摘要
[目的]为实现对鸡胴体质量等级的非接触在线自动判定,本文提出一种基于机器视觉和机器学习技术的鸡胴体等级在线检测方法。[方法]首先用图像采集装置获取鸡胴体图像,对图像进行预处理并提取包括投影面积、胴体长度、轮廓长度、鸡胸长度、鸡胸宽度和鸡胸面积等图像特征参数。在对所得到的数据进行无量纲处理后,再以这6个特征参数为输入、质量为输出,分别利用随机森林(random forest,RF)算法、自适应提升算法(Adaboost,AB)和梯度提升算法(gradient boosting,GB)3种机器学习方法,建立鸡胴体质量等级预测的非线性回归模型,对鸡胴体质量等级进行自动判定。[结果]在对鸡胴体质量进行预测时,梯度提升模型的判定系数最大,为0.996 0,明显优于线性模型,也优于其他2种非线性模型;在对鸡胴体质量等级进行判定时,也是梯度提升模型的判定正确率最高,为96%。[结论]可利用梯度提升模型对鸡胴体质量和等级进行精确预测和判定。
[Objectives]In order to realize non-contact on-line automatic determination of chicken carcass weight grade,an on-line detection method based on machine vision and machine learning technology was proposed.[Methods]Firstly,the chicken carcass image was acquired by image acquisition device.The image was preprocessed and the characteristic parameters including projection area,carcass length,contour length,chicken breast length,chicken breast width and chicken breast area were extracted.And then take these six characteristic parameters as input and quality as output.After dimensionless processing of the data,the nonlinear regression model for predicting the weight grade of chicken carcass was established by using three machine learning methods such as RF(random forest),AB(Adaboost)and GB(gradient boosting).[Results]The results showed that the gradient boosting model had the largest coefficient of 0.996 0,which was obviously superior to the linear model and the other two non-linear models.The gradient boosting model had the highest accuracy rate of 96%when judging the weight grade of chicken carcass.[Conclusions]The gradient boosting model can be used to predict and determine chicken carcass weight and grade accurately.
作者
戚超
徐佳琪
刘超
吴明清
陈坤杰
QI Chao;XU Jiaqi;LIU Chao;WU Mingqing;CHEN Kunjie(College of Engineering,Nanjing Agricultural University,Nanjing 210031,china)
出处
《南京农业大学学报》
CAS
CSCD
北大核心
2019年第3期551-558,共8页
Journal of Nanjing Agricultural University
关键词
鸡胴体
机器视觉
机器学习
预测模型
分级
chicken carcass
machine vision
machine learning
prediction model
classification
