Precise Z-Block positioning and dimension measurement using improved Canny edge detection and sub-pixel contour fitting
Jie Xiong,Dongsheng Wang,Jian Yin,Runfang Wu
DOI: https://doi.org/10.1007/s11227-024-06769-4
IF: 3.3
2024-12-04
The Journal of Supercomputing
Abstract:Z-Block-type optical combining device based on Lan-WDM technology are widely used in the transmitter end of optical communication in data centers, and with the wide range of applications, higher requirements are put forward for their production efficiency and dimensional accuracy. In order to solve the problems of low efficiency, susceptibility to human factor interference and low measurement accuracy in the traditional manual method of sorting and dimensioning measurement in the Z-Block production process, this study proposes a Z-Block centroid positioning and dimensioning scheme based on a machine vision method. Firstly, the Z-Block visual automatic sorting and dimensioning measurement system scheme and optical imaging scheme were designed to achieve high-quality imaging of Z-Block. Then, according to the problems of image edge noise, poor contrast, and weak edge caused by the complex production environment in the process of Z-Block visual imaging, an improved Canny edge detection algorithm was proposed, which used Blob analysis for denoising and ROI positioning, optimized gradient strength and direction calculation method and local dynamic threshold selection method, and realized high-accuracy edge detection. Finally, in order to solve the problems of inaccurate measurement point division and unsmooth overall contour in the edge detection results, a contour fitting method based on sub-pixel accuracy was proposed to achieve high-precision dimensional measurement. Experimental results show that the proposed method, on the basis of high-accuracy edge detection and high-precision contour fitting method, achieves the average error of center point positioning of (5, 5) pixel, the average error of dimensional measurement (0.03, 0.03) mm, and the average execution time of the algorithm is 143.34 ms, which can meet the automatic production of Z-Block.
computer science, theory & methods,engineering, electrical & electronic, hardware & architecture