摘要
单板的质量决定单板类人造板的使用价值,单板上的缺陷处理成为木材加工中的重要环节。为处理单板的缺陷,提高木材的利用率,提出一种基于匹配语义感知的单板缺陷图像修复方法。首先使用匹配语义感知模块获取远距离的特征,提升模型的精度;然后使用双卷积模块,捕获多尺度上下文信息,并在整个网络中使用区域归一化,避免均值和方差偏移。使用峰值信噪比(Peak signal-to-noise ratio,PSNR)和结构相似性(structural similarity index,SSIM)为评价指标。研究结果表明,改进后方法的PSNR达到28.48,SSIM达到0.91,与全局和局部判别器网络(Globally and Locally Consistent Image Completion,GL)相比,PSNR和SSIM分别提升1.03%和0.05%。研究结果表明该方法可取得结构、纹理一致的修复效果,为单板缺陷修复提供指导性意见。
The quality of veneer determines the grade of veneer wood-based panels and the treatment of defects on veneer becomes an important part of wood processing.In order to deal with veneer defects and improve wood utilization,an image inpainting method of venerr defect based on match attention is proposed.The method proposes a match attention module to acquire features at a distance to enhance the accuracy of the model and uses a double convolution module that captures multi-scale contextual information.Then region normalization is used throughout the network to avoid mean and variance bias.Peak signal-to-noise ratio(PSNR) and structural similarity index(SSIM) are used as evaluation indicators.The results show that the PSNR of the improved method reaches 28.48,and the SSIM reaches 0.91.Compared to the globally and locally consistent image completion(GL) method,the PSNR and SSIM are improved by 1.03% and 0.05%,respectively.This method can achieve consistent effect in structure and texture,which providing guidance for the inpainting of the veneer defects.
作者
葛奕麟
孙丽萍
王頔
GE Yilin;SUN Liping;WANG Di(College of Computer and Control Engineering,Northeast Forestry University,Harbin 150040,China;College of Petroleum Engineering,Harbin Institute of Petroleum,Harbin 150028,China)
出处
《森林工程》
北大核心
2024年第1期98-105,共8页
Forest Engineering
基金
中央高校基础研究基金(2572019BF08)。
