Research On UAV Path Planning Technology Based On Visual SLAM

In recent years,with the rapid development of AI technology represented by intelligent sensing and decision-making technology,the intelligent and autonomous degree of UAV is becoming higher.In particular,UAV’s intelligent sensing mapping of the surrounding environment,autonomous path planning and navigation on this basis,has become a research hotspot in this field.Visual Simultaneously Localization and Mapping(v SLAM)is the core technology of Intelligent perception and location.It relies on camera sensors to obtain very rich information and realize intelligent perception of the environment.Therefore,it is widely used on unmanned aerial vehicles with limited payloads.However,because the visual SLAM method is more sensitive to illumination,extracts more redundant features,it causes a large amount of calculation,real-time performance is not strong,and the robustness is not high.Therefore,this article firstly proposes an adaptive threshold feature extraction method.This method expands the edges of the collected images to get edge corners,and then grids the image,applies adaptive thresholds in the grid to extract the corners,and can obtain more stable and reliable feature information.Aiming at the defects of extracting more redundant features and large algorithmic calculations,the feature allocation method is improved,and different quadtree heights are set according to the expected feature points of different image sampling layers,thereby reducing redundant features and reducing system calculations the amount.Making sure the real-time of SLAM technology.In the UAV path planning based on the environment map constructed by slam,the calculation amount of traditional a* algorithm will increase with the increase of nodes,which reduces the calculation efficiency leads to the problem of low real-time of UAV autonomous path planning.This paper presents an improved a* algorithm,which uses weight value to balance the cost function and heuristic function of a* algorithm.Then,the constraint model is established for heuristic function and the minimum value is solved,obtain the optimal solution of improved a* algorithm path planning in a specific environment.Then,in the process of node search,the search angle is optimized to reduce the number of nodes to search,and avoid the problem of small angle inflection point in flight.Finally,this article builds a drone vision SLAM path planning platform based on the above work in Ubuntu 16.04 system.The proposed method is verified by using public data set and self-made data set.The results show that drone vision SLAM path planning platform designed in this paper can complete positioning and mapping well,and can perform path planning on the establishment of accurate environmental maps,and prove the viability of the algorithm.