| Summary: | 碩士 === 國立高雄大學 === 資訊工程學系碩士班 === 101 === The Light Detection and Ranging (LiDAR) device has become an increasingly common tool for the acquisition of 3D data. Such devices are capable of acquiring 3D measurement quickly and accurately over relatively large coverage areas. However, a LiDAR device is unable to capture the color information of the surfaces. Binocular stereo, on the other side, can be easily deployed to acquire colored depth data but requires image features to perform correspondence analysis for depth calculation and has difficulties in providing dense depth data. The LiDAR and the binocular stereo systems have complementary characteristics and by integrating the two systems, it is possible to construct a system that possesses the advantages of the two methods.
In this work, we constructing a system mounted with a LiDAR device and two pan-tilt cameras for large scale 3D scene reconstruction. Spatial feature detection and matching are then performed on the images to obtain sparse point cloud. As the system moves with time, temporal feature tracking is also performed on the images sequence to calculate intra-frame motion. With the proposed auto-calibration method, the extrinsic parameters of the LiDAR and the cameras are calculated, which are in turn used to align the point clouds scanned by the LiDAR along the path. Experimental results have shown that the proposed system and method is able to achieve integrated dense reconstruction for large scale scene.
|
|---|
