关键词：自动化产线；普洱茶饼；机器视觉；图像处理；缺陷检测；可编程控制器中图分类号：S571.1；TP391.4；TP18 文献标识码：A DOI：10.7525/j.issn.1006-8023.2025.03.015

Abstract：With the popularization of automated production lines，thepresing process of tea cakeshas become particularlyimportantfor product quality.However，thequalitycontrolofPu'er mini tea cakes producedbyautomated productionlines oftenfalsshortofthose madebyhand.Therefore，the detectionof the appearance qualityof teacakes after production byautomated lines has become an urgent isse to beaddressd.To this end，this study proposes an automated quality detection algorithm for Pu'er tea cakesbased onmachine vision.Thealgorithmcomprehensively applies various image processing techniques，including Otsu threshold segmentation and Cannyedge detection，and introduces multiplealgorithmoptimization strategies to improve detectionaccuracyand eficiency.The algorithmcanautomatically completethedetectionand evaluationoftheappearance qualityof tea cakes and transmittheresults inreal-time toaProgrammable Logic Controler（PLC）．Experimentalresults show thatthe algorithm can accurately identify appearancedefects of tea cakes，with an average computational accuracy of 95. 7 5 % ，demonstrating high robustness and reliability. It is suitable forqualitycontrol in automated production lines and hasa widerange of application prospects，especially in the intelligent transformation of the tea production industry，where it has significant reference value.

Keywords：Automated production line； Pu'er tea cake;machine vision;image processing;defect detection； PLC

0 引言

代信息技术和智能农机装备在智慧农业领域的应用越来越广泛[1]。特别是茶叶产业，作为中国参与国际市场竞争的重要产业之一，其加工过程也越来越多地引随着科技的不断进步和市场需求的日益增长，现入了自动化生产线，尤其是茶叶的自动化压制设备。茶饼的压制过程对产品质量有着至关重要的影响[2]。然而，自动化生产线生产的茶饼质量控制相较于人工制作的茶饼仍显不足，因此，外观质量的检测成为关键环节。传统的人工检测方法效率低，且受主观判断影响较大，导致检测结果不稳定，且不便与自动化产线相结合[3]。

为解决这一问题，本研究提出一种基于机器视觉的自动化外观检测方法，专注于对自动化生产线生产出的茶饼进行形状和裂纹的精确分析，并将分析的结果传人可编程控制器（programmablelogic controller，PLC）中。该方法旨在提高检测效率和准确性，以确保茶饼外观质量的一致性，与自动化产线相连接，进而提升茶叶加工产业的自动化程度。而图像处理在工业产品质量检测中的应用已有多项研究。典型的研究分析主要是对产品外观进行评估。Pannawan等[4]在汽车方向包括通过边缘检测、形态学操作和轮廓分零件检测中采用了Canny边缘检测结合霍夫变换（houghtrans-form）方法来检测零件的形状缺陷，取得了良好的效果[5]。此外，李强等[在陶瓷工业中通过轮廓分析和形状匹配实现了陶瓷表面裂纹检测。……