One of the time consuming tasks in the timber industryis the manually measurement of features of wood stacks.Such features include, but are not limited to, the numberof the logs in a stack, their diameters distribution, and theirvolumes. Computer vision techniques have recently beenused for solving this real-world industrial application. Suchtechniques are facing many challenges as the task is usuallyperformed in outdoor, uncontrolled, environments. Furthermore,the logs can vary in texture and they can be occludedby different obstacles. These all make the segmentation ofthe wood logs a difficult task. Graph-cut has shown to begood enough for such a segmentation. However, it is hardto find proper graph weights. This is exactly the contributionof this paper to propose a method for setting theweights of the graph. To do so, we use Circular HoughTransform (CHT) for obtaining information about the foreandbackground regions of a stack image, and then use thistogether with a Local Circularity Measure (LCM) to modifythe weights of the graph to segment the wood logs from therest of the image. We further improve the segmentation byseparating overlapping logs. These segmented wood logsare finally scaled and used to acquire the necessary woodstack measurements in real-world scale (in cm). The proposedsystem, which works automatically, has been testedon two different datasets, containing real outdoor imagesof logs which vary in shapes and sizes. The experimentalresults show that the proposed approach not only achievesthe same results as the state-of-the-art systems, it producesmore stable results.
展开▼
