algorithm is proposed.Firstly,an SSR image enhancing algorithm combined with illumination adjusting and guided filtering is adopted to overcome the shortcoming of uneven illumination in feature images.Secondly,the Canny algorithm is adopted to detect the edge feature,and the Hough transforming is adopted to fit the line.Finally,the same linearity judgment is added to reduce the influence of mesh size of Hough transforming on the fitting precision.The simulating results show that the proposed algorithm improves the accuracy of line segment fitting in low light environment.The above works have certainly reference value for further achieving the autonomous navigation of underground drone.