Simulation Of Flaw Defect Based On Qualitative Vision Features Semantic Description

Casting defect simulation techniques have received increasing attention in the field of automatic casting inspection. Simulation results are used to validate the system's sensitivity and tune the automatic inspection parameters, so as to improve the performance and guarantee the quality of mechanical products. As one of the most complex defects, real-time and reliable simulation of flaw defect becomes the difficulty of defect simulation, it will greatly develop the automatic defect simulation techniques.On the basis of qualitative vision features, an extraction model of semantic is built to describe the semantic of flaw defect by ontology description logics. Several math descriptors are designed to reduce the semantic gap between bottom-level vision features and high-level semantic. The weighed sum for logic description of descriptors represents the membership of flaw.On the research of ontology description logics, semantic extraction and fuzzy geometric features, a new simulation algorithm including four steps is proposed. Firstly, three methods are introduced to generate the skeleton of flaw defect: extracting from flaw defect X-ray images, drawing in the setting window or generating it randomly. The last results got by the first way are compared with the samples about the fuzzy geometric features to adjust the parameters of simulation algorithm. Secondly, motion filter is applied to each pixel on skeleton. Then, according to the rough flaw width needed, fuzzy skeletons are weighted stacked after migration. Finally, the whole image is filtered considering the integral direction, and then superimposed onto the background image. The membership of flaw obtained by calculating can be used to validate the eligibility of flaw. Experimental results show that the approach can produce a large range of simulated flaw defects with randomness and fuzziness.