机器学习在植物病害识别研究中的应用

Application of machine learning in plant diseases recognition

植物病害识别是植物生长过程最重要也是最基本的环节,其既可以为高效除害提供最有力的依据,也可以减少一定的经济损失。随着信息技术的不断发展,在植物病害识别方面的研究工作已有一段历程,本文主要对机器学习技术在植物病害识别中的应用研究进行详细的综述。首先,通过调研植物病害问题的主要特征,明确植物病害识别研究中的识别任务;其次,阐述传统机器学习方法到深度学习的模式分类技术变迁,重点提出深度学习在植物病害识别中的应用优势;然后,调研机器学习在植物病害应用的相关研究文献,对文献所使用的模型、技术细节、数据来源、数据处理技术以及性能指标评价进行详细综述与对比,分析该领域研究存在的问题;最后,基于调研结果对植物病害识别的进一步研究展开讨论,同时对研究对象的特点与大规模数据集合的构建提出相关意见,在技术上提倡深度学习算法的使用,鼓励更加先进的模型尝试等建议。另外,还整理目前已经公开且可以下载使用的关于植物病害识别研究的数据库集合,为相关的研究提供便利。

Plant diseases recognition plays important role in plant growth process. It can not only provide the most powerful basis for eliminating diseases, but also reduce certain economic losses. As machine learning has been successfully applied in various domains, it has entered also the domain of plant diseases recognition. In this paper, we performed a survey of the application of machine learning technology in plant diseases recognition. We examined the main characteristics of plant diseases under study, the classical models and frameworks employed, the sources, nature and pre-processing of data used, and the overall performance achieved according to the metrics used at each work under study. On the basics of analyzing the technical transformation from traditional machine learning to deep learning in the application of pattern recognition, the developing trends of plant diseases recognition technologies were prospected, and the advantages of deep learning in this application were presented. Thus, we advocated the use of deep learning and analyzed the challenges of current research. For future work, we encourage researchers to develop more advanced learning architectures. It is also important to emphasize that the image databases used in the experiments is the most basic requirement. To address this problem, some datasets were made openly and freely available as collected in this article.