FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W154584071> ?p ?o ?g. }

• W154584071 abstract "The thesis describes efficient and robust algorithms and representations for extracting high level geometry, random variation, and local deformation from images. The extraction is done by fitting parameterized templates at different levels of abstractions.High level geometry, such as polyhedral and curved shapes subject to invariant constraints, is extracted by fitting to image data, a template of parameterized local shapes constrained by polynomial equations. Topology, rough configuration, a priori constraints, and model to image correspondences are assumed given. The application is a modeling system for fast extraction of polyhedral shapes from single and multiple images, without accurate placement and fit-by-eye.The main contribution of this modeling system is the over-constraint method which works well for constrained least square error problems that have sparse Jacobian and Hessian matrices. This method converges reliably to the global constrained minimum by first going to the best-fit surface, then moving onto the constraint surface by weighting in the constraint equations. It thus avoids local singularities and minima on the constraint surface, that are typical of the Lagrange method. It is also better conditioned than the penalty method, and so has a faster linear convergence rate.Random variation is extracted with a template of normal distributions learned at each pixel in the image. Once learned, these distributions are used to detect abnormalities. This scheme is applied to the automatic inspection of X-ray images of turbine blades. The geometry of the blades is assumed invariant, up to a global linear transformation and local deformations within known small tolerances. Defects are assumed infrequent, but have arbitrary location, size, shape, and strength.The main contribution of this inspection system is a flaw detection algorithm based on detecting abnormality. Critical to this algorithm is a very accurate registration of the test to the model surfaces, so that random variations can be extracted as differences at corresponding pixels. Also critical is an estimation of the normal distribution based on the median and mean absolute deviation, which have lower order moments than the mean and standard deviation, and so are less sensitive to outliers.Local deformation from ideal geometry is extracted with a template of displacement vectors at strong local features. These displacement vectors are bounded, smoothed, and compensated to account for a priori manufacturing tolerances. The application is a screening of X-ray images of turbine blades that out-performs human inspectors and requires no human visual interpretation." @default.
• W154584071 created "2016-06-24" @default.
• W154584071 creator A5065436865 @default.
• W154584071 date "1995-11-20" @default.
• W154584071 modified "2023-09-26" @default.
• W154584071 title "Extract models from images with parameterized templates" @default.
• W154584071 hasPublicationYear "1995" @default.
• W154584071 type Work @default.
• W154584071 sameAs 154584071 @default.
• W154584071 citedByCount "0" @default.
• W154584071 crossrefType "journal-article" @default.
• W154584071 hasAuthorship W154584071A5065436865 @default.
• W154584071 hasConcept C11413529 @default.
• W154584071 hasConcept C165464430 @default.
• W154584071 hasConcept C200331156 @default.
• W154584071 hasConcept C28826006 @default.
• W154584071 hasConcept C33923547 @default.
• W154584071 hasConceptScore W154584071C11413529 @default.
• W154584071 hasConceptScore W154584071C165464430 @default.
• W154584071 hasConceptScore W154584071C200331156 @default.
• W154584071 hasConceptScore W154584071C28826006 @default.
• W154584071 hasConceptScore W154584071C33923547 @default.
• W154584071 hasLocation W1545840711 @default.
• W154584071 hasOpenAccess W154584071 @default.
• W154584071 hasPrimaryLocation W1545840711 @default.
• W154584071 hasRelatedWork W1494199114 @default.
• W154584071 hasRelatedWork W18433737 @default.
• W154584071 hasRelatedWork W1975089360 @default.
• W154584071 hasRelatedWork W2000018820 @default.
• W154584071 hasRelatedWork W2015753863 @default.
• W154584071 hasRelatedWork W2025062188 @default.
• W154584071 hasRelatedWork W2057926617 @default.
• W154584071 hasRelatedWork W2080259523 @default.
• W154584071 hasRelatedWork W2102913354 @default.
• W154584071 hasRelatedWork W2119789474 @default.
• W154584071 hasRelatedWork W2122926404 @default.
• W154584071 hasRelatedWork W2130472498 @default.
• W154584071 hasRelatedWork W2150149968 @default.
• W154584071 hasRelatedWork W2168399949 @default.
• W154584071 hasRelatedWork W2252704262 @default.
• W154584071 hasRelatedWork W2254602467 @default.
• W154584071 hasRelatedWork W2495532800 @default.
• W154584071 hasRelatedWork W3022130175 @default.
• W154584071 hasRelatedWork W3164877266 @default.
• W154584071 hasRelatedWork W4231444 @default.
• W154584071 isParatext "false" @default.
• W154584071 isRetracted "false" @default.
• W154584071 magId "154584071" @default.
• W154584071 workType "article" @default.