Computer Vision

Presentation of lectures

Department of Computer Graphics and Multimedia

 Topic

Introduction, motivation,
basic principles, application
Motto

„What is computer vision?

How is it used at all?
What are the applications?“

Computer Vision / Introduction, motivation, basic principles, applications 3

Contents:
• What is computer vision
• Basic principles
• Application

Computer Vision / Introduction, motivation, basic principles, applications 4

What is computer vision?
• Computer vision, as the title suggest, deals with „modeling“
of vision of live creatures/humans by machines
• Computer Vision is VERY open problem
• Computer vision, however, can be VERY useful now
• Computer vision today should be understood as a „set of
individual techniques“ for modeling of vision rather than
„complex replacement of human vision“

• An open problem of computer vision is even what should be

its result – modeling of how „reality“ is seen at all is also a
very open problem

Computer Vision / Introduction, motivation, basic principles, applications 5

What is computer vision?
• Sicentific/technical discipline that explores methods for
acquisition of Information about the world from images and
that attempts to implement these principles using available
technical means/hardware.
• The goal is not only to model human (live creatures) vision.
Sometimes this is not feasible and sometimes biolgical
means are developed (live creatures did not need it).
• Image acquisition means often have better features than
human eyes and available computer vision tasks often
achieve better results than human in individual tasks:
- Face recognition from photographs
- Scene and surface reconstruction
- Image seach in large data bases
• Humans are unable to do some tasks that computers can.
Computer Vision / Introduction, motivation, basic principles, applications 6
What is vision at all?
• Process of creation of believable/trustworthy
representations of the outside world based on images
and/or image sequences.

… well, is this true at all ??? …

Computer Vision / Introduction, motivation, basic principles, applications 7

What is image?
• This is notorically known but …

Pinhole camera – rays go through infinitely small hole and they

are captured on the opposite side of camera.

Computer Vision / Introduction, motivation, basic principles, applications 8

What can be understood from image?
• Shape and opsition of objects, theis optical features (colour,
texture, smoothness), motion.
• Semantic information
- Cathegory and identity of objects
- Activity
- Meaning from the observer point of view that he/she/it
needs for decisions.

Computer Vision / Introduction, motivation, basic principles, applications 9

Image is almost always ambiguous!
• Every image has infinitely many gometrical interpretations
http://www.youtube.com/watch?v=4q0zBWG-Sso&feature=related http://www.youtube.com/watch?v=4q0zBWG-
Sso&feature=related http://www.youtube.com/watch?v=gcw1IIGSGMM

Computer Vision / Introduction, motivation, basic principles, applications 10

…image is almost always ambiguous!

Computer Vision / Introduction, motivation, basic principles, applications 11

How to deal with ambiguity?
• Hard to say but humans (probably) use:
- knowledge about possible status of the world.
- most of the possible interpretations of images do not
match real world reality (so that we can „forget them“).

Vision… Estimation of the most probable interpretation of

image based on the individual experience?

Computer Vision / Introduction, motivation, basic principles, applications 12

How to deal with ambiguity?
• Straight lines in image are probably straight also in reality
• Lines whose ends „touch“ in the image probably touch also
in reality, the same with objects.

Computer Vision / Introduction, motivation, basic principles, applications 13

How to deal with ambiguity?
• Objects do overlap…

Computer Vision / Introduction, motivation, basic principles, applications 14

How to deal with ambiguity?
• Humans see 3D objects where they are not…

Computer Vision / Introduction, motivation, basic principles, applications 15

How to deal with ambiguity?
• Humans use expectations about perrspective (correctly)?

Computer Vision / Introduction, motivation, basic principles, applications 16

How to deal with ambiguity?
• … well, sometimes incorrectly …

Computer Vision / Introduction, motivation, basic principles, applications 17

How to deal with ambiguity?
• Humans interpret shadows…

Computer Vision / Introduction, motivation, basic principles, applications 18

How to deal with ambiguity?
• … sometimes wrong …

Computer Vision / Introduction, motivation, basic principles, applications 19

How to deal with ambiguity?
What is the difference between the images?
What is the expectation about the real world…
…and causes differences in interpretations?

Computer Vision / Introduction, motivation, basic principles, applications 20

More and more features
Clustering:
• Humans tend to see patterns and clusters (of close) objects.
• Why?
o It organizes the scene and simplpifies its representation.

Motion:
• Simple motion can have more interpretations, can you see them?
http://www.youtube.com/watch?v=DLBkwig3M2U&feature=related
• Old interesting video http://www.youtube.com/watch?v=KT89CQ2nRpo&feature=relmfu
http://www.youtube.com/watch?feature=endscreen&NR=1&v=1F5ICP9SYLU

Computer Vision / Introduction, motivation, basic principles, applications 21

Basic principles of computer vision
• Direct processing of intensity/colour information, simple
filters, etc. (image processing)
• Subdivision of image into regions of similar
intensity/colour/texture. (segmentation, image processing)
• Creation of scene models – construction of shapes,
topology, 3D models including their surfaces from 1 image
• Processing of more images (sequences and ad-hoc) for
model creation and mutual position (including camera)
• Detection of objects in images based on different principles
similarity, models of shapes, topology, etc.
• Recognition of objects of known type (biometry)
• Motion detection, motion analysis, tracking, etc.

