Research On The Three-dimensional Panoramic Stitching Technique

The three-dimensional imaging technique is a kind of measurement technique for acquiring the stereo space coordinates of objects to be measured,which can provide more abundant spatial shape and position information,currently in the parts measurement,unmanned aerial vehicle navigation,virtual reality,3D printing,and medical diagnosis and other fields have a wide range of applications.Due to the limited view of a single depth camera,it is more and more important to study how to combine multiple depth images and expand the three-dimensional image mosaic technology of three-dimensional imaging field.At present,the three-dimensional point cloud stitching algorithm has made great progress in three-dimensional scanning and virtual reality applications,but the calculation of three-dimensional point cloud data is relatively large,the real time is low and the hardware requirements are high,which limits the large scale application of three-dimensional stitching technology.In this paper,a 360-degree real time three-dimensional panoramic imaging stitching system is designed to solve the short comings of the single depth camera detection field.Begin from the camera imaging model,this paper analyzes the theoretical relationship between the three-dimensional image stitching technique and the camera calibration technique,explains the basic problems of three-dimensional image stitching technology and the solution process of the rigid body transformation matrix,and analyzes and introduces the basic image processing methods.Finally,a three-dimensional panoramic imaging stitching system based on active calibration is presented,which completes the real-time stitching of 360-degree panoramic images quickly and effectively.In this paper,the second generation Kinect V2 camera is used as the three-dimensional image acquisition device,the conversion relationship between the color camera and the deep camera coordinate system of the Kinect V2 equipment is calibrated by the checkerboard,so the reconstruction of the rgb-d color three-dimensional image is completed,and then the two-dimensional image calibration board is designed,The two adjacent Kinect V2 devices are calibrated to obtain the rotational component R and Shift component T of the rigid transformation parameters of the two cameras,and then the accurate registration parameters of the adjacent camera are obtained by iterative optimization of ICP algorithm,and the point cloud data obtained from eight depth cameras are unified into the same coordinate system.Completed a 360-degree real-time panoramic mosaic system.