基于机器视觉的马铃薯自动分级与缺陷检测系统设计

Design of Potato Automatic Grading and Defect Detection System Based on Machine Vision

针对现有的马铃薯分级和检测需要大量的人力物力、检测效率不高,设计了基于机器视觉的马铃薯自动分级与缺陷检测系统。工作时,自动分级系统对大量马铃薯进行快速表皮去泥和分级工作,得到3种规格的马铃薯并逐个运输到缺陷检测系统进行马铃薯缺陷的识别检测;通过多种图像处理算法对比分析,以平均值法灰度化、中值滤波处理、大津法分割等方法得到最佳的马铃薯图像,且目标图像能与背景图像很好分割,提高了缺陷检测的准确度和效率;采用RGB彩色模型对马铃薯图像进行分析,以马铃薯图像设定的阈值与标准差值相比较,得到图像中所有缺陷点,并对马铃薯图像缺陷部分的连通区域进行标记。选择1000个试验样本进行系统和人工分级与检测的试验,结果表明:自动分级系统对不同类别大小的马铃薯分级有较高的准确度,缺陷检测系统对多种缺陷的检测准确精度很高,并验证了马铃薯缺陷检测系统的可行性。

The existing potato grading and testing needs a lot of manpower and material resources,and the detection efficiency is not high,so the design of automatic potato grading and defect detection system based on machine vision is proposed.A large number of potatoes were rapidly desilted and graded by the automatic grading system,and three potato specifications were obtained and transported to the defect detection system one by one for identification and detection of potato defects.Through the comparative analysis of various image processing algorithms,it is concluded that the best potato image can be obtained by means of mean value graying,median filter processing and OTSU segmentation,and the target image can be well segmined with the background image,which improves the accuracy and efficiency of defect detection.RGB color model is adopted to analyze potato image.All defect points in the image can be obtained by comparing the threshold value set by the potato image with the standard deviation value,and the connected region of the defect part of the potato image can be marked.1000 test samples were selected for systematic and manual grading and testing.The results show that the automatic grading system has high accuracy in grading potatoes of different sizes.The accuracy of defect detection system is very high,which verifies the feasibility of potato quick detection system.