Research And Application Of 3D Map Modelling For Indoor Environment Based On Kinect

Computer vision is a kind of science which is used to acquire and process natural data to simulate the biological vision system by computer and related equipment.Three dimensional(hereafter 3D)map modeling has been playing an important role in the field of computer vision.The traditional 3D map modeling technology is complex and costly,also the speed and accuracy of this technology need improvement.Therefore,how to reduce the cost of 3D map modeling,improve the speed and accuracy of 3D modeling,simplify the operation of 3D modeling has a high research value and good application prospects.In this paper,it mainly studies on the 3D map modeling and its application in indoor environment based on Kinect.Firstly,this paper will introduce and analyze the basic principle and functions of Kinect V2,obtaining RGB images and depth images by Kinect V2 for the 3D map modeling,and according to the RGB image and depth image collected by the sensor to generate 3D point cloud map of each frame.After the comparison and analysis of different visual feature detection methods,this paper chooses the siftGPU which has better accuracy and faster speed to extract and match the feature points of the image.In the 3D map modeling process,this paper uses singular Value Decomposition as the core of the Iterative Closest Point for motion estimation,at the same time,in order to avoid the low accuracy of pose estimation problem due to the larger error caused by using this method to process images feature matching,this paper uses the method of random sample consensus to match results further.After that,this paper acquires the final position by using the data from loop detection to process the graph-based optimization,splicing out high precision point cloud map and motion trajectory.Simultaneously,to create a practice of 3D map modeling in daily life,this paper designs a 3D house showing system which is a kind of easy access to structure information of house.The system by acquiring real data of indoor environment to create a 3D model for user's access to real 3D structure information and color information for indoor environment.It effectively shows the 3D map modeling of the indoor environment based on the application of Kinect in the field of daily life.