Unmanned aerial vehicle image rapid three-dimensional reconstruction method for field unknown environment

Abstract

The invention discloses an unmanned aerial vehicle image rapid three-dimensional reconstruction method for a field unknown environment. According to the invention, the method includes using an OpenMPmulti-thread processing mechanism to resize image returned by an unmanned aerial vehicle, sorting the images, extracting feature points in a parallel grid format, obtaining sparse point clouds of allimage clusters through a global motion recovery structure, calculating depth images of the images, and constraining regions of interest through image Mask; and based on a depth map fusion method, obtaining the dense point cloud of each image cluster, and combining the reconstruction results of each cluster to obtain the three-dimensional structure of the whole scene. The method can be realized without depending on complex image acquisition equipment and computer equipment, and the reconstruction speed is greatly improved by using parallel grid type feature extraction; through point cloud partitioning and grid partitioning, the memory pressure is reduced, and the reconstruction speed is increased.

Description

technical field [0001] The invention relates to the field of unmanned aerial vehicle technology and the field of three-dimensional reconstruction, in particular to a method for rapid three-dimensional reconstruction of unmanned aerial vehicle images facing unknown environments in the field. Background technique [0002] At present, in outdoor or field applications, UAVs can realize rapid and comprehensive three-dimensional reconstruction of a scene, such as forestry resource protection in the wild, post-disaster rescue of fires and earthquakes, border security, oil fields or offshore platform oil and gas facilities, and power facilities. Monitoring and other fields have very important application value. [0003] For the 3D reconstruction requirements of wild scenes, if it is realized based on the satellite platform, there is a problem of insufficient resolution, and it is impossible to clearly construct a 3D scene in the wild. For more accurate radar laser scanning reconstr...

AI Technical Summary

Problems solved by technology

[0003] For the 3D reconstruction requirements of wild scenes, if it is realized based on the satellite platform, there is a problem of insufficient resolution, and it is impossible to clearly construct a 3D scene in the wild

For more accurate radar laser scanning reconstruction, it has the disadvantage of being expensive and heavy, and it is very inconvenient to perform 3D reconstruction in the field environment

Most of the traditional image-based 3D reconstruction methods have shortcomings such as long running time, poor timeliness, and high requirements for equipment, which cannot meet the needs of rapid reconstruction such as emergency rescue and disaster relief, military reconnaissance, and border security.

For fast and even real-time 3D reconstruction under this requirement, traditional methods cannot solve this problem well

Specifically, the traditional image-based 3D reconstruction methods have the following three problems: (1) This type of method has high requirements for the quality of the input image data, and the 3D reconstruction effect for large scenes in the wild is poor; (2) ) In the process of using drones for 3D reconstruction in this type of method, generally the drone is used to shoot all the scenes first, and then after returning, the captured images are processed to perform 3D reconstruction of the captured scenes, which does not have timeliness (3) The pictures taken by drones have high resolution and large amount of data. The traditional reconstruction method is very time-consuming and cannot meet the requirements of fast reconstruction or even real-time reconstruction.

Images