摘要
木窗是一种以木材或木质复合材料为主要构件的门窗产品,具有良好的生态性能和美观效果,适用于多种建筑形式和风格,其中木窗尺寸是衡量木窗加工是否合格的重要指标。对于传统木窗双端铣削加工中人工测量尺寸方式存在的精度低、效率低等问题,提出一种基于机器视觉的木窗双端铣削加工尺寸测量方法,以期提高尺寸测量精度及加工效率。该方法针对木窗厚度引起的透视效应,提出一种物平面提升法,以消除透视投影带来的误差。首先对木窗图像采取灰度化、平滑去噪、图像增强及轮廓分割等举措,完成图像预处理,提取出木窗内外轮廓区域。对轮廓区域应用Canny算子获取木窗像素级轮廓。通过优化的Zernike矩亚像素边缘提取算法对木窗像素级边缘进行更精确的定位,得到亚像素级轮廓坐标。通过最小二乘法联合RANSAC算法对亚像素轮廓坐标进行拟合,得到拟合轮廓及角点坐标,并使用透视矫正模型计算出木窗尺寸。实验利用3种厚度规格相同但尺寸不同的松木材质矩形木窗,分别测量其内框和外框的边框尺寸及对角线尺寸,并与对应的实际物理尺寸对比,验证了所提木窗尺寸测量方法的检测精度。研究结果表明,所提方法与实际物理尺寸值相比,其绝对误差范围在±0.12 mm之内,相对误差在±0.1%之内,且效率及精度高,可以满足对木窗的在线尺寸检测。
Wooden window is usually made from a wood or wood-based composite material as the main components.The precise dimension of a wooden window is a critical factor that determines the quality and functionality of the window.The window components should be accurately measured to ensure a proper fit within the window opening.This is crucial for functionality,preventing issues such as drafts,leaks,or difficulty in opening and closing.To address the limitations associated with traditional wood window double-end milling processing manual size measurements,such as the low precision and low efficiency,the machine vision-based wood window double-end milling processing offers a technological solution to improve the size measurement accuracy and processing efficiency.This study proposed an object plane lifting method for the perspective effect caused by the thickness of wooden windows,in order to eliminate the error caused by perspective projection.Firstly,the image of the wooden window was grayed out,smoothed and denoised,enhanced and segmented to complete the image preprocessing,and the inner and outer contour regions of the wooden window were extracted.The Canny operator was applied to the contour region to obtain the pixel-level contour of the wooden window.The optimized Zernike moment sub-pixel edge extraction algorithm was used to locate the pixel-level edges of the wood window more accurately and obtain the sub-pixel-level contour coordinates.The subpixel contour coordinates were fitted by the least squares method combined with the RANSAC algorithm to obtain the fitted contour and corner coordinates,and the wood window dimensions were calculated using the perspective correction model.The experiments utilized three types of rectangular wooden windows made of pine with the same thickness specifications but different sizes to measure the border dimensions and diagonal dimensions of their inner frames and outer frames respectively and compare them with the corresponding actual physical dimensions to validate the detection a
作者
任长清
张佳林
杨春梅
宋文龙
吴哲
REN Changqing;ZHANG Jialin;YANG Chunmei;SONG Wenlong;WU Zhe(College of Mechatronics Engineering,Northeast Forestry University,Harbin 150040,China)
出处
《林业工程学报》
CSCD
北大核心
2024年第1期141-149,共9页
Journal of Forestry Engineering
基金
黑龙江省重点研发项目(GA21A405)
中央高校基本科研业务费专项资金(2572020DR12)。
关键词
机器视觉
木窗
尺寸测量
边缘检测
亚像素
machine vision
wood window
size measurement
edge detection
sub-pixel
