基于自适应阈值的表面缺陷纸张图像识别技术  

Image Recognition Technology of Paper with Surface Defects Based on Adaptive Threshold

作　　者：杨品[1,2] 尹超 YANG Pin;YIN Chao(Department of Mechanical and Electrical Information,Anhui Vocational and Technical College of Press and Publishing,Hefei 230601,China;College of Light Industry and Food Science,Nanjing Forestry University,Nanjing 210037,China)

机构地区：[1]安徽新闻出版职业技术学院机电信息系,安徽合肥230601 [2]南京林业大学轻工与食品学院,江苏南京210037

出　　处：《造纸科学与技术》2025年第4期89-93,共5页Paper Science And Technology

基　　金：安徽省高等学校自然科学研究重点项目(2024AH050629)。

摘　　要：纸张质量控制是纸张生产过程中的关键环节之一,其不仅决定着纸张质量优劣,还会影响后续的加工、保障等环节。已有技术由于表面缺陷分割不合理,致使表面缺陷识别性能较差,故提出基于自适应阈值的表面缺陷纸张图像识别技术研究。综合考虑纸张图像的整体灰度分布和局部特征,计算自适应阈值,并制定纸张图像分割规则,获取表面缺陷纸张图像。从几何形态、纹理结构、统计分布等方面提取表面缺陷特征——缺陷轮廓复杂度指数、曲率分布熵、多尺度Gabor能量谱、小波包能量熵与空间自相关系数。通过神经网络模型输入层、隐藏层与输出层的协同操作,实现纸张图像表面缺陷的分类识别。实验结果显示,在单一表面缺陷与多个表面缺陷背景下,设计技术能够精准识别各类纸张图像表面缺陷,并且表面缺陷误识率降低至2%。Paper quality control is one of the key links in the paper production process,which not only determines the quality of paper,but also affects the subsequent processing and guarantee.In the prior art,due to unreasonable surface defect segmentation,the performance of surface defect recognition is poor,so the paper image recognition technology based on adaptive threshold is proposed.Considering the overall gray distribution and local characteristics of paper image,the adaptive threshold is calculated,and the paper image segmentation rules are formulated to obtain the paper image with surface defects.The features of surface defects,such as defect contour complexity index,curvature distribution entropy,multi-scale Gabor energy spectrum,wavelet packet energy entropy and spatial autocorrelation coefficient,are extracted from geometric morphology,texture structure and statistical distribution.Through the cooperative operation of input layer,hidden layer and output layer of neural network model,the classification and recognition of paper image surface defects are realized.The experimental results show that under the background of single surface defect and multiple surface defects,the design technology can accurately identify the surface defects of various paper images,and the false recognition rate of surface defects is reduced to 2%.

关 键 词：表面缺陷 识别性能测试与分析 纸张图像 自适应阈值 图像预处理 

分 类 号：TS73[轻工技术与工程—制浆造纸工程]