Is beststudentpaper hard to get?

It is even more difficult to get the best student papers. In 2022, CVPR * * submitted more than 8,000 papers, and only one was selected. It is not an exaggeration to say that one in a million.

The core problem of his thesis is PnP, which determines the position and posture of 3D objects through a single picture.

The PnP problem originated in the field of 184 1 photogrammetry, and was formally proposed by Fishler and Bolles in 198 1. It is a classic problem in the field of computer vision. The PnP problem is also called the pose estimation problem of given control points, that is, under the condition of given camera internal parameters, shooting an image containing n spatial reference points, knowing the coordinates of these n reference points in the world coordinate system and their corresponding projection points in the image coordinate system, and calculating the position and pose of the camera when shooting the image through their corresponding relationships.

PnP problem can be divided into two categories according to the number n of spatial reference points: one is the case of 3 ≤ n ≤5, and the other is the case of n ≥6.

The first case can be divided into P3P problem, P4P problem and P5P problem. For P3P problem, when the plane determined by three spatial reference points does not pass through the optical center, there are at most four solutions. For the P4P problem, there is a unique solution when all four spatial reference points are * * * planes, and there are at most five solutions when there is no * * * plane. For the P5P problem, when any three reference points are not * * * lines, there can be at most two solutions. This kind of PnP problem usually has multiple solutions because of the small number of reference points used for calculation, and it is extremely sensitive to image noise and has weak anti-interference ability. In the second kind of PnP problem, there are many reference points to calculate the attitude, so the research focus of PnP problem in this case is the accuracy, time complexity, robustness and noise resistance of the algorithm.