Bogazici University, Electric Electronic Eng. H. Isl Bozma .

Date: 1.3.2012 Due date: 14.3.2012

EE 576 - Project 1
The rst project aims to familiarize you with image formation and OpenCV along with binary images and thresholding. Please be sure to read the Project Grading Policy in the course web site. Please also make sure that your hand-in complies fully with the instructions as specied therein. 1. Consider the setup as shown in Fig. 1. There is a rectangular prism located located as shown in the gure. The image plane of the camera that is imaging this setup is also shown.

Figure 1: Problem 1

(a) What will be the corresponding image points of the ve 3D points Pi , i = 1, . . . , 5 (b) Now suppose that the rectangular prism is rotated around X3 axis by 45 clockwise. What will be the corresponding image points of the ve 3D points. (c) Now suppose that the rectangular prism is translated 1 m away from the camera. What will be the image points. If, instead of being translated backwards, if the object was translated 1 m towards the camera, what will be the image coordinates. What will be image points be, if the translation is along the X2 axis by the same amount. (d) Now suppose steps in (b) and (c) are done one after the other. Again, what are the corresponding image points. (e) Now suppose you have the 5 image points plus the image point of the center point on the front face of the object. Write a program that tries to compute the corresponding 3D points best as it can using the methods we have learned in the course for solving the calibration matrix. You may use OpenCV api if you like. If you do so, pls make sure you understand the approach tht is used. Please explain your reasoning. 2. Now add new rectangular prisms on the scene. Do this upto 5 objects. You decide where. Project them onto the image plane. Write a C/C++ program using OpenCV that that does the following: It should

consider the setup as presented in Problem 1. It should read in the following from a text le. 1. Number of objects 2. Object number 3. Orientation (X1,X2,X3 axis) + translation of the object Considering an image plane to consist of 576 480 evenly spaced imaging cells, it should then generate the corresponding image and show it. The above steps should be repeated in a loop until the user chooses exit. Do this for a variety of dierent object numbers.

Q3. Following, segment the objects in the image using either a region or contour based approach.

Notes regarding the programming part of the projects: Make sure that your code has the appropriate header les as necessary. Put all the methods in a separate C/C++ le. If you are familiar with OOP, you may use C++ and generate the appropriate classes as required. The main program should be in a separate le that is ready to run. You can either hand in a CD containing your project or email a zipped version. Please make sure that you have all the required les in the turned in CD or le.