Detection of Wood Features Extraction Region using Convolutional Neural Network
Detection of Wood Features Extraction Region using Convolutional Neural Network
复制标题
利用卷积神经网络检测木材特征提取区域
DOI：
10.1145/3468081.3471136
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Weiwei Du
中科院分区：
文献类型：
--
作者：
Yumei Zhang;Keiko Nagashima;Weiwei Du
木材被应用于不同的行业，如建筑房屋，桥梁，并根据其强度。根据研究[12]，木材的强度可以通过木材的特征来推断。虽然[3]可以提取木材特征，但[3]未能准确地提取木材特征的一些木材图像，如室内木材图像的未经处理的高噪声原木。如果能将未经处理的高噪声原木的室内木材图像的区域去除，使相同的木材特征出现在低噪声区域的地方，则可以更有效地提取木材特征。本文提出了一种简单的卷积神经网络模型来检测室内木材图像中未处理的低噪声原木区域，以提取木材特征。实验结果表明，在未经处理的高噪声原木的室内木材图像上可以更有效地提取木材特征。
Wood is applied with different industries such as building houses, bridges and depending on their strength. According to the research [12], the strength of wood can be inferred by wood features. Although [3] can extract wood features, [3] fails to extract wood features accurately on some wood images such as the indoor wood images of the unprocessed high-noise logs. If the region of the indoor wood images of the unprocessed high-noise logs can be removed as the same wood features appear in the places where are low-noise region, wood features can be extracted more effectively. This paper proposes a simple Convolutional Neural Network model to detect the region of the unprocessed low-noise logs in the indoor wood images for wood features extraction. Experimental results show that wood features can more effectively be extracted on the indoor wood images of the unprocessed high-noise logs.