兰 韬,初 侨,刘 文,戴 岳,杨 博,杨 丽,张晓芳,席兴军.基于深度学习的牛肉大理石纹智能分级研究[J].食品安全质量检测学报,2018,9(5):1059-1064
基于深度学习的牛肉大理石纹智能分级研究
Research on intelligent grading of beef marbling based on deep learning
投稿时间:2017-12-11  修订日期:2018-01-23
DOI:
中文关键词:  深度学习  牛肉  大理石纹  智能分级
英文关键词:deep learning  beef  marbling  intelligent grading
基金项目:中央级公益性科研院所基本科研业务费专项资金(562016Y-4489)
作者单位
兰 韬 中国标准化研究院食品与农业标准化研究所 
初 侨 中国标准化研究院食品与农业标准化研究所 
刘 文 中国标准化研究院食品与农业标准化研究所 
戴 岳 中国标准化研究院食品与农业标准化研究所 
杨 博 武汉大学测绘遥感信息工程国家重点实验室 
杨 丽 中国标准化研究院食品与农业标准化研究所 
张晓芳 中国标准化研究院食品与农业标准化研究所 
席兴军 中国标准化研究院食品与农业标准化研究所 
AuthorInstitution
LAN Tao Sub-Institutes of Food and Agriculture, China Institute of Standardization 
CHU Qiao Sub-Institutes of Food and Agriculture, China Institute of Standardization 
LIU Wen Sub-Institutes of Food and Agriculture, China Institute of Standardization 
DAI Yue Sub-Institutes of Food and Agriculture, China Institute of Standardization 
YANG Bo State Key Laboratory of Information Engineering in Surveying, Wuhan University 
YANG Li Sub-Institutes of Food and Agriculture, China Institute of Standardization 
ZHANG Xiao-Fang Sub-Institutes of Food and Agriculture, China Institute of Standardization 
XI Xing-Jun Sub-Institutes of Food and Agriculture, China Institute of Standardization 
摘要点击次数: 610
全文下载次数: 398
中文摘要:
      目的 开发客观、准确、无损的基于深度学习的牛肉大理石纹智能化分级技术。方法 将深度学习的图像识别方法应用于牛肉大理石纹的特征提取和分类上, 并进行相应的调试和学习。结果 通过计算机调试和学习, 评级正确率分别达到84.2%(一级)、89.4%(二级)、81.9%(三级)、84.1%(四级)、82.6%(五级)。各级牛肉的识别率均在80%以上, 识别时间都在1 s以内, 达到了预期目标。结论 将深度学习的图像识别方法应用于牛肉大理石纹的特征提取和分类上, 评级准确率非常高, 且随着图片数据库样本数的不断增多, 其识别的准确度将不断提高, 可进行大量推广使用。
英文摘要:
      Objective To develop an objective, accurate and nondestructive intelligent grading method of beef marbling based on deep learning. Methods The image recognition method based on deep learning algorithm was used for feature extraction and classification of beef marbling, and the corresponding debugging and learning was carries out. Results The accuracy rates of beef marbling grading reached to 84.2% for the first level, 89.4% for the second level, 81.9% for the third level, 84.1% for the forth level and 82.6% for the fifth level, respectively, through the computer debugging and learning. The recognition rates of beef at all levels were above 80%, the recognition time was all within 1 s, and the desired goal was achieved. Conclusion The image recognition method based on deep learning algorithm can be used for feature extraction and classification of beef marbling, and the accuracy of this method is good. With the increasing number of samples in the picture database, the accuracy of its recognition will be improved continuously, and this method can be widely used.
查看全文  查看/发表评论  下载PDF阅读器