摘要
深度学习方法因其具有学习能力强、覆盖范围广、自适应力强、可移植性好等优点,适用于解决实际生产的非线性问题,在农业领域得到了广泛研究和应用。文章简述了深度学习的概念及其特点,从种植业和养殖业2个方面阐述了深度学习在农业领域的研究现状;详细介绍了在分类识别、病虫害识别及预测、动物身份识别及行为监测等方面的研究进展及效果;总结了目前制约深度学习方法进一步应用的原因是样本数据量大、处理要求高和硬件不匹配等;最后对深度学习的发展趋势进行了展望。
With advantages of high self-study ability,wide coverage area,strong adaptability and sound portability and etc.,deep learning methods could be used to solve non-linear problems in actual production.They have been widely applied in agricultural fields.In this paper,the concepts and characteristics of deep learning methods were described.The research status in agricultural fields was illustrated,especially in planting industry and breeding industry.The research progress and effects were also introduced in detail in the fields of classification recognition,pest identification and predication,animal identification and behavior monitoring and etc.The influencing factors were also studied,such as large sample size,high-level processing requirements and mismatched hardware and etc.In the end,the development tendency of deep learning was prospected.
作者
马聪
张建华
陈学东
朱丹
Ma Cong;Zhang Jianhua;Chen Xuedong;Zhu Dan(Institute of Agricultural Economy and Information Technology,Ningxia Academy of Agriculture and Forestry Sciences,Yinchuan,Ningxia 750002)
出处
《宁夏农林科技》
2020年第1期35-37,42,F0003,共5页
Journal of Ningxia Agriculture and Forestry Science and Technology
基金
宁夏农林科学院科技创新引导项目“基于机器视觉及深度学习的枸杞智能精选算法研究”(NKYQ-20-04)。
关键词
深度学习
卷积神经网络
病虫害识别
分类
行为监测
Deep learning
Convolutional neural network
Pest identification
Classification
Behavior monitoring
