Practicing Vision Integration, Evaluation and Applications

Model Database Size100000100001000100 10Single object unoccludedMultiple objects unoccludedMultiple objects occludedScene ComplexityObject ComplexityFlat(2D)Polyhedral Quadric SuperquadricFree-Form(sculpted)Type of processPhysical frames(discrete parts)Logical frames(continuous flow/web processing)StationaryBottles (500/min)Clutch plates (100/min)MovingCastings (2/min)Solder (10/min)Color sorting of peas(millions/min; 600 ft/min)Glass coatings (1 ft/min)Continuous foil (10 ft/min)Aluminum foil (2,000 ft/min)Digital Image EnhancementSegmentationRegion DescriptionRelationship Between RegionsObject Identification and LocationProblem requirements analysis(identify specific goals and requirements of the system)Problem mapping to image data(identify image cues useful to achieve the goals)Image feature extraction(develop algorithms to extract image features)Generation of hypotheses(combine domain knowledge with image features to formhypotheses)Preparation of hypotheses to meet problem requirements(evaluate and recompute hypotheses to address the goals) Implementation tuning to meet problem specifications(rework implementation if necessary, to addressperformance issues)10.950.9Recognition rate0.850.80.750.70.650.60.550.524681012Feature subset size141618