| Coal is one of the most important energy resources for the development of China's national economic,thus coal mining has drawn lots of attention.However,due to the structure compelicity of the coal mine and the diversity of underground geological environment,manual operation is often inefficient which caused frequent accidents.With the continuous development of information technology,Unmanned Aerial Vehicle(UAV)will gradually replace manpower and play a more and more significant role on the unmanned coal mining in the future.In this thesis,the autonomous navigation of underground UAV is chosen as the research object.Aiming at the special environment of mine tunnel,the key technologies involved in autonomous navigation based on monocular vision sensor,such as pose estimation,obstacle detection and path planning,are studied.The main contributions are as follow:(1)Aiming at the pose estimation problem in autonomous navigation of underground UAV,this thesis proposed a pose estimation method of underground UAV based on monocular vision and artificial landmarks.Reflective tags with the same shape are used as the artificial landmarks,which are identified quickly by the visual attention mechanism.According to the geometric relationship between the center point of the airborne camera and the adjacent reflective tags and its projection points,the position and attitude of the underground UAV are solved.The simulation results show that the proposed pose estimation method has higher algorithm efficiency,which can satisfy the requirements of accuracy and real-time performance of UAV's pose estimation in mine tunnel.(2)Aiming at the obstacle detection problem in autonomous navigation of underground UAV,this thesis proposed an obstacle detection method of underground UAV based on monocular vision.First,in order to improve the low contrast and blur of mine images,a weighted mean filtering algorithm based on extremum suppression and median is proposed,and the images are enhanced by CLAHE algorithm.Then,an intersection method of global threshold binarization and local threshold binarization based on morphological processing is proposed to separate the obstacle from the background,and the contour of the obstacle is extracted by the Suzuki contour tracking method.Finally,according to the shape characteristic parameters of the obstacle,the minimum external rectangle is used to precisely locate the area of the obstacle.The simulation results show that,the obstacle detection algorithm proposed in this thesis aiming for clear contour can lead to correct recognition result of the obstacle,which means the proposed method can satisfy the requirements of obstacle detection of underground UAV.(3)Aiming at the path planning problem in autonomous navigation of underground UAV,this thesis proposed a path planning method of underground UAV with dynamic obstacle avoidance capability based on monocular vision.First,ORB algorithm based on quadtree and triangulation method are combined to locate the obstacle in the mine tunnel.Then,a Gauss-Newton method based on scale Huber function is proposed to optimize the obstacle location results,and the mine tunnel environment is modeled by hybrid grid method and configuration space method.Finally,based on the optimized obstacle location information and the mine tunnel environment model,the A*algorithm is used to implement the path planning of underground UAV.The simulation results show that the proposed path planning algorithm can successfully avoid the obstacle ahead and reach the destination safely,which satisfy the requirements of optimal path planning of underground UAV.The above work not only provide with a certain technical basis for the actual development and flight test of the underground UAV autonomous navigation system,but also promote some research references for the design of the UAV autonomous navigation system in other special scenarios. |
PDF Full Text Request