| 摘要 |
<p>The invention relates to an industrial intelligence-based method and installation for evaluating the quality of the industrial products, directly from the production line, with different levels of rejecting the non-conforming products. According to the invention, the method is based on image processing and consists of a first stage of taking over an image resulting from the X ray control of a checked product, a second stage of low-level image processing, which comprises the radiographic image processing, the segmentation thereof by using a Hopfield neural network and the extraction of objects resulting from segmentation, the third stage of high level faults detection, comprising the characteristics extraction and the detection proper based on fuzzy logics and the fourth stage of admission/rejection decision with three possible answers for the analyzed product, namely: "conforming product", "recoverable waste" or "non-recoverable waste". According to the invention, the installation comprises an automatic device () for handling, moving and rotating for automatic move of the examined product, an equipment () for examination with penetrating radiations, a system () for automatically processing the radiographic image and the transmission by a computing unit (), a main belt carrier (), two belt carriers () for handling the recoverable or non-recoverable waste type products, two limiters () for stopping the mentioned products and two pneumatic pushers () for directing the same, the computing unit () comprising, in its turn, a module () for processing the image, a module () for extracting the characteristic, a high level detection module () and a module () for actuating a mechanized device for handling, moving and rotating.</p> |
