3D Scanning Based on Computer Vision By:Hansen Chen Xiaobo Dong Xingqian Xu TA: Mustafa Mir 04/27/2012
Contents Introduction Motivation Review original Design Describe project build and function test Discuss Success and Challenges Recommendation Special Thanks
Introduction 3D scanning will move from industry to home. People would like to see 3D Models when shopping online. Project features: Computer Vision Technique Web camera Matlab
Motivation Economical --- using computer vision technique instead of laser. Interactive --- 3ds format. Fun Project --- We like it.
Review Original Design What is changed?
Project build Hardware Boost converter with feedback control. Integrated Platform with camera and Led light control. 5 hours operating time with 2 AA batteries.
Boost Converter With feedback control
Microcontroller We use Microcontroller 430G2131 to control the PWM wave to the transistor, the microcontroller can modify the duty cycle from 15% up to 65% depending on the feedback voltage.
Project build Hardware (3)
Segmentation (1) Segmentation Overview Prerequisite: Dark surrounding with constant white light background Input: Recorded Video Method: Intensity thresholding method (Otsu's method for threshold value) Output: Images for the objects’ shape in 36 different view points.
Segmentation inside boundary region Finding Image boundary Input Video Finding Image boundary
Select the marker region and processing Segmentation (3) Calculating the angle for each frames given by the position of the markers Image Cleaning (Delete Noisy) Select the marker region and processing
Selecting the frames given by the angles Segmentation (4) Selecting the frames given by the angles Output Reconstruction step
Angle Calculation(1) Detect the position of the Marker
Angle Calculation(2) By Theorems in geometry There is a relation between x,y and angle theta. r is the radius
3D Reconstruction Algorithm: use 360 degree projection view to reconstruct test object.
3D Reconstruction Step 1: Object Zoom as camera distance change
3D Reconstruction Step 2: 3D Matrix rotate to correct angle
3D Reconstruction Step 3: 3D Matrices merge into one
3DS formation Step 1: Writing the correct protocol:
3DS formation Step 2: Meshing (Layer Algorithm):
Matlab GUI Start Button Camera Select Lots of GUI Codes!!! Test Cases Fast Run Are honestly generated automatically by matlab
Successes Successful motor control Segmentation algorithm success, accurate segmentation results are obtained 3D reconstruction algorithm success, 3D model can be reconstructed. 3ds formation file can be generated correctly.
Challenges Concave surfaces Environmental sensitive Object off platform center Converting 3ds cubic structure file into mesh structure file
Verification Case Mouse
Verification Case Toy SpongeBob
Recommendation Implementation(if time allows) Cover case Mesh the output Front light to capture color information Application Online retailing Add on for 3D printer 3D gaming etc..
Special thanks Professor P. Scott Carney TA Mustafa Mir Group 36—Jing Guo, Yuqiao. Power Lab Kevin Colravy Professor Philip Krein Free Samples From TI