A29401 No Calculator permitted in this examination -1- Turn Over School of Computer Science First Year – MSc Advanced Computer Science First Year – MSc Robotics 06 25024 Robot Vision Summer Examinations 2013 Time allowed: 1 hr 30 min [Answer ALL Questions] No Calculator A29401 -2- Turn Over [Answer ALL Questions] 1. (a) Describe a pinhole camera. [3%] (b) Describe the role of lenses in cameras. [3%] (c) What is a chromatic aberration? What causes it? [2%] (d) What is vignetting? What does it cause? [2%] 2. (a) Describe “kernel separability” and its importance for image filtering. [5%] (b) For the kernel given below determine if it is separable. Justify your answer. [5%] -1 -2 -1 2 4 2 3 6 3 3. (a) In pseudocode, describe the Hough transform for detecting circles with fixed radii. [7%] (b) Demonstrate the algorithm on a pair of points A=(3,4) and B=(5,4) for the values of radii r=1 and r=2. Map the two points into the parametric space. What can we infer about the points A and B from the analysis of the parametric space? [8%] 4. (a) In pseudocode, describe RANSAC for a general case of searching for the best model in the set of data. [10%] (b) Explain the role of RANSAC parameters. [5%] 5. (a) Describe the process of detecting interest points with a DoG (Difference of Gaussians) pyramid (as proposed by Lowe). [10%] (b) How do we achieve in-image-plane rotational invariance of a SIFT descriptor? [5%] No Calculator A29401 -3- End of Paper 6. (a) Given the sketch of a two-camera system (Ol, Or denote the left centre and the right centre of the cameras, respectively), and a 3D point P, draw and explain the terms epipols, epipolar plane, and epipolar lines that correspond to the point. [10%] (b) A mobile robot is equipped with two cameras (assume a pinhole model). The optical axes of the cameras are parallel and lie on a horizontal plane. The baseline between the cameras is 12 cm. The disparity of a point T in space is 16 mm. When the robot moves 20 cm in the viewing direction, the disparity increases to 30 mm. Calculate the focal length. [10%] 7. (a) Describe the steps involved in object learning and recognition using PCA. [10%] (b) Describe object learning and categorization using “bag of features”. [5%] . P Ol . .Or































































































































学霸联盟