http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
( Alvin Muhammad Savero ),( Imam Wahyudi ),( Istie Sekartining Rahayu ),( Andi Detti Yunianti ),( Futoshi Ishiguri ) 한국목재공학회 2020 목재공학 Vol.48 No.5
Muna teakwood, especially from old stands, has been popular as raw material for timber industries in Indonesia for the past ten decades. Due to the scarcity of this wood, superior-grown seedlings of Muna teakwood have been developed and widely planted. Since there is no information on its characteristics, therefore, the aim of this research was to investigate wood characteristics of the 8-year-old superior-grown teak from Muna Island to ensure their proper utilization as raw material for wooden furniture. Wood discs and boards from basal area of three different trees were used as the samples. Macroscopic and microscopic anatomical characteristics were observed following the IAWA's list, while their physicalmechanical properties were measured following British Standard 373-57. Results showed that anatomical characteristics of this wood sample are similar to regular teakwood, but its heartwood portion is higher. Differences among trees are found in regards to wood texture, growth ring width, as well as early and latewood portion. The green moisture content was lower than that of fast-growing teak of a similar age. The wood is more stable than the old teakwood, but its specific gravity is lower. In general, mechanical properties of this wood were higher than those of the regular fast-growing teakwood, but lower than the old one. Based on its specific gravity, this superior Muna teakwood was categorized as a Strength Class of III. The wood is suitable enough for wooden furniture manufacturing.
Anatomical Characteristics of Acacia spp. from Vietnam
( Alvin Muhammad Savero ),( Jong-ho Kim ),( Byantara Darsan Purusatama ),( Denni Prasetia ),( Se-hwi Park ),( Nam-hun Kim ) 한국목재공학회 2022 한국목재공학회 학술발표논문집 Vol.2022 No.2
To investigate and compare the anatomical characteristics of Acacia spp. (Acacia mangium and Acacia hybrid) from Vietnam, the anatomical characteristics were observed by optical microscopy connected to an image analysis system and analyzed according to the IAWA list for hardwood identification. Both wood species had different bark surface characteristics in color and roughness. Heartwood-sapwood color and microscopic features of both species were similar. Both species showed differences in cell dimension such as vessel diameter and vessel number.
양고운 ( Goun Yang ),비안타라다르산푸루사타마 ( Byantara Darsan Purusatama ),김종호 ( Jongho Kim ),알빈무함마드사베로 ( Alvin Muhammad Savero ),데니프라세티아 ( Denni Prasetia ),김남훈 ( Namhun Kim ) 한국목재공학회 2022 한국목재공학회 학술발표논문집 Vol.2022 No.1
전통적으로 대금 제작용 재료로 사용되어온 쌍골죽의 특성을 이해하기 위하여 해부학적 특성을 조사하였다. 공시재료는 국립국악원에서 제공받은 왕대(Phyllostachys bambusoides) 수종인 쌍골죽 및 비교를 위해 쌍골죽과 골의 유무로 분류하는 민죽을 활용하였으며, 공시재료의 간(幹)과 뿌리의 내측부와 외측부로 나누어 비교하였다. 해부학적 특성을 광학현미경(Nikon ECLIPSE, E600)으로 관찰 후 분석프로그램(IMT I-Solution lite)을 이용하여 정량 분석을 수행하였다. 또한 X선 회절법에 의해 셀룰로오스 결정특성을 분석하였다. 수간부의 섬유길이는 쌍골죽 내측부 1,987㎛, 외측부 2,220㎛, 민죽 내측부 2,323㎛, 외측부 2,253㎛로 민죽이 더 긴 것으로 나타난 반면, 뿌리부의 섬유길이는 쌍골죽 내측부 1,190㎛, 외측부 1,246㎛, 민죽 내측부 894.1㎛, 외측부 958㎛로 쌍골죽이 더 길었다. 유관속 형태는 Grosser와 Liese(1971)의 방법에 따라 두 수종 모두 유관속 Ⅰ형으로 분류되었으며, 유관속의 배열은 민죽이 쌍골죽보다 더 규칙적이었다. 방사단면 및 접선단면 관찰 시 내측부보다 외측부에서 유관속초 영역이 더 넓었다. 정량적 분석 결과, 4mm<sup>2</sup>당 유관속의 수는 쌍골죽 8.6개, 민죽 16.3개로 민죽이 더 조밀하게 분포하였다. 도관 직경은 쌍골죽 123.6㎛, 민죽 73.8㎛로 쌍골죽이 더 큰 것으로 나타난 반면, 방사단면에서 유세포 높이는 쌍골죽 100.0㎛, 민죽 115.9㎛, 접선단면에서 쌍골죽 89.0㎛, 민죽 95.4㎛로 쌍골죽보다 민죽이 더 높았다. 또한 방사단면에서 유세포의 직경은 쌍골죽 47.2㎛, 민죽 41.1㎛, 접선단면에서 쌍골죽 47.1㎛, 민죽 38.2㎛로 쌍골죽이 민죽 대비 더 높았다. 수간부의 상대결정화도는 쌍골죽 내측부 62.2%, 외측부 72.5%, 민죽 내측부 49.1%, 외측부 58.2%로 쌍골죽이 민죽보다 높았으며, 뿌리부는 쌍골죽 내측부 50.1%, 외측부 51.8%, 민죽 내측부 49.1%, 외측부 58.2%로 두 수종 모두 내측부에 비해 외측부가 더 높은 것으로 확인되었다. 본 연구의 결과는 향후 쌍골죽의 재질판단 지표로 활용될 수 있을 것으로 판단되었다.
