Research On Indoor 3D Environment Modeling Technology Based On Kinect

With the continuous development of intelligent robots,Three-dimensional reconstruction technology has been widely used in the field of machine vision.It has become a new method to complete 3D reconstructionby using a RGBD depth camera.The purpose of this paper is to acquire the color information and depth information of the indoor environment through Kinect v2 sensor,and finally reconstruct the 3D model of the indoor environment.Firstly,get the camera's internal and external parameters.The kinematical calibration algorithm was used to calibrate the Kinect camera,and the internal and external parameters of the color camera and the depth camera were obtained.The position relationship between the two was derived,then the collected color information of the same perspective is merged into the corresponding point cloud data.Secondly,preprocess the acquired depth image.The point cloud data needed for 3D reconstruction is derived from the depth information acquired by the Kinect sensor.Because of the noise and voids in the original depth image,a large amount of depth information error and invalid values in the resulting point cloud data are obtained.In order to obtain a better reconstruction result,we analyze the defects of the acquired depth image so the depth map needs to be repaired.First,cut off the edge invalid value in the depth image,then fill in the hole,and then use the joint bilateral filter based on RGB image to repair it.Finally,the outliers and down-sampling preprocessing are performed on the obtained 3D point cloud data.Finally,register the collected point cloud data.The traditional iterative nearestpoint algorithm is analyzed and improved.For the traditional ICP algorithm,a good initial registration is required.In this paper,Normal Distributions Transform(NDT)is used as initial registration,and optimized the objective function of registration error,the GICP is used to do the exact registration,and multi-frame registration of indoor scene point cloud data is completed.The three-dimensional model of the experimental results illustrate that the multistep joint depth image restoration and registration method have certain practical application value in the three-dimensional reconstruction of indoor environment.In the range of sensor accuracy,although the detailed information of the three-dimensional model is not accurate enough,the basic outline is clear and can be applied in daily life.