From Pherographia To Color Pherographia
Download as PPTX, PDF0 likes1,016 views
This document describes the evolution of models for simulating ant colony behavior and its application to image processing techniques called pherographia. It begins with the original 1995 ant model by Chialvo and Millonas and describes subsequent extensions by other researchers to apply it to digital images. These include adding terms for pheromone deposition rates and ant reproduction. It also summarizes the development of color pherographia using different color spaces and rules for ant movement. The document concludes by discussing results and potential future work extending the techniques.
From Pherographia To Color Pherographia
- 1. From Pherographia to Color Pherographia ColorSketchingwith Artificial AntsC.M. Fernandes12C. Isidoro2F. Barata2J.J. Merelo1A.C. Rosa21University of Granada2Technical UniversityofLisbon
- 2. Summary2Original b/w photoPheromonedensityafter 100 iterationsAntsafter 100 iterationsThe Ant System: from Chialvo and Millonas’s Ant Model to Pherographia.Artwork created with monochromatic pherographia. Color Pherographia (four variations).Results.Conclusions and future work.CongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 2011
- 3. The Original ModelCongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 20113Dante Chialvo and Mark Millonas, How Ants Build Cognitive Maps, 1995 The model simulates a swarm of ants moving in an homogeneous environment.
- 4. A population of ants is randomly distributed in a two-dimensional array.
- 5. The ants move one step (cell) in each iteration, following simple rules.
- 6. Global and complex behaviour emerges from the simple rules and from the indirect interaction of the ants via the environment.
- 8. Stigmergy
- 9. Simple ((local, no explicit memory, homogeneous and isotropic)The Original Model (Chialvo and Milonas)4Evaporation: pheromonedensityineachcellisdecreasedby a constantamount.Pheromonedeposition rate: T= ηincreasingthecell’spheromonedensity σ Pheromone weighting function:β: degree of randomnesswithwhichtheantsfollowthepheromonegradientδ: saturationeffect, ant’sabilitytosensepheromonedecreases at highconcentrationsNormalised Transition probabilities (probability to go from cell k to cell i)w = 1/20w = 1/12w = 1/12w = 1/4w = 1/4w = 1/2w = 1/2CongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 2011w = 1
- 10. The Original ModelCongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 20115
- 11. Ramos and Almeida’s Model6Ramos and Almeida, Artificial AntColoniesin Digital Images Habitat, ANTS 2000- Instead of constant pheromone deposition rate, a term not constant is included:ConstantGives a measure of similaritybetween two different latticewindows, in terms of grey levelspatial arrangement.CongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 2011
- 12. Final Model7Fernandes, Ramos and Rosa, Self-Regulated Artificial AntColonieson Digital Image Habitats, InternationalJournalof Lateral Computing, 2005An “evolutionary” component is added to the ant system.Each ants has an initial energy that decreases in each time step: the probability of surviving depends on the energy: P = 1-e(a)Each ant is allowed to reproduce in each time step. The reproduction probability depends on the number of neighbouring ants and the pheromone density. W(0) = W(8) =0; W(4) = 1; W(5) = W(3) =0.75; W(6) = W(2) =0.5;W(7) = W(1) = 0.25. CongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 2011
- 13. Timor Mortis Conturbat Me (2008)“Timor Mortis...” was exhibited at the P4Photography art gallery, in Lisbon.About this work: http://carlosmfernandes.com/index_archivos/Page768.htmCongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 20118
- 15. The Horse and the Ants (2009)The Horse and the Ants has been exhibited in several art and science shows.CongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201110
- 16. Criatividade Computacional, ISCTE, Abril de 201011
- 17. Studies for a Modern Zoetrope (2011)CongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201112
- 18. Becher’s typologies: analysis and synthesisCongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201113Bernd e HillaBecherIdris Kahn
- 19. Becher’stipologies: analysis and synthesisCongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201114
- 20. ColorPherography15The main difference is the equation:
- 21. In b/w pherography, ∆ measuresthecontrastintheregionaroundtheant’sposition.
- 23. First, RBG isconverted to L, a and b (Labcolorspace)
- 25. a and b measurethecolor.
- 26. ∆ istheEuclideandistancebetweentheaverageof L, a and b, eachaveragedoverthecellandits 8 neighborigcells, andtheaverageof L, a and b averagedoverthepreviouscellandits 8 neighboringcells. CongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 2011
- 27. ColorPherographyFour variations were tested, each one with different rules for the ants movements and for handling occupied cellsOriginally, the objective was to remove a bias introduced by constraint of having no more than one ant in each cell. CongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201116
- 28. ColorPherographyCongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201117Variation 1: when an ant tries to move to an occupied cell, the ant that occupies that cell is moved in the same direction and its directional vector changes.
- 29. ColorPherographyCongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201118Variation 2: the ants are not allowed to move to occupied cells. If one tries to move to an occupied cell, it will stop, and the direction of the ant occupying that cell changes to that of the vectorial sum between the original direction and the one towards which this ant tried to move.
- 30. ColorPherographyCongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201119Variation 3: ants “bounce off” if they try to move to occupied cells.
- 31. ColorPherographyCongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201120Variation 4: introduces speed and implements ants with variable speed, that is, the speed may change when an ant moves to a cell occupied by another ant.
- 32. CongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201121
- 33. CongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201122
- 34. Conclusions and Future WorkThe monochromatic pherographia has been extended to colorpherographia. Colorpherographia also detects the edges of the image.Investigate other local and simple rules that generate different global behaviour. Emergence, memory and readaptation (videos).Implement pherographia (B/W and color) in a digital camera using LUA programming language. CongressonEvolutionaryComputation, CEC’11, NewOrleans, USA, 201123