Ben Schneider
Zack Sorenson
Neil Agarwal
Bailey Zhang
Vision Recognition Outline
What is Vision Recognition?
Definition
Computational Vision
Attempts to understand visual perception and understanding Machine Vision
Attempts to create practical application for computer vision
How does it relate to AI?
Creating useful autonomous agents
-Robocup is the tournament where they have dogs play soccer - obviously
vision is useful here. They have a goal to by 2050 create bipedal robots
capable of beating the best human soccer players.
-Asimo is a bipedal robot which has cameras that allow it to interact with
its surroundings.
Human/computer interfacing
-Gesture and gait recognition.
-GRACE is a robot that uses vision to recognize her surroundings.
GRACE attended a conference, navigating her way through a lobby and up
through and elevator. GRACE also has a newly installed binocular vision
system that allows her to recognize human gestures.
-Recognition of handwritten text and drawings.
Passive observation & analysis
-For example observing a landscape and looking for changes over time.
-Traffic and surveillance cameras that automatically identify passing objects.
Understanding biological vision
-"Spatial reasoning" is something we all talk about but no one really
understands.
What's being done in Vision Recognition?
Penn State Computer Vision Lab
Hand gesture recognition as supplement to speech recognition.
Text recognition and clarification in video
Current Implementation
Optical Character Recognition (OCR)
Types of OCR
Character-based word recognition
Segmentation-based word recognition
Word-shape recognition
Current Limitations of OCR
Underlines in text
`I' and `i' confusions
Distortion of `e'
Examples of OCR software
Calera Wordscan Plus 1.0
Caere Omnipage Professional 3.0
Xerox Imaging Systems AccuText 3.0
Scarecrow
- Floating, autonomous robot platform powered by solar panels
- Searches for birds using its vision system
- Goes around, when its senses birds, bumps them or shoots water at them
- Floats in catfish ponds to protect the stock which is frequently depleted
by hungry birds
Successes
Machine Vision is being developed to control the quality of products, detecting
defects automatically
The technology is reaching a point where people can be tracked in 3D, leading to
more robust security systems.
And for your entertainment, Machine vision is being developed to track actual
movements, for more realistic and exciting games
Plus, more specific examples
Problems and setbacks
Finding 3D objects in 2D images
-Researchers are trying out binocular vision systems, but they work only
for hardware, not software.
Making raster images usable for a computer
-Usually all the computer has is a bunch of pixels that it has to convert into
meaningful shapes.
-The general steps for recognizing an object within an image are:
-Distinguish the object from its surroundings
-Find the shape of the object
-Recognize the object as a chair, apple, etc.
-Mathematical algorithms like edge detection and averaging are used
Incomplete understanding of human vision
-Even as AI researchers look into computer vision, biologists continue to
look into human and animal vision
-AI people genereally like to look at vision as a bunch of discrete, isolate
processes that each do one thing, but in biological vision systems the
different systems may overlap and meld together
Future of Vision Recognition
PC's Recognizing Users (Microsoft Project)
- When a user sits in front of the computer, it will bring up his work, etc.
- PC equipped with a camera on top
- Long term: aiming for a system that can recognize eye movements, so
that software could be controlled by eye movement
- Only a couple years from reaching mass market?
Augmented Reality
Used for surveying
More exciting gaming
Tourist "goggles"