1 Artificial Intelligence: Vision Stages of analysis Low level vision Surfaces and distance Object Matching.

Slides:



Advertisements
Similar presentations
ARTIFICIAL PASSENGER.
Advertisements

BELLWORK SINGLE POINT PERSPECTIVE
Measuring the height of Lunar Mountains using data from the Liverpool Telescope.
Imaging Science FundamentalsChester F. Carlson Center for Imaging Science The Human Visual System Part 2: Perception.
Computer Vision Lecture 16: Region Representation
Stills From Pan’s Labyrinth
December 5, 2013Computer Vision Lecture 20: Hidden Markov Models/Depth 1 Stereo Vision Due to the limited resolution of images, increasing the baseline.
1 of 25 1 of 22 Blind-Spot Experiment Draw an image similar to that below on a piece of paper (the dot and cross are about 6 inches apart) Close your right.
Advanced Computer Vision Introduction Goal and objectives To introduce the fundamental problems of computer vision. To introduce the main concepts and.
Imaging Science FundamentalsChester F. Carlson Center for Imaging Science Binocular Vision and The Perception of Depth.
Processing Digital Images. Filtering Analysis –Recognition Transmission.
Vision Computing An Introduction. Visual Perception Sight is our most impressive sense. It gives us, without conscious effort, detailed information about.
Introduction to Computer Vision 3D Vision Topic 9 Stereo Vision (I) CMPSCI 591A/691A CMPSCI 570/670.
1 Lecture 11 Scene Modeling. 2 Multiple Orthographic Views The easiest way is to project the scene with parallel orthographic projections. Fast rendering.
Computer Vision Marc Pollefeys COMP 256 Administrivia Classes: Mon & Wed, 11-12:15, SN115 Instructor: Marc Pollefeys (919) Room.
CS292 Computational Vision and Language Visual Features - Colour and Texture.
December 2, 2014Computer Vision Lecture 21: Image Understanding 1 Today’s topic is.. Image Understanding.
Composition (The elements). What is Composition Composition is the arrangement of shapes (forms) in an image – their position, relationship to one another.
November 29, 2004AI: Chapter 24: Perception1 Artificial Intelligence Chapter 24: Perception Michael Scherger Department of Computer Science Kent State.
Geometry Thru Composition. rectangles Using rectangles is a close likeness to Rule of Thirds. However, rather than keeping each section of your frame.
CIS 601 Fall 2004 Introduction to Computer Vision and Intelligent Systems Longin Jan Latecki Parts are based on lectures of Rolf Lakaemper and David Young.
October 8, 2013Computer Vision Lecture 11: The Hough Transform 1 Fitting Curve Models to Edges Most contours can be well described by combining several.
A Brief Overview of Computer Vision Jinxiang Chai.
Recap Low Level Vision –Input: pixel values from the imaging device –Data structure: 2D array, homogeneous –Processing: 2D neighborhood operations Histogram.
1 Perception and VR MONT 104S, Spring 2008 Lecture 22 Other Graphics Considerations Review.
CAP4730: Computational Structures in Computer Graphics 3D Concepts.
October 14, 2014Computer Vision Lecture 11: Image Segmentation I 1Contours How should we represent contours? A good contour representation should meet.
Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A.
Active Vision Key points: Acting to obtain information Eye movements Depth from motion parallax Extracting motion information from a spatio-temporal pattern.
How do we take the 2 -dimensional image that is projected onto the back of the eye, and figure out from that what the 3 -dimensional world that caused.
CIS 601 Fall 2003 Introduction to Computer Vision Longin Jan Latecki Based on the lectures of Rolf Lakaemper and David Young.
December 4, 2014Computer Vision Lecture 22: Depth 1 Stereo Vision Comparing the similar triangles PMC l and p l LC l, we get: Similarly, for PNC r and.
Photographic Composition Basic elements of composition help strengthen all types of photographs and digital images Quality and eye catching images are.
Computer Vision Why study Computer Vision? Images and movies are everywhere Fast-growing collection of useful applications –building representations.
1 Perception, Illusion and VR HNRS 299, Spring 2008 Lecture 8 Seeing Depth.
DIEGO AGUIRRE COMPUTER VISION INTRODUCTION 1. QUESTION What is Computer Vision? 2.
Vision Part 2 Theories on processing colors. Objectives: The Student Will Compare and contrast color theories (VENN) Explain the Gestalt Theory List your.
Elements of Art The elements of art are a set of visual techniques that describe ways of presenting artwork. Elements of art also refer to the visual language.
CS 376b Introduction to Computer Vision 03 / 21 / 2008 Instructor: Michael Eckmann.
Principles of Design. The Principles of Design are a set of guidelines artist’s use for two main reasons… To help them create artwork that is both pleasing.
Realtime NPR Toon and Pencil Shading Joel Jorgensen May 4, 2010.
Copyright Howie Choset, Renata Melamud, Al Costa, Vincent Lee-Shue, Sean Piper, Ryan de Jonckheere. All Rights Reserved Computer Vision.
Design Studies 20 ‘Show Off’ Project How to make a computer monitor In Google Sketchup By: Liam Jack.
Autonomous Robots Vision © Manfred Huber 2014.
(c) 2000, 2001 SNU CSE Biointelligence Lab Finding Region Another method for processing image  to find “regions” Finding regions  Finding outlines.
Visual Perception Principles Visual perception principles are ‘rules’ that we apply to visual information to assist our organisation and interpretation.
Visual Computing Computer Vision 2 INFO410 & INFO350 S2 2015
APECE-505 Intelligent System Engineering Basics of Digital Image Processing! Md. Atiqur Rahman Ahad Reference books: – Digital Image Processing, Gonzalez.
Reference books: – Digital Image Processing, Gonzalez & Woods. - Digital Image Processing, M. Joshi - Computer Vision – a modern approach, Forsyth & Ponce.
Colour and Texture. Extract 3-D information Using Vision Extract 3-D information for performing certain tasks such as manipulation, navigation, and recognition.
1 By Mike Maloney © 2003 Mike Maloney2 Light as a Ray Light very often travels in straight lines. We represent light using rays, which are straight lines.
October 1, 2013Computer Vision Lecture 9: From Edges to Contours 1 Canny Edge Detector However, usually there will still be noise in the array E[i, j],
1Ellen L. Walker 3D Vision Why? The world is 3D Not all useful information is readily available in 2D Why so hard? “Inverse problem”: one image = many.
Perception and VR MONT 104S, Fall 2008 Lecture 8 Seeing Depth
Anaglyph overview stereoscopic viewing technology.
Robotics Chapter 6 – Machine Vision Dr. Amit Goradia.
How we actively interpret our environment..  Perception: The process in which we understand sensory information.  Illusions are powerful examples of.
Another Example: Circle Detection
A Plane-Based Approach to Mondrian Stereo Matching
DIGITAL SIGNAL PROCESSING
Mean Shift Segmentation
Properties of human stereo processing
Common Classification Tasks
Fitting Curve Models to Edges
Brief Review of Recognition + Context
Perception.
Introduction to Artificial Intelligence Lecture 24: Computer Vision IV
CSE (c) S. Tanimoto, 2007 Image Understanding
Introduction to Artificial Intelligence Lecture 22: Computer Vision II
Presentation transcript:

