摘 要: 针对铝材入库计数过程存在效率低和耗时长的问题,提出了一种基于机器视觉和改进的图像金字塔的铝材计数方法。该方法应用Sobel边缘检测提取模板轮廓,采用加权最小二乘法进行直线拟合实现模板轮廓的优化处理,进而得到最优模板,改进后的图像金字塔算法实现了对图像模板匹配的快速计数。实验结果表明,所提出的方法应用于铝材入库计数,具有较高的召回率和较快的匹配速度,测试数据集的平均检测准确率达到98.4%,匹配效率相较于传统图像金字塔算法提升了37.0%。 |
关键词: 边缘检测;模板匹配;直线拟合;图像金字塔 |
中图分类号: TP391.4
文献标识码: A
|
基金项目: 国家自然科学基金面上项目(52175513) |
|
Application of Improved Multi-objective Template Matching in Aluminum Counting |
GAO Xusheng, CHENG Xiaorong, WANG Yuanye, WANG Zixuan, XU Tin
|
(School of Optical-Electrical and Computer Engineering, University of Shanghai f or Science and Technology, Shanghai 200093, China)
15800961656@163.com; cxrsjtu@163.com; 1194701571@qq.com; 823196574@qq.com; 1565912192@qq.com
|
Abstract: This paper proposes an aluminum counting method based on machine vision and an improved image pyramid, to address the problems of low efficiency and time consumption in the process of counting stock-in aluminum products. In this method, Sobel edge detection is used to extract the template contour that is optimized by weighted least squares method for line fitting, so as to obtain the optimal template. The improved image pyramid algorithm achieves fast counting of image template matching. The experimental results show that the proposed method has a high recall rate and fast matching speed when applied to aluminum stock-in counting. The average detection accuracy of the test dataset reaches 98.4% , and the matching efficiency is improved by 37.0% compared to traditional image pyramid algorithms. |
Keywords: edge detection; template matching; linear fitting; image pyramid |