Scanning And Reconstruction Of3D Objects And Human Bodies Using Depth Cameras

Capturing the3D shapes of objects and human bodies is a basic task for computer graphics and computer vision applications. Compared with structured light, laser scan or stereo vision, depth camera has its advantages. It enables acquiring depth data in real time with little consideration about texture or lighting condition. The camera is compact, low-price and as easy to use as a video camera, which has the potential to be used by everyday users. Unfortunately, the captured data cannot be directly used due to the low image resolution and high noise level. Another challenge is the non-rigid registration of noisy deforming objects. In this paper, novel approaches are proposed to scan3D objects and human body with the help of time-of-flight camera and Microsoft Kinect. The main content is as follows:This paper presents an algorithm for3D reconstruction of rigid object using one TOF camera. First, the captured data from different views are segmented and denoised. Second, the meshes are roughly aligned using rigid alignment, and key frames are selected to reduce motion blur and information redundancy. Then the key frames are further aligned using global non-rigid alignment. Finally, a qualified mesh is generated by using Poisson surface reconstruction.A novel approach is proposed to scan3D body with hairstyle using one TOF camera. By capturing depth data at video rate, temporal average meshes can be obtained from different views. After some analysis, we found that the local geometric details of real surfaces, after the hair scanning process and low-frequency body deformation, are still preserved in the average meshes. Utilizing the restriction that the corresponding surfaces in different views should overlap, a global optimization process is proposed to iteratively improve the average meshes, while still preserving the geometric details. Our method can also scan static objects with normal material, and got very impressive results compared with the state-of-the-art method.In this paper, we present a novel scanning system for capturing3D full human body models by using multiple Kinects. We propose a practical approach for registering the various body parts of different views under non-rigid deformation. First, a rough mesh template is constructed and used to deform successive frames pairwisely. Second, global alignment is performed to distribute errors in the deformation space, which can solve the loop closure problem efficiently. Misalignment caused by complex occlusion can also be handled reasonably by our global alignment algorithm. The experimental results have shown the efficiency and applicability of our system.Finally, with the help of proposed scanning methods, the reconstructed3D body models can be applied in generating personalized digital avatar, virtual try on, and3D printing of static object with self-moving illusion similar to the captured human motion.