Summary: | Agarwood is a fragrant dark resinous wood formed when Aquilaria tress infected with a certain type of mold and appears like wood defects. It is the most valuable non-timber product has been traded in international markets because of its distinctive aroma, and can be processed into incense and perfumes. Agarwood grade is determined by several characteristics, such as black colour intensity, smell, texture and weight through visual inspection. However, this could lead to several problems such as false grading results. Traditionally, the carving process of separation of the uninfected Aquilaria wood that lacks of the dark resinous accomplished by using simple tools like knife and chisel. Hence, an expert worker is required to complete the task. In this paper, the Artificial Neural Network (ANN) technique is used to classify the Agarwood based on the features extraction from Gabor Filter and percentage of black colour estimation. At first, the images of seven groups of wood defects or knots are identified: dry, decayed, edge, encased, horn, leaf, and sound defect with total sample of 410 knots. Then, these images of knots are matched into three grade groups of Agarwood. Next, the experimental results show that the Agarwood can be classified into three grades groups based on knot and black intensity. A set of selected images of knots were used as trace patterns and carved on pieces of wood blocks by using a Computer Numerical Control (CNC) machine where the total time taken for each carving process was calculated. For each image, two Gabor Filter features and percentage of black colour were used as ANN inputs. In conclusion, the total accuracy of the experiments is 98% and the total time of carving is increased with the increased of grade group number.
|