基于机器视觉的孵化早期群体受精蛋鉴别  

Identification group of fertilized eggs and infertile eggs in early hatching stage based on machine vision

作　　者：吴林峰 余怀鑫 祝志慧[1] WU Lin-feng;YU Huai-xin;ZHU Zhi-hui(College of Engineering,Huazhong Agricultural University, Wuhan ,Hubei 430070 , China)

机构地区：[1]华中农业大学

出　　处：《食品与机械》2019年第4期152-156,共5页Food and Machinery

基　　金：中央高校基本科研业务费资助(编号:2662017PY057);公益性行业(农业)科研专项(编号:201303084)

摘　　要：针对群体种蛋信息在线检测困难问题,采用机器视觉技术对基于工业蛋托的群体鸡种蛋受精信息进行检测,将整托蛋直接从孵化箱放进检测装置获取群体种蛋图像,对图像进行分割、平滑去噪,提取图像RGB、HIS、灰度均值以及蛋重作为特征参数,分别建立了基于多元线性回归、支持向量机(SVM)和BP神经网络鉴别模型。试验结果表明,3种模型中,SVM模型具有较高的稳定性和准确率,在第3天和第7天分别达到了81.7%和96.7%,为群体种蛋信息在线检测提供了一种可行的方法。In the early stage of incubation.the infertile eggs still have certain edible value, and if removed earlier, they can not only reduce economic losses, but also avoid the impact on other normally hatched eggs.In order to solve the problem of online detection of group eggs information.this paper uses machine vision technology to detect the fertilization information of group chicken eggs based on industrial egg trays for the first time.The whole eggs were directly put into the detection device from the incubator to obtain the group egg images, which reduced unnecessary damage to the eggs and improves the efficiency, the image was segmented and smoothing denoised.and RGB.HIS, gray mean and egg weight of it were extracted as characteristic parameters by establishing identification models of support vector machine (SVM) and BP neural network.The experimental results show that among the three models, the SVM model has higher stability and accuracy, reaching 81.7% and 96.7% on the 3rd and 7th days respectively, which provides a feasible method for online detection of group egg information.

关 键 词：机器视觉 群体 受精蛋 无精蛋 无损检测 

分 类 号：S879.3[农业科学—畜牧兽医] TP391.41[自动化与计算机技术—计算机应用技术]