Real-time object three-dimensional reconstruction method based on depth camera

Abstract

The invention provides a real-time object three-dimensional reconstruction method based on a depth camera. The method comprises the steps that a frame pf depth image is sequentially acquired from thedepth camera as a current frame and is preprocessed; the relative pose of the current frame and the previous frame is estimated through a centroid ICP algorithm, and the precise pose of the camera ofthe current frame is calculated by using the precise pose of the previous frame and the relative pose of the current frame and the previous frame; the current frame data are fused into the local TSDFby using the precise pose of the camera; and point cloud integrated into the local TSDF in the step 3 is acquired from the local TSDF, and the point cloud is used as the point cloud of the previous frame, or match fusion is carried out on the local TSDF and the global TSDF, and the local TSDF is initialized. According to the invention, the failure of the ICP matching algorithm is avoided; accumulated errors are reduced; the accuracy of a model is improved; and the method is suitable for reconstructing a specified object or person.

Description

technical field [0001] The invention relates to the field of three-dimensional imaging, in particular to a method capable of real-time three-dimensional reconstruction of objects or human bodies. Background technique [0002] 3D reconstruction technology has always been a hot topic in the field of computer graphics and computer vision. 3D reconstruction is to build a 3D model from the input data. With the emergence of various depth cameras for ordinary consumers, the technology of 3D scanning and reconstruction based on depth cameras has been developed rapidly. For points in the real scene, each frame of data scanned by the depth camera includes not only the color RGB image of the point in the scene, but also the distance value from each point to the vertical plane where the depth camera is located. This distance value becomes the depth value (depth), and these depth values ​​together form the depth image of this frame. [0003] The depth image can be regarded as a graysc...

AI Technical Summary

Problems solved by technology

[0004] The core issues of 3D reconstruction based on depth cameras are: 1. The data volume of depth cameras is very large, how to deal with huge data in the reconstruction process

[0008] 1. The object or person is not separated from the scene. When the reconstructed object is an object or person in the scene, the object needs to be segmented in the reconstructed 3D model. The complexity of segmentation is high, so it is not suitable for reconstructing the object For the case of objects or people in the scene

[0009] 2. Since the frame rate of the camera is fixed, when the scanning speed is fast, the camera poses of the two frames before and after are quite different, resulting in the failure of the ICP+RGBD matching algorithm used by KinectFusion and Kintinuous

[0010] 3. Accumulated errors lead to a decline in the accuracy of the 3D object model. Although loopback detection has been added to Kintinuous and can effectively detect loopbacks and perform loopback optimization to correct the model during scene reconstruction, this loopback detection method often fails during human body scanning. Wrong loopback or no loopback occurs

Images