Computer Vision / Introduction, motivation, basic principles, applications 22

Applications of vision in industry
• Quality control of products
- shape (also 3D)
- location
- surface
- description
• Safety of work
• Various control tasks

Computer Vision / Introduction, motivation, basic principles, applications 23

Measurements
• Advanced edge detection
• Subpixel dimension measurements

Computer Vision / Introduction, motivation, basic principles, applications 24

Surface evaluations
• Light sources in differents parts of spectrum
• Detection of irregularities and cracks

Computer Vision / Introduction, motivation, basic principles, applications 25

Search of shapes in images
• Character recognition
• Shape verification (logo)
• Localization of labels and shapes

Computer Vision / Introduction, motivation, basic principles, applications 26

Further applications
• Quality control of beer bottles (special optics)

Computer Vision / Introduction, motivation, basic principles, applications 27

3D measurements
• Evaluation of glue drop volumes
• Evaluation of welding quality

Computer Vision / Introduction, motivation, basic principles, applications 28

3D measurements
• Evaluation of quality of car wheel discs

Computer Vision / Introduction, motivation, basic principles, applications 29

Evaluation of large surfaces
• Monitoring of non-wowen textiles

Computer Vision / Introduction, motivation, basic principles, applications 30

Applicaion of vision in traffic
• Enforcement and documentation
• Detection of communications status
• Travel time evaluations
• Traffic researches
• Detection of irregular states
• Search of vehicles (individual)

Computer Vision / Introduction, motivation, basic principles, applications 31

Images in traffic
• HDTV and better resolution (quality B/W)

Computer Vision / Introduction, motivation, basic principles, applications 32

Intelligent cameras
• All in one construction

Computer Vision / Introduction, motivation, basic principles, applications 33

Night vision
• UnicamINFRA – infrared reflector

Computer Vision / Introduction, motivation, basic principles, applications 34

Night vision
• Night shot of a fast moving vehicle

Computer Vision / Introduction, motivation, basic principles, applications 35

Camera + IR reflector
• Installation in tunnels

Computer Vision / Introduction, motivation, basic principles, applications 36

Registration plate reading
• Image and read license plate

Computer Vision / Introduction, motivation, basic principles, applications 37

Section speed measurement

Computer Vision / Introduction, motivation, basic principles, applications 38

Section speed measurement

Computer Vision / Introduction, motivation, basic principles, applications 39

Section speed measurement

Computer Vision / Introduction, motivation, basic principles, applications 40

Red light violations

Computer Vision / Introduction, motivation, basic principles, applications 41

Red light violations

Computer Vision / Introduction, motivation, basic principles, applications 42

Traffic monitoring device
• Profile scan, camera, IR reflector

Computer Vision / Introduction, motivation, basic principles, applications 43

Vehicle profile scan
• Passenger car

• Lorry (below 3.5t)

Computer Vision / Introduction, motivation, basic principles, applications 44

Vehicle profile scan
• Lorry

• Lorry

Computer Vision / Introduction, motivation, basic principles, applications 45

Vehicle profile scan
• Lorry (alternative)

• Lorry (alternative)

Computer Vision / Introduction, motivation, basic principles, applications 46

Applications in human machine interaction
• Detection of humans (presence)
• Detection of identitiy
• Detection of activity
• Detection of human body parts
• Detection of human face expression

Computer Vision / Introduction, motivation, basic principles, applications 47

„Feature extraction“ image and video
• Feature extraction

Computer Vision / Introduction, motivation, basic principles, applications 48

„Feature extraction“ image and video
• Localization based on classifiers and
„Gabor Wavelet Networks“

Computer Vision / Introduction, motivation, basic principles, applications 49

„Feature extraction“ image and video
• Body parts detection

(video…)

Computer Vision / Introduction, motivation, basic principles, applications 50

 video editing
Setup (Desired
information, editing
properties)

Features
extraction

Input Camera
video Video editing selection
streams algorithm
(rules)

Scenario Output
video
stream
Video editor

Computer Vision / Introduction, motivation, basic principles, applications 51

Recording and tracking in video sequences
„Mobile meeting room“

(video…)

Computer Vision / Introduction, motivation, basic principles, applications 52

Point of interest localization
• Repeatable searchable points
(changes camera location, intensity, colour…)
• Such a set of points can be used in 3D reconstruction e.g.
using RANSAC (RAndom SAmple Concensus)

Computer Vision / Introduction, motivation, basic principles, applications 53

Automatic orientation and image assembly
• Automatic localization of camera in 3D space and/or
assembly of images acquired using such camera

Computer Vision / Introduction, motivation, basic principles, applications 54

Acceleration of processing in FPGA/DSP/GPU
• The purpose is to reduce the CPU load and/or
creation of standalone systems
• Acceleration in DSP / FPGA – often better in price/performance and
power consumption comparing to traditional solutions

Computer Vision / Introduction, motivation, basic principles, applications 55

Thanks for your attention!