VAlue proposition
Camera calibration is typically the first step in numerous vision and robotics applications that involve 3D sensing. Classic methods enable accurate camera calibration by imaging a specific 3D structure such as a checkerboard. With the rapid growth of monocular 3D vision, there is an increasing focus on 3D sensing for in-the-wild images, such as monocular depth estimation, 3D object detection, and 3D reconstruction. While techniques of 3D sensing over in-the-wild monocular images developed, camera calibration for such in-the-wild images continues to pose significant challenges. Classic methods for monocular calibration use strong geometry prior. However, such 3D structures are not always available in in-the-wild images. As a solution, alternative methods are needed.
Description of Technology
This technology is motivated by the consistency between the monocular depthmap and surface normal map. An incorrect intrinsic distorts the back-projected 3D point cloud from the depthmap, resulting in distorted surface normals. Based on this, intrinsic is optimal when the estimated monocular depthmap aligns consistently with the surface normal. The complete intrinsic is recovered by leveraging the consistency between the surface normal and depthmap. An additional novel 3D monocular prior is introduced in complementation to the depthmap and surface normal map. This incidence field depicts the incidence ray between the observed 3D point and the projected 2D pixel on the imaging plane. The combination of the incidence field and the monocular depthmap describes a 3D point cloud. The incidence field is a direct pixel-wise parameterization of the camera intrinsic. The incidence field is invariant to the image cropping or resizing. This enables its generalization over in-the-wild images.
Benefits
Applications
IP Status
Patent Pending
LICENSING RIGHTS AVAILABLE
Full licensing rights available
INVENTORS: Xiaoming Liu, Shengje Zhu, Abhinav Kumar
Tech ID: TEC2023-0137