김종호 ( Jong-ho Kim ),비안타라다르산푸루사타마 ( Byantara Darsan Purusatama ),데니프라세티아 ( Denni Prasetia ),알빈무함마드사베로 ( Alvin Muhammad Savero ),김남훈 ( Nam-hun Kim ) 한국목재공학회 2022 한국목재공학회 학술발표논문집 Vol.2022 No.2
본 연구에서는 인공신경망을 활용하여 수종간 분류 가능성을 검증하고, 수종의 분류 과정에서 작용하는 영향 인자를 분석하기 위해 기존 합성곱신경망(CNN) 모델을 개선한 모델을 이용, 수종분류 정확도에 영향을 미치는 인자를 분석하였다. 본 연구에서 검증한 인공신경망 모델은 최종적으로 95% 이상의 높은 분류정확도를 나타내어 인공신경망을 활용하는 딥러닝 기술을 통해 목재의 수종 분류, 나아가 수종 식별을 매우 높은 정확도로 수행 가능한 것으로 검증되었다. 학습횟수의 증가에 따라 데이터의 손실률은 감소하고 분류 정확도는 증가하는 경향이 나타났으며, 전체부위 및 만재부에서 확보한 데이터셋을 활용할 경우 손실률과 분류 정확도에 영향이 있었으나, 전체부위 데이터셋은 성능 하락이, 만재부 데이터셋은 성능이 향상되는 경향이 나타났으며, 데이터셋의 증폭여부는 손실률 및 분류 정확도에 영향은 있었으나 그 영향의 정도가 크지 않았다.
( Intan Fajar Suri ),( Byantara Darsan Purusatama ),( Jong Ho Kim ),( Go Un Yang ),( Denny Prasetia ),( Muhammad Alvin Savero ),( Se Yeong Park ),( Nam Hun Kim ) 한국목재공학회 2022 한국목재공학회 학술발표논문집 Vol.2022 No.1
This study aimed to evaluate the weathering properties of Paulownia tomentosa and Pinus koraiensis woods heat-treated in palm oil (OHT) and air (AHT) at 180℃, 200℃, and 220℃ for 2 hours. The untreated and heat-treated samples were exposed to UV and water for artificial weathering test according to ASTM G53-96. The artificial weathering test of heat-treated and untreated wood samples was performed by exposing to UV lamps in the QUV accelerated weathering tester (QUV/se Accelerated Weathering Tester, Q-LAB, USA) for 168 h and 336 h. The weathering cycle involved a continuous light irradiation of UV exposure for 2 hours and condensation for 2 hours. Color change and dimensional stability of the weathered samples were determined. Color change was measured by the CIEL*a*b* system (Esteves et al. 2008). Macroscopically, there was hardly shown on color difference in the wood samples before and after weathering. After weathering test, heat-treated wood of both species showed lower total color change than the control samples, and the total color change decreased with increasing treatment temperature. The total color change of OHT wood was smaller than that of AHT wood samples. In both species, the heat-treated wood samples showed higher dimensional stability than the control samples. The volumetric shrinkage of heat-treated wood samples decreased as the temperature increased. The OHT wood samples showed smallest volumetric shrinkage than AHT wood samples. In conclusion, after the weathering test, the OHT wood samples displayed better color and dimensional stability than the AHT wood samples.
