摘要
为研究小模数齿轮参数检测方法,针对小模数齿轮(模数0.2~0.8 mm)机器视觉检测中边缘模糊导致的精度下降问题,提出一种基于高斯核与非极大值抑制结合的亚像素边缘提取方法。通过融合一阶直线边缘和二阶弧形边缘检测技术结合边缘阈值优化,提升边缘定位精度。结果表明,该方法在0.2~0.3 mm模数齿轮检测中误差小于0.0471 mm,0.6~0.8 mm模数齿轮误差小于0.0512 mm,较传统Sobel和Canny算法最大降低81.71%误差,齿数识别准确率达100.00%。该方法具有强抗噪性,能够获取齿轮连续、平滑、完整的轮廓边缘,检测精度高,满足小模数齿轮分度圆,齿顶圆,齿根圆,齿数的检测要求。
To study the method of small module gear parameter detection,to address the problem of reduced accuracy caused by edge blurring in machine vision detection of small-module gears(module O.2—0.8 mm),and then propose a sub-pixel edge extraction method combining Gaussian kernels with non-maximum suppression.By integrating first-order linear edge detection and second-order curved edge detection techniques with optimised edge thresholds,this project improved edge localisation accuracy.Experimental results showed that this method achieved an error of less than O.0471 mm for gears with a module of 0.2—0.3 mm and an error of less than 0.0512 mm for gears with a module of 0.6—0.8 mm,and reduced errors by up to 81.71%compared to traditional Sobel and Canny algorithms,with a gear tooth count recognition accuracy rate of 100.00%.This method exhibited strong noise resistance,enabling the acquisition of continuous,smooth,and complete gear contour edges.It achieved high detection accuracy and met the detection requirements for the pitch circle,tooth tip circle,tooth root circle,and tooth count of small-module gears.
作者
史阳阳
张学军
鄢金山
王彤
马菲璠
SHI Yang-yang;ZHANG Xue-jun;YAN Jin-shan;WANG Tong;MA Fei-fan(College of Mechanical and Electrical Engineering,Xinjiang Agricultural University,Urumqi 830052,China;Xinjiang Key Laboratory of Intelligent Agricultural Equipment,Urumqi 830052,China)
出处
《新疆农业大学学报》
2025年第4期347-354,共8页
Journal of Xinjiang Agricultural University
基金
新疆农业大学实验室和基地建设项目(XNSJ202521)
新疆农业大学大学生创新项目(dxscx2025238)。
关键词
小模数齿轮
亚像素
边缘检测
非极大值抑制
图像处理
fine-pitch gears
sub-pixel
edge detection
non-maximum suppression
image processing