1 Artificial Intelligence: Vision Stages of analysis Low level vision Surfaces and distance Object Matching

2 Introduction Another “mundane” task involves being able to make sense of what we see. We can handle images of objects differing in: size orientation color lighting expression (for faces etc) obscured by other objects And recognize the objects in the scene, and what is happening in the scene.

3 Vision task Ultimate task: from visual signal (digitized image) to representation of the scene adequate for carrying out actions on the objects on the scene. E.g., image of parts of device --> representation of location, orientation, shape, type etc of parts enabling robot to assemble device. More limited task: recognize objects (from limited set) - is it a widget, wodget or wadget?

4 Stages of processing Like NLP, we are mapping from an unstructured raw signal to a structured meaningful representation. Like NLP we do it in stages: Digitization - raw data -> digitized image (e.g., 2d array of intensity/brightness) Low level processing - identify features like lines/edges from the raw image. Medium level - determine distances and orientation of surfaces. High level - Create useful high level representation (e.g., 3-d models, with objects and parts identified)

5 Low level Processing Ignore digitization. First task then is to extract some primitive features from the image. We might have a 512x512 image, where each image point (pixel) has a certain image intensity or brightness, represented by a number For color need three numbers per image point (blue, green, red), but start just considering b&w. We start with a “grey-level” image. Image intensity sometimes called grey level.

