課程名稱︰三維電腦視覺與深度學習應用
課程性質︰資工所選修
課程教師︰陳祝嵩
開課學院：電機資訊學院
開課系所︰資工所
考試日期（年月日）︰2021/04/22
考試時限（分鐘）：160分鐘
試題 :
憑印象，不一定是原卷用字遣詞
可帶雙面A4大抄
1.
(a) Write and explain the components of a camera intrinsic matrix.
(b) What is nonlinear distortion? Write one example and explain.
2.
Prove that the homography H_R describing a camera with only pure rotation
(rotation matrix R) and known intrinsic matrix K is H_R = KRK^-1
3.
Write and explain the minimal 2D-3D correspondences required for solving the
following problem:
(a) PnP problem using DLT method with unknown intrinsics.
(b) PnP problem using EPnP method with known intrinsics.
4.
Suppose we have a 2-camera system which is rectified. Write down the
essential matrix describing the system. Introduce your own terms if
necessary.
5.
Explain the process of retrieving the visual word from an image
(using Bag of Words for example).
6.
Suppose using RANSAC on a line fitting problem with a set of 100 points,
in which 20 are outliers. Calculate the number of draws if we want a
99% chance of getting a pure-inlier sample.
(log 2 = 0.3010, log 3 = 0.4771)
7.
Suppose there are 2 cameras looking at a common world point P. The rays from
P to the camera origin C1 and C2 passes through both image planes at the
center, which is (0, 0), i.e. the rays C1P and C2P coincides with the
principal axes of 2 cameras respectively.
Prove that in this case, the fundamental matrix F describing the system has
term F_33 = 0.