基于鱼群社会等级和空间行为分析的养殖鱼类饥饿程度表征模型研究

Basic theory of quantitation for the hunger level of fishes in industrial recirculating aquaculture systems (RAS) is studied in this project with the help of the experimental object of juvenile Micropterus salmoides. The goal of this study is to construct the representation model with time delay for the whole hunger level of the juvenile Micropterus salmoides in industrial RAS. Four major parts are included in this project: study for the method in fast reconstruction for the spatial information of the fish school, containing individual sizes, swing degree and spatial structure; seeking for the method in unsupervised and quickly identification for the social hierarchy of the individual fish using its relevant characteristics; establishing the method in quantifying the behaviors caused by the hunger level of the fish school by means of analyzing the characteristics of individual social hierarchy, individual spatial behaviors, interaction of the spatial dynamic information between each other, peripheral distribution and synergistic effect of the flow field between each other; constructing the representation model with time delay for the whole hunger level of the fish school, and multiple indicators are adopted to test and optimize the constructed model mentioned above. This project is based on the field test data and carried out using the methods of the theoretical analysis and model fitting. The anticipated achievements of this project would provide not only the strategy of quantifying the hunger level of fishes in industrial RAS but also the theoretical guidance for the research and development of the intelligent feeding machine in aquaculture.

本项目将以加州鲈幼鱼为研究对象，针对工业化循环水养殖模式下的鱼类饥饿程度量化的相关基础理论进行研究，旨在构建适用于工业循环水养殖模式下的加州鲈幼鱼鱼群整体饥饿程度表征时滞模型。本项目研究内容包括：研究鱼群空间信息快速重建方法，计算鱼群个体大小、摆动程度及其空间分布信息；利用鱼群不同个体间的相关特征，探寻鱼群个体社会等级无监督快速判别方法；通过对鱼群个体社会等级特征、个体自身空间行为特征、个体间空间动态信息交互特征和个体间周边流场分布及其协同作用特征分析，建立鱼群整体饥饿程度表征行为的量化方法；构建鱼群整体饥饿程度表征时滞模型，并引入多重指标对所构建模型进行验证和参数优化。本项目拟以现场测试数据为基础，采用理论分析与模型模拟相结合的研究方法开展研究工作。项目预期成果不仅能够