面向地衣的多模态交互式在线识别方法研究

The species and genetic resources of lichens is extremely rich, and the development of lichens resources has just started currently in China. The recognition of lichens is the primary task of lichens resources development. Faced the automatic fast and accurate recognition of lichen, this project will introduce the multi-modal intelligent human-computer interaction technology to carry out research, based on the knowledge and experience of lichen fungi experts. Faced lichen, this project will focus on the effective feature extraction method of multi-modal large data, the construction of space-time evolution recognition model and the interactive online updating method, etc. At the end, the project will realize a multi-modal interactive online recognition prototype system of stone ear lichen. The Beijing Forestry University, Institute of Software Chinese Academy of Science and Institute of Microbiology Chinese Academy of Science will take part in the project and play their respective advantages of multi-modal data effective processing, recognition model construction, new human-computer interaction, lichen fungal knowledge ontology and biodiversity research, etc. There will be many key technology breakthroughs like building a mapping model from the space of multimodal physical parameters to that of fungi effective recognition and online model update method by introducing active interactive pattern. All of these will promote the positive application of new human-computer interaction technology in terms of fungus automatic and fast recognition.

地衣的物种资源和基因资源极为丰富，而我国目前对地衣资源的开发才刚刚起步。对地衣的识别是其资源开发的首要任务，本项目面向地衣的自动快速准确识别，拟基于地衣真菌专家的知识和经验，引入多模态智能人机交互技术展开研究。项目将重点研究面向地衣的多模态大数据的有效特征提取方法、时空演化识别模型的构建与交互式在线更新方法等内容，最后实现一套面向石耳科地衣的多模态交互式在线识别原型系统。项目将联合中科院软件研究所和中科院微生物研究所，发挥各自在多模态数据有效处理、识别模型构建、新型人机交互、地衣型真菌知识本体及生物多样性研究等方面的优势，突破多模态物理参数空间到真菌有效识别特征空间的映射模型构建、引入主动交互模式的模型在线更新等关键技术，共同推进新型人机交互技术在真菌自动快速识别方面的积极应用。

地衣型真菌的物种资源和基因资源极为丰富，而我国目前对地衣资源的开发才刚刚起步，对地衣的识别是其资源开发的首要任务。本项目针对地衣等真菌的快速、准确智能识别问题开展研究，融合真菌专家的知识和经验，主要研究实现了细粒度图像分类方法、特征自适应的在线增量学习方法等关键技术，并研发了一套菌物智能识别与管理系统，目前已在中科院微生物研究所进行应用示范。相关成果在ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies（UbiComp 2021）、Computational Intelligence and Neuroscience等国内外知名会议和期刊上发表论文4篇，获批软件著作权2项，培养硕士生2名，获2018年度《农业机械学报》优秀论文。