합성곱신경망 모델에 따른 침엽수재 수종식별 성능 비교 및 최적 모델 개발
김종호 ( Jongho Kim ),비안타라다르산푸루사타마 ( Byantara Darsan Purusatama ),양고운 ( Goun Yang ),프라세티아데니 ( Prasetia Denni ),알빈무함마드사베로 ( Alvin Muhammad Savero ),김남훈 ( Namhun Kim ) 한국목재공학회 2022 한국목재공학회 학술발표논문집 Vol.2022 No.1
The four convolutional neural network(CNN) architectures such as GoogLeNet, ResNet50, VGG16, and modified CNN were analyzed for investigating the effect of environmental variables on the accuracy of species identification like focused position, epochs, data augmentation and optimizer. Totally 5,535 cross-section images including 1,535 images of low-magnification(40X), 2,000 images of earlywood focused image(200X), and 2,000 images of latewood focused image(200X) were prepared for the dataset establishment. After the preparation, each dataset was randomly separated into 80% of the training group and 20% of the verification group. Data augmentation was applied only in training group for verifying the effectiveness of the dataset amount. As a result of training and verification process, the GoogLeNet architecture increased classification accuracy in proportion to the number of training epochs, and its classification accuracy achieved 99.0% at training process and 98.1% at verification process when applied non-augmented latewood dataset. In augmented latewood dataset, classification accuracy achieved 91.1% and 91.6% at training and verification process, respectively. In contrast, the best classification accuracy of ResNet50 architecture was 87.7% at training process and 71.3% at verification process. VGG16 architecture showed poor performance with around 10% accuracy at both training and verification processes under all conditions. The modified CNN architecture showed excellent classification accuracy with 95.9~99.8% at training process and 95.1~96.9% at verification process when using the earlywood dataset with 100 epochs condition. Moreover, the latewood dataset with 100 epochs condition also makes remarkable results as 96.2~99.2% at training process and 96.5~96.7% at verification process. Based on the results, data augmentation was not significantly affected to classification accuracy of CNN based softwood identification system in this research. In contrast, classification accuracy showed the increased tendency with an increment of training epochs and adoption of the latewood dataset.
딥러닝기반 수종식별체계 데이터셋확보를 위한 상용 목재제품의 수종식별 (II)
김종호 ( Jongho Kim ),비안타라다르산푸루사타마 ( Byantara Darsan Purusatama ),양고운 ( Goun Yang ),프라세티아데니 ( Prasetia Denni ),알빈무함마드사베로 ( Alvin Muhammad Savero ),김남훈 ( Namhun Kim ) 한국목재공학회 2022 한국목재공학회 학술발표논문집 Vol.2022 No.1
For the purpose of securing a dataset of deep learning-based softwood species identification systems, wood species were analyzed using anatomical features to verify more reliable species in the process of purchasing and utilizing commercial wood products with uncertain wood species. The three samples, such as Red pine laminated board from Finland, aromatic cube block of domestic Cypress, and Cedar board of unknown origin were used as identification resources. The general features and anatomical features of the three species were analyzed. The features were classified into standardized code according to the IAWA list of microscopic features for softwood identification and compared with the InsideWood database managed by North Carolina State University to infer the features matched species. Red pine laminate board was classified as Mugo pine(Pinus mugo) due to observation of IAWA feature code 2, 40, 43, 44, 79, 82, 85, 90, 97, 103, 107, 109, and 110. Cypress aromatic block had IAWA feature code 4, 40, 43, 44 72, 74, 76, 80, 93, 98, 103, and 107, and these features imply Cypress or Hinoki(Chamaecyparis obtusa), Taiwan cypress(Chamaecyparis formosensis), Sawara cypress(Chamaecyparis pisifera) and Buthan cypress(Cupressus duclouxiana). However, it is most likely to Cypress due to its commercial name matches. Cedar solid board was identified as Cedar(Cryptomeria japonica), Taiwan coffin fir(Cunninghamia konishii), and Chinese swamp cypress(Glyptostrobus pensilis) because of the existence IAWA feature codes 40, 42, 44, 72, 73, 76, 80, 85, 94, 99, 103, and 107. Nevertheless, the probability of cedar was the highest due to the consistency with the commercial name.