UAV Navigation Map Construction Method Based On Aerial Image And Implementation

Drones can be widely used in many important fields,such as industry,agriculture,transportation,and national defense.These applications rely on breakthroughs in a series of key common technologies.Autonomous navigation is one of them.This is because drones need to cope with complex surrounding environments and avoid obstacles in these applications.UAVs must have autonomous navigation capabilities.Reconstructing the3 D map of the environment to obtain a 3D map of the environment is an effective way to solve its autonomous navigation problem,and it is also the key to the large-scale application of drones.Traditional 3D reconstruction methods include geometric modeling and laser scanning methods,but each has obvious disadvantages.Today’s image-based3 D reconstruction methods are gaining more attention because of their convenience and efficiency.With the advent of small multi-rotor drones,it has also brought more ideal image data acquisition methods to the field of 3D reconstruction.Therefore,the research content of this article combines 3D reconstruction and UAV,constructing a 3D navigation map construction system,collecting scene image data through drone aerial photography,3D reconstruction of the image data set to obtain a point cloud map,and then constructing from the point cloud map Provide a three-dimensional navigation map that can assist UAV autonomous navigation.The main work and contributions of this paper include:1)Based on the comparison and analysis of literatures in related fields at home and abroad,we conducted in-depth research on several key steps in the construction of 3D maps.We used Zhang Zhengyou calibration method to carry out camera calibration experiments and determined camera parameters.2)For the extraction and matching of image feature points,several commonly used algorithms are studied,and these algorithms are experimentally tested,and the algorithms suitable for the construction of three-dimensional maps in this paper are given priority.For the matching process of image features in the three-dimensional map construction process,a selective image matching method is used,thereby reducing unnecessary time consumption.3)Several three-dimensional map models are studied,an octree three-dimensional map suitable for navigation is established,and the point cloud map is compressed and constructed as an octree three-dimensional map.4)The multi-layer screening algorithm is used to screen the collected video stream images for key frames,which improves the efficiency of reconstruction.A hybrid image enhancement algorithm based on histogram equalization and retinex enhancement algorithm is proposed.The image is subjected to mixed image enhancement preprocessing before feature extraction,which increases the number of feature point detection and matching.Filter the k-d tree based on the reconstructed point cloud model to remove noise,so that the final constructed octree map has higher quality.Finally,build a hardware and software platform,design and implement a threedimensional map construction system.Then,according to the drone,aerial photography is used to collect images in different scenes,and a three-dimensional map is obtained after three-dimensional reconstruction.The research work in this paper is of great significance to the ultimate realization of autonomous navigation of drones.