1 / 33

System Architecture

System Architecture. Intelligently controlling image processing systems. Introduction. So far Presented methods of achieving goals Integration of methods? Controlling execution Incorporating knowledge. What knowledge?. What do algorithms achieve?

ranae
Télécharger la présentation

System Architecture

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. System Architecture Intelligently controlling image processing systems

  2. Introduction • So far • Presented methods of achieving goals • Integration of methods? • Controlling execution • Incorporating knowledge Image Processing and Computer Vision: 7

  3. What knowledge? • What do algorithms achieve? • What is known about the problem being solved? • Relationship between problem and algorithm? Image Processing and Computer Vision: 7

  4. Knowledge representation • Implied • Feature vectors • Relational structures • Hierarchical structures • Rules • Frames Image Processing and Computer Vision: 7

  5. Implied knowledge • Knowledge encoded in software • Usually inflexible in • Execution • Reuse • Simple to design and implement • Systems often unreliable Image Processing and Computer Vision: 7

  6. Feature vectors • As seen in statistical representations • Vector elements can be • Numerical • Symbolic coded numerically Image Processing and Computer Vision: 7

  7. Example: A N Image Processing and Computer Vision: 7

  8. Relational structures • Encodes relationships between • Objects • Parts of objects • Can become unwieldy for • Large scenes • Complex objects Image Processing and Computer Vision: 7

  9. Hierarchical structures Follow natural division of scene objects parts of object Image Processing and Computer Vision: 7

  10. Example: scene grassland roadway building road junction grass tree edges Image Processing and Computer Vision: 7

  11. Uses • Structure defines possible appearance of objects • Structure guides processing Image Processing and Computer Vision: 7

  12. Rules • Rules code quanta of knowledge • Interpretation • Forwards • Backwards <antecedent>  <action> <two antiparallel lines>  <road> Image Processing and Computer Vision: 7

  13. Forward chaining • If <antecedent> is TRUE • Execute <action> • Antecedent will be a test on some data • Action might modify the data • Suitable for low level processing Image Processing and Computer Vision: 7

  14. Backward chaining • Action is some goal to achieve • Antecedent defines how it should be achieved • Suitable for high level processing • Guides focus of system Image Processing and Computer Vision: 7

  15. Inference engine Database Rulebase System architecture Image Processing and Computer Vision: 7

  16. Frames A “data-structure for representing a stereotyped situation” Slot (attribute) Filler (value: atomic, link to another frame, default or empty, call to a function to fill the slot) Image Processing and Computer Vision: 7

  17. Methods of control • How to control how the system’s knowledge is used. • Hierarchical • Heterarchical Image Processing and Computer Vision: 7

  18. Hierarchical control • “Algorithm” defines control • Compare other software: • Main programme calls subroutines • Achieve a predefined sequence of tasks • Two extreme variants • Bottom-up • Top-down Image Processing and Computer Vision: 7

  19. Bottom-up control Object recognition Decision making Extracted features, Attributes, Relationships Feature extraction Image Image Processing and Computer Vision: 7

  20. Top-down control Hypothesised object Prediction Predicted features, Attributes, Relationships Directed feature extraction Features in image that Support or refute the hypothesis Image Processing and Computer Vision: 7

  21. Critique • Inflexible methods • Errors propagate • Hybrid control • Can make predictions • Verify • Modify predictions Image Processing and Computer Vision: 7

  22. Hybrid control Object recognition Decision making Prediction Extracted features, Attributes, Relationships Predicted features, Attributes, Relationships Feature extraction Direciction Image Image Processing and Computer Vision: 7

  23. Heterarchical control • “Data” defines control via knowledge sources • KSs contribute to process image • KS fires in response to presence of data • Creates new data • Modifies existing data • Can be chaotic • Blackboard Image Processing and Computer Vision: 7

  24. KS KS KS Blackboard architecture Blackboard scheduler Blackboard Image Processing and Computer Vision: 7

  25. Information integration • Hypotheses boolean • True or false • Facts are real valued True  certainty = 1.0 False  certainty = 0.0 Unsure  0.0 < certainty < 1.0 How is this represented? Image Processing and Computer Vision: 7

  26. Example Recognising cars Shape analyser - certainty = 0.56 Position analyser - certainty = 0.78 Texture analyser - certainty = 0.40 How to combine evidence? Image Processing and Computer Vision: 7

  27. F1 F2 F3 Bayesian methods • Define a belief network • A tree structure • Reflects evidential support of a fact Image Processing and Computer Vision: 7

  28. Propagation of certainty • Leaf nodes • Certainty given by basic operations • Non-leaf nodes • Combine child nodes’ certainties • Results propagate to root node Image Processing and Computer Vision: 7

  29. Dempster-Shafer • Bayesian theory has confidence in belief only • No measure of disbelief • D-S attempts to define this Image Processing and Computer Vision: 7

  30. Certainty interval 0 .. A = measures of belief A .. B = measures of uncertainty B .. 1 = measures of disbelief [A,B] starts large. As evidence accumulates to support or refute a hypothesis, A and B change Image Processing and Computer Vision: 7

  31. Other formalisms • Belief calculi exist • Not yet widely used • A result is important • Confidence in result is not quantified Image Processing and Computer Vision: 7

  32. Summary • Intelligent (vision) systems • Knowledge representation • Control strategies • Integration of belief Image Processing and Computer Vision: 7

  33. Everything that can be invented has been invented Charles Duell, Commissioner U.S. Office of Patents, 1899 Image Processing and Computer Vision: 7

More Related