Procesamiento paralelo multi-núcleo para la detección de defectos en tablas de pino utilizando PDI

A proposal for the implementation of an autonomous defects detection in wooden boards and their implementation with multi-core parallel processing is described in this paper. Also an adaptive histogram enhancement method as a pre-processing stage for automatic histogram threshold segmentation of defects is presented. Specifically, experiences were made with pine samples extracted manually by an expert operator of a wood splitter machine. The results show that it is possible to detect the faults observed by the expert and rightly determine the best place to split the board with the splitter machine. The implementation of the histogram thresholding segmentation is appealing because of its low computational cost compared to traditional pattern recognition techniques. The implementation on a parallel structure in quad core processor with eight threads, achieved a speedup of 2.28 times compared to the sequential processing. In addition, it was found that the inspection of both sides by digital image processing could recover more than 24% of the wood that today is discarded because the operator inspects only one of the faces and makes the cut with a handicap on either side.