You are on page 1of 39

Research and Machine Vision

Joel P. Ilao November 23 & 26, 2011

Some Philippine Facts


2,180 HEIs: 607 public, 1573 private ( from CHED website) no PH university in top 300 of World University Rankings, 2011 (DLSU: 451-500). 3 PH universities within world top 500. PH ranked 75th out of 142 countries in Global Competitiveness Report 2011-2012 (up by 10 spots from last year) by World Economic Forum (WEF)
improved by 2 points in Higher Educ. Category and training

Taken from Unesco Institute for Statistics (UIS) Fact Sheet (August, 2011)

country PH SG TH

Year 2005 2007 2005 2005 2002

Population (in Millions) 92.2 5 63.4 231.4 85.8

RSE per million 81 6088 311 162 115

Source: Unesco Science Report, 2010

ID VN

DOST National Science and Technology Plan (NSTP) for 2002-2020


Long Term Area thrusts
agriculture, forestry and natural resources health / medical sciences biotech ICT microelectronics material sciences and engineering earth and marine sciences fisheries and aquaculture environment natural disaster mitigation Energy manufacturing and process engineering

Research Process

Research Process

Digital Image Processing


DIP is the analysis and manipulation of digitized images through the use of computer

Sub-Areas of DIP
Image Compression 3-D Vision Image Segmentation

Pattern Recognition

Filtering

Noise Reduction in Images

System Flowcharts
Input Feature Extractor

Pre-processing Algo 1 Algo 2

Database

Main Algo

Post-processing Combine Intermediate Output

Output

Tools for conducting research


Hardware:
computer camera

software:
Matlab and Matlab clones
Octave, Mathematica, SciLab

IDEs with image processing libraries


openCV for C/C++, AForge.Net for C#, etc.

REVIEW OF DSP

Signal Modification
Amplification

y (t ) = x(t ) y (t ) = x(kt )

Sub-sampling / scaling

Linear Systems
Definition: A system is linear iff it satisfies: 1 y1 (t ) + 2 y2 (t ) = H [1 x1 (t ) + 2 x2 (t )] Example in DIP:
AWGN riding on digital images

Transfer function: y = f(x,y)


one interesting problem in DIP is estimation of transfer function application in super-resolution

Time vs. Frequency Domain


Converting from time to freq. domain and back
Fourier Transform Inverse Fourier Transform other transforms: (i.e. wavelet transform, discrete cosine transform, etc.)

Histogram Analysis
Image Segmentation

SAMPLE MV AND DIP PROJECTS

Vision-based Vehicle Tracking

Vision-based Vehicle Tracking

Overall Block Diagram

Hypothesis Generation Block

Finding vehicles based on contour (Hypothesis Verification)

Finding vehicles based on contour (Hypothesis Verification)

Vision-based Vehicle Tracking

License Plate Recognition

Vision-based Object Counting System

Vision-based Object Counting System


Noise suppressed image Segmented objects

Image Preprocessing

Object Segmentation

Attribute Extraction
Attribute values Reference image attribute values

Input image file

User

Selected attributes

Comparison
Results

Reference Attribute Database

Display Results

Human Gait Analysis

70

60

50

40

30

20

10

50

100

150

200

250

300

350

400

Humain Gait Analysis

Depth Estimation and Image Segmentation using Monocular Images

Depth Estimation and Image Segmentation using Monocular Images

Depth Estimation and Image Segmentation using Monocular Images

Facial Expression Tracking and Mimicking

Facial Expression Tracking and Mimicking

Fingerprint Authentication

Fingerprint Identification

Fingerprint Identification

Vision-based Hand Mimicking System

Left Image

Right Image

Vision-based Hand Mimicking System


User

Hand Movements

Image Acquisition Module Segmented Colored Image 2D Coordinates of Hand Parts 2D Points Tracker Module Angles of Hand Parts Angle Reconstruction Module 3D Reconstructed Model Angles of Hand Parts Instruction

Simulation Module

Assignment for next meeting


Group yourselves into groups of 4 Each group will propose system architectures /block diagrams and flowcharts for (a) a sample MV problem, and (b) MV problem of their choice.
keep proposals unique and creative include references when needed clearly state reasons for choices

Sample MV problem (chick counter)

You might also like