AmeriranikistanProject OverviewSynopsisSystem Block DiagramHardwareSelf Organizing Map (SOM)SoftwareMatlab DemoSystem SetupDivision of LaborTimelineCost EstimateRisk analysisSafeguards against risksPossible ImprovementsApplicationsReferences¿Questions?AmeriranikistanAmeriranikistanMuhammad AhmadKyle HustonFarhad MajdeteimouriDan MackinProject OverviewProject Overview•Uses Self Organizing Maps to Model Fiber Optic Cables in Matlab.•Speckle Pattern is recorded in your typical webcam and then passed into Matlab.•We will use the Image Acquisition Toolbox, in Matlab, to analyze the speckle pattern.•The speckle pattern will change as the user taps on the fiber strand in various locations.•The Self-Organizing Map algorithm will learn the topology of the fiber.SynopsisSynopsisLaser light is shone through an optical fiberCamera (CCD) receives the light on the other endUser taps random points on the fiberAlgorithm gradually learns to make a map of the fiberDisplay shows the location of all touches going forwardSystem Block DiagramSystem Block DiagramLaser Multi-Mode Fiber CameraAlgorithmDisplayHardwareSoftwareUSBHardwareHardwareOptical BreadboardRed Laser DiodeFocusing and Collimating LensesMirrorsMulti-Mode FiberCamera (USB Web Cam or CCD Sensor)Personal ComputerSelf Organizing Map (SOM)Self Organizing Map (SOM)Unsupervised Artificial Neural Network (ANN) AlgorithmPattern RecognitionCluster the Distribution of the Input SpaceReduce the dimensionality and visualize high dimensional data preserving the most significant featuresExtract the underlying topology in the input (signal) spaceSoftwareSoftwareMatlab is our choice of software:–SOM is based on linear algebra–Efficient matrix and vector computations–Easy creation of scientific and engineering graphics–Extensibility (Tool Boxes)–I/O functionsMatlab DemoMatlab DemoSystem SetupSystem SetupDivision of LaborDivision of LaborHardware/Calibration – Kyle, DanImage Processing – Muhammad, FarhadSOM Algorithm – Farhad, KyleDisplay Routines – Dan, MuhammadTimelineTimelineID Task Name Duration Start Finish1 PDR 9 day s Fri 8/31/07 Tue 9/11/072 CDR 23 day s Tue 9/11/07 Thu 10/11/073 Milestone 1 16 day s Thu 10/11/07 Thu 11/1/074 Milestone 2 21 day s Thu 11/1/07 Thu 11/29/075 Expo 10 day s Sat 12/1/07 Thu 12/13/076 Hardware Setup 6 day s Mon 9/10/07 Mon 9/17/077 Image Acquisition 6 day s Mon 9/17/07 Mon 9/24/078 Front End Image Processing 9 day s Wed 9/19/07 Mon 10/1/079 SOM Algorithm Dev elopment 34 day s Mon 9/17/07 Thu 11/1/0710 Display 30 day s Mon 10/1/07 Fri 11/9/0711 Integration 40 day s Thu 10/11/07 Tue 12/4/0712 Testing 54 day s Mon 10/1/07 Wed 12/12/0713 Users Manual 69 day s Sat 9/8/07 Tue 12/11/0714 Technical Manual 70 day s Sat 9/8/07 Wed 12/12/07PDRCDRMilestone 1Milestone 2ExpoHardwarImage AFront End ImaSOM Algorithm DevelopmentDisplayIntegrationTestingUsers ManualTechnical Manual8/19 8/26 9/2 9/9 9/16 9/23 9/30 10/7 10/14 10/21 10/28 11/4 11/11 11/18 11/25 12/2 12/9 12/16September October Nov ember DecemberCost EstimateCost EstimateOptical Breadboard $700Red Laser Diode $20Focusing and collimating lenses $20Multi-Mode Fiber $30Camera (USB Web Cam) $80Personal Computer N/ATotal $850Risk analysisRisk analysisThe experimental nature of SOM could lead to delays in the project timelineAny learning curve for the Matlab Image Acquisition ToolboxRunning the algorithm using a scripting language (Matlab) on a best effort OS (Windows) could prove costlyHardware issues (e.g. Fiber breakage and calibration issues)Safeguards against risksSafeguards against risksEarly Integration and TestingPrototyping the SOM AlgorithmPossible Improvements Possible Improvements Automated fiber tapping mechanism for training the SOM algorithmOptimizing the SOMApplicationsApplicationsUsing multiple fibers to model the behavior of flexible structures in spaceAnimation development without any sensor dependenciesReferencesReferencesSelf-Organizing Maps by T. KohonenProfessor Dana Anderson in the physics
View Full Document