Abstract

Sawmills in Thailand demand an automatic approach to correctly detect rubberwood pith. This is a starting point to maximize the yield of slabs. Knowing the pith location at both end faces of a log makes it possible for a log scaler to turn the log to align both piths parallel to a bandsaw. Then, the likely to defect part, the knot, can be removed. The rest of the log is then cut to produce a set of slabs. In this paper, an algorithm is proposed to accelerate estimating the pith location of rubberwood with lower estimation error from the untreated cross section image captured by a normal camera. The algorithm employs the histogram of oriented gradients and a set of relevant histogram bin indices to reduce the number of line segments to be successively used in a computational intensive group of line intersection part of the algorithm. The results confirm that the execution time of the proposed algorithm is 187 times and the estimation correctness is 2.69 times compared to the formerly proposed method for European species on a set of 35 cross section rubberwood images taken from a realistic sawmill environment.