A dual-threshold image sweep-and-mark algorithm for defect detection in veneer

An improved statistical Image Sweep-and-Mark (ISM) algorithm that uses neighborhood relationships among image tiles is investigated to determine its potential for use in automated systems for identifying defects in Douglas-fir (Pseudotsuga menziesii) veneer. Many conventional ISM algorithms mark tiles by computing an index that quantifies how different the feature measurements for a tile are from those of other tiles. A tile is marked if its index is less than a given threshold. Using such a rule, each tile is classified without reference to its neighbors. By employing an additional (secondary) threshold for tiles whose neighbors have been marked using the primary threshold, the new ISM algorithm is able to achieve better results than conventional statistical ISM algorithms, especially when operating on images containing pitch pockets/streaks. The new algorithm reduces the overall error rate in the identification of clear wood by 17.3 percent, and reduces the overall error rate in the identification of defects by 66.9 percent when compared to conventional algorithm results.

You must be logged in to download any documents. Please login (login accounts are free) or learn how to Become a Member