A real-time 3D reconstruction method based on UAV

Abstract

The invention discloses a real-time three-dimensional reconstruction method based on an unmanned aerial vehicle, comprising the following steps: step 1, collect scene data, obtain depth data through a depth sensor according to the collected scene data, and obtain a depth map; step 2, obtain step 1 Noise preprocessing is performed on the depth image of the noise preprocessing, and the noise preprocessing includes morphological operations and filtering operations; step 3, accelerates the depth image after the noise preprocessing in step 2, and the acceleration processing includes the binary in the depth image after the noise preprocessing. Dimensional points are uniformly sampled; step 4, convert the depth map after accelerated processing in step 3 into a point cloud; step 5, model the point cloud obtained in step 4; step 6, perform noise reprocessing on the result obtained in step 5 , to achieve 3D reconstruction. The invention can carry out real-time three-dimensional reconstruction of the scene, provides the possibility for the subsequent intelligent processing of the drone, and helps to promote the application of the drone.

Description

technical field [0001] The invention belongs to the technical field of unmanned aerial vehicle control, and in particular relates to a real-time three-dimensional reconstruction method based on an unmanned aerial vehicle. Background technique [0002] With the rapid development of information technology and the continuous maturity of equipment technology, there have been unmanned aerial vehicles, referred to as unmanned aerial vehicles (UAV). Because of its characteristics of non-contact, zero casualties, and long-term work, it has been widely used in military and civilian fields. [0003] From a technical point of view, drones can be divided into several categories: unmanned helicopters, unmanned fixed-wing aircraft, unmanned multi-rotor aircraft, unmanned airships, and unmanned paragliders. In the civilian field, UAVs can be used in aerial photography, news reports, environmental monitoring, express delivery, disaster monitoring, traffic patrols and power inspections, etc...

AI Technical Summary

Problems solved by technology

[0007] (2) When the deflection angle between pictures is large or the light, season, etc. are different, it will be difficult to match the feature points;

[0008] (3) The feature matching speed is slow, very time-consuming, and not real-time

[0009] Using lidar to quickly scan the scene can efficiently and accurately obtain the information of the three-dimensional points of the scene, but the cost of lidar is high, and it is difficult to popularize and apply

[0010] Based on the reconstruction algorithm of 3D voxel, the method is to initialize a 3D body first, then divide it to obtain smaller cube grids, which are voxels, and then design a cost functional function for each The consistency between a voxel is evaluated, and finally the cost function is minimized to complete the estimation of the three-dimensional surface; the common problem with this method is that the reconstruction result contains more noise

Images