Method for filling indoor light detection and ranging (LiDAR) missing data based on Kinect

Abstract

The invention relates to a method for filling indoor LiDAR missing data based on Kinect. The method comprises extracting a key frame in a Kinect scanning process to obtain sparse scanned data; using a scale invariant feature transform (SIFT) algorithm to perform feature extraction on a RGB-D image collected by a Kinect device, and rejecting abnormal feature matching points through a random sample concensus (RANSAC) operator; merging the extracted features; extracting features of an LiDAR image, matching the features with the features of the Kinect device roughly to obtain a transfer matrix; using an improved iterative closest point (ICP) algorithm to achieve fine match of the LiDAR image with the RGB-D image of the Kinect; and performing the missing data fusion between an LiDAR model and the part scanned by the Kinect. The method for filling indoor LiDAR missing data based on Kinect has the advantages that the device cost is low, the acquisition process is flexible, scene depth and image information can be obtained, partial or missing data of indoor complex scenes can be acquired and filled rapidly.

Description

technical field [0001] The invention relates to a method for filling missing data of indoor LiDAR, in particular to a method for filling missing data of indoor LiDAR based on Kinect. Background technique [0002] LiDAR (Light Detection And Ranging), called laser radar, is the abbreviation of laser scanning and detection system. LiDAR systems are mainly divided into two categories: airborne LiDAR systems and ground LiDAR systems. The present invention is mainly aimed at ground LiDAR systems in complex indoor environments. Compared with the traditional image-based 3D reconstruction, the 3D reconstruction based on 3D laser scanning has the advantages of fast, accurate and non-contact. Due to the discontinuity of the laser scanner, in order to convert the coordinates of the point cloud data acquired by multiple scanning points into a unified coordinate system to form a completed point cloud model, image registration is required. Image registration technology is mainly divided ...

AI Technical Summary

Problems solved by technology

[0005] The present invention mainly solves the technical problem that traditional laser scanners have limited scanning range in the prior art, especially for complex indoor scenes, and cannot complete data collection work in narrow areas; it provides a new type of somatosensory device Kinect is cheap, can acquire scene depth and image information at the same time, output at a rate of 30 frames per second, can acquire a large amount of depth information, and can maintain the integrity of collected information, etc., to realize the Kinect-based indoor LiDAR missing data collection for local data collection. method

Images