基于Python-OpenCV的管柱图像快速识别与管径分类研究及优化  

Study and Optimization of Pipe Column Images Fast Recognition and Pipe Diameter Classification Based on Python-OpenCV

作　　者：牛柯 彭斌[1] Niu Ke;Peng Bin(School of Mechanical and Electrical Engineering,Lanzhou University of Technology,Lanzhou 730050,China)

机构地区：[1]兰州理工大学机电工程学院,兰州730050

出　　处：《科技通报》2024年第11期50-56,共7页Bulletin of Science and Technology

基　　金：国家自然科学基金项目(51675254);国家重点研发计划项目(SQ2020YFF0420989)。

摘　　要：本文提出一种基于OpenCV图像处理技术的管柱零件目标检测方法。通过对传统检测方法的改进和优化,实现了对管柱图像的快速识别与管径分类。经过图像增强处理和预处理操作,包括图像灰度化、二值化、滤波和边缘检测,对灰度图像反色处理,提高灰度图像的质量,根据灰度直方图的分布确定二值化最佳阈值范围。中值滤波PSNR(peak signal-to-noise ratio)值与SSIM(structural similarity)值与双边滤波和非局部均值滤波相比提高了56%和85%;确定中值滤波内核值为13×13。通过仿真确定采用Otsu的分割方式分割图像与Canny边缘检测阈值相结合,可以有效提高Canny算法的自适应能力以及提高边缘的连接度和准确性。验证结果表明:改进后的Canny-like算法在边缘检测准确率上对比传统Canny、Sobel、Prewitt、Laplacian等算法均有16%左右的提升,该算法优化了传统算法检测的连续性和准确率。最后将该检测方法应用于不同环境下的管柱检测,取得较高的检测精度,满足实时准确检测的要求。Research on string recognition technology based on Python-OpenCV in this paper.Through the improvement and optimization of the traditional detection method,the fast recognition of the string image and the classification of the pipe diameter are realized.After image enhancement processing and pre-processing operations,including image grayscale,binarization,filtering,and edge detection;the quality of the gray image is improved by inverting the gray image,and the optimal threshold range of binarization is determined according to the distribution of the gray histogram.Compared with bilateral filtering and non-local means filtering,the PSNR(peak signal-to-noise ratio)value and SSIM(structural similarity)value of median filtering are increased by 56%and 85%on average;The median filtering kernel value was determined to be 13×13;Through the simulation,it is determined that combining the Otsu segmentation method and the Canny edge detection threshold can effectively improve the adaptive ability of the Canny algorithm and improve the edge connectivity and accuracy.The verification results show that the improved Canny-like algorithm has a 16%improvement in edge detection accuracy compared with the traditional Canny,Sobel,Prewitt,Laplacian,and other algorithms.The algorithm optimizes the continuity and accuracy of the traditional algorithm.Finally,the detection method is applied to the pipe string detection in different environments,and high detection accuracy is obtained,which meets the requirements of real-time and accurate detection.

关 键 词：目标检测 OPENCV 边缘检测 中值滤波 OTSU 

分 类 号：TH45[机械工程—机械制造及自动化]