| 基于机器视觉的槟榔分级方法研究 |
| |
| 引用本文： | 李志臣,凌秀军,李鸿秋.基于机器视觉的槟榔分级方法研究[J].中国农机化学报,2023,44(1):137. |
| |
| 作者姓名： | 李志臣 凌秀军 李鸿秋 |
| |
| 作者单位： | 金陵科技学院机电工程学院，南京市，211169 |
| |
| 基金项目： | 国家自然科学基金面上项目(51775270) |
| |
| 摘 要： | 针对槟榔人工分级劳动生产率低、准确率低的问题，开展基于遗传神经网络的机器视觉槟榔分级研究。以4种类别的槟榔图像为研究对象，首先设计一个6层结构的遗传神经网络对槟榔进行分级，虽然分级准确率较高但是网络结构复杂。然后对运用主成分分析法降低图像特征的维数并将遗传神经网络简化为3层结构的方法进行研究。最后用400幅和100幅槟榔图像对这个3层神经网络进行训练和验证，经过调整网络的学习率等参数，训练和验证的准确率达到95%以上。通过神经网络模型测试试验，槟榔正确分级的准确率为90%。数据降维后的三层遗传神经网络能够实现对槟榔的实时分级，为机器分级提供了技术支持。
|
| 关 键 词： | 槟榔 神经网络 分级 主成分分析法 机器视觉 |
Study on betel nut classification method based on machine vision |
| |
| Abstract: | Aiming at the problems of low labor productivity and low accuracy of betel nut manual classification, a study on machine vision betel nut classification based on genetic neural network was carried out. Using four kinds of betel nut images as the research object, a 6-layer structure genetic neural network was designed to grade betel nut. Although the classification accuracy was high, the network structure was complex. The method of reducing the dimensionality of the image features and reducing the genetic neural network to a 3-layer structure was studied. The 3-layer neural network was trained and validated with 400 and 100 betel nut images. After adjusting the learning rate of the network, the trained and validated accuracy reached more than 95%. And the accuracy of betel nut classification was 90%. The three layer genetic neural network after data dimension reduction can realize the real time betel nut grading, which provides technical support for machine grading. |
| |
| Keywords: | betel nut neural network grading principal component analysis machine vision |
|
| 点击此处可从《中国农机化学报》浏览原始摘要信息 |
| 点击此处可从《中国农机化学报》下载免费的PDF全文 |