6 Edge Detection Consider the image below: First task is to find the “edges” of the image. We obtain, from the array of grey levels, a “sketch” consisting of a number of lines.

7 Simplifying.. Lets see what it might look like as an array of intensity values (ignoring door, window) Edges occur where the intensity value changes significantly. We find the difference between intensity values at neighboring points, and if large, mark poss. edge.

8 Applying difference operation Just considering horizontal differences, and marking when the difference is greater than a threshold of 3, we get the following: Have found vertical sides of house and bits of roof. Similar operations let us find other edges.

9 Line Fitting We’ve now got a simplified image with points corresponding to edges in an image marked in. Next task is to get from that to a set of lines. This reduces the amount of data and gets closer to useful representation.

10 Simple Approach: Tracking Find an edge point. Look at all surrounding points to find connected edge points. Keep going while the points you are finding form a straight line. When no more points in that direction, stop and make last one end point of line.

11 Problems.. What about: curved lines obscured lines (e.g., edge of an object, when parts of that edge are obscured by another object). Solution is to try and find candidate likes such that the number of edge points falling on that line is maximized. We consider all lines, and find those that seem to have lots of edge points on them.

12 Surfaces We’ve looked first at low level vision: Find candidate edge points where intensity level changes quickly. Find lines (where many edge points fall on possible line). Next stage is to find surfaces, and their distance from viewer and orientation. This gives us a 3-d model of object(s) in the scene.

13 Consider Is this: Rectangle with right hand side near viewer OR 4 sided shape with RHS longer than LHS. A small surface near the viewer, or a large surface a long way away from the viewer? What cues to we have to help us determine this?

14 Sources of depth info Stereo vision: Two eyes give slightly different images, allowing distance estimates. Motion: If the viewer is moving, again we get multiple images which give us a clue. Shading and texture. Consider:

15 Stereo Vision How do the different images from our two eyes enable us to guess at distances? Try holding pencil close to your eyes - close one eye then another. Pencil will be in different locations. Now move it further away. The amount the pencil “moves” will be less. The difference in direction of an object, from one eye and from the other, depends on the distance away. eyes

16 Stereo Vision 1. Find corresponding image features in two images. 2. Work out from that the angle that the feature would be from the camera (for each camera). 3. Do some geometry to get the distance. Image from left eye/camera Image from right eye/camera

17 Stereo Vision The math ends up quite easy. For those who can recollect school trigonometry.. Z = b sin  sin  / sin(180-  -  ) z is distance we are looking for b = distance between eyes theta is angle from one eye to object alpha is angle form other eye to object. But tricky bit is the feature matching - how do we tell that an image point from one camera corresponds to a particular point from other.

18 Depth from Motion Humans also use motion as an important cue for depth/orientation. Try holding pencil near you, and moving head while keeping pencil still. Relative positions of pencil and background will change. Diagram/Math similar to stereo vision.. (“eyes” now correspond to camera location at two time points). Techniques however are different.

19 Texture and Shading Texture, or regular repeated patterns helps in determining orientation (see slide 4). We assume that the patterning is regular and conventional. Shading also helps, especially if we know the light source. Impression of curving?

20 Object Recognition We now have the tools to extract from an image a set of surface features, with given distances and orientation. E.g., feature 1 is a 20cmx40cm rectangular surface 2m away, sloping away from the viewer at angle 40º. Next step is to: Put these together into a 3-d model. Recognize that it is a widget..

21 Object Models We need to have a way of describing shapes of objects of interest (e.g., widgets), and also describing shapes in the scene. Need 3-d models, so we can recognize objects from different viewing angles. Base these on “volumetric primitives” (ie, 3d shapes) (e.g., cube, cylinder). Now our first stage is to get from our surfaces, to possible shapes. Reasonably easy, if surfaces are right and no obscuring objects.

22 Object Models Image of a house might end up with model: Pyramid + Cube

23 Matching.. Now if we have stored the fact that house =pyramid on top of cube We should be able to recognize the image as a house whatever orientation we view the house from. So we match candidate object models to model of object(s) in scene, and find the closest match.

24 Summary Vision - from grey level image to recognized object and model of scene. Start with low level vision: Find candidate edge points where intensity level changes quickly. Find lines (where many edge points fall on possible line). Then find surfaces, distances, match to 3d primitives and object models.