Font Size: a A A

3D Scene Reconstruction Based On2D-Laser-Scan Alignment

Posted on:2015-04-26Degree:MasterType:Thesis
Country:ChinaCandidate:X Q LinFull Text:PDF
GTID:2298330467985863Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
In recent years, the emergence of multi-propeller aircraft has opened up a new direction for the research of robot. But due to the limited payload and endurance, these aircrafts cannot be equipped with existing3D laser range finders at present to accomplish the3D scene reconstruction task. Therefor efficient3D reconstruction technique based on2D laser range finder becomes a new research hotspot in the field.This paper presents a new3D scene reconstruction method based on2D laser scan alignment along with the modeling of system data and optimization. All the measurement data from sensors contains noise, so the primitive reconstructed scene based on these measurements presents irregular distortion. Aiming at these problems, probabilistic models of these measurement data are constructed as well as a spatial smoothness model based on the smoothness assumption of the target environment. Then the Maximum Likelihood Estimation is adopted to transform the alignment problem into an optimization problem. After solving the problem using iterative method, a set of more precise pose parameters can be obtained. Finally the integration of multi-sensor and calibration of primitive reconstruction result can be achieved.On the foundation of theoretical study, this paper builds a portable data acquisition system. The system is composed of a2D laser range finder, an AHRS, a GPS and an optical flow module for the purpose of scanning the scene, sensing attitude, measuring locations and estimating velocity respectively. All the data produced is stored using a portable computer. To verify the validity of the proposed alignment method, many3D reconstruction experiments of structured and semi-structured scenes are conducted in indoor and outdoor environment. By comparing and analyzing the experimental results, a conclusion can be drawn that the proposed alignment method can correctly calibrate the reconstruction error produced by sensor noise. Compared to traditional3D reconstruction system based on the stationary scanning mode, the introduced system in this paper can complete data acquisition in the duration of moving. Thus the operating efficiency can be greatly improved as well as flexibility. This system can be hold in hand or attached to a multi-propeller aircraft in real applications.
Keywords/Search Tags:Scene Reconstruction, Laser range finder, Scan alignment, LikelihoodEstimation, Iterative Optimization
PDF Full Text Request
Related items