MorphoHawk: True Machine Vision
Trainableobject identification and analysis Correctlyidentify and categorize objects in different environmental conditionsIdentifynoise or artifacts and clean up images and signalsTrackchanges over timeSeparatevarious changes to help identify root causes of problems
Wirelesscommunications (e.g. message extraction from noisy signal)Medical(e.g. comparison of medical images to detect changes in a body like a new tumoror extracting an x-ray or ultrasound from a noisier signal, reducing time for aclear image to be taken)Manufacturing(e.g. automated defect detection, quality control on processes, etc)Imageprocessing (e.g. sharpening filters)Tracking(bar code reading)Surveillance(e.g. automatically detecting what type of vehicle enters a scene)AstronomyCounterfeitdetectionObjectclassification for autonomous vehicles and aircraftAnyapplication requiring a machine to identify and classify information to bevisualized
Researchers at INL havedeveloped an analysis system known as MorphoHawk for automatic featuredetection and classification across a host of applications in changingenvironmental conditions. In general, the system can be trained on what to lookfor, and will then group objects in the scene (e.g. image, signal, etc),analyze those groups, and categorize them according to the rules it wasconditioned with. After it has categorized an image, it can compare the objectsit has identified with subsequent images, allowing it to detect changes insimilar objects (e.g. manufacturing quality control) or detect the introductionof new objects (e.g. a tumor in medical scans or a person entering a scenemonitored by a camera). It has shown that it can discern between an object andits shadow, meaning it can handle differences in lighting on a scene.MorphoHawk has shownutility as a signal filtering tool to differentiate between noise andmeaningful data in analysis of digital images and electronic signals, resultingin sharp, cleaned images and clearly extracting the message of the signalswhile removing the noise. MorphoHawk can be applied to analyze images formanufacturing defects, enhancing the capability of existing inspection systems.Feature extraction is another unique capability of MorphoHawk. For example, metal surface topology can beseparated into effects of rolling and grinding, allowing discrepancies to beassigned to the appropriate process. It has even been used to identify afacture path in materials and examine structural changes in battery electrodesto predict battery lifetime.
LICENSINGOPPORTUNITY: IdahoNational Laboratory (INL) and its M&O Contractor Battelle Energy Alliance,LLC (BEA) are currently looking for commercialization partner(s) interested inentering into a license agreement for the purpose of commercializing thetechnology described below. INTELLECTUALPROPERTY STATUS: This invention has associated intellectual property US Patent No. 9,342,876,BEA Docket No. BA-481: “Methods, apparatuses, and Computer-Readable Media forProjectional Morphological Analysis of N-Dimensional Signals,” issued 17 May2016. DEVELOPMENTSTATUS: This technology has been tested and validated at thebench scale. Additional development will be required to demonstrate a pilotscale process. ADDITIONAL INFORMATION: Summaryvideo: https://www.youtube.com/watch?v=-JCoEKEZ5W0 , also found on OSTI Publications:
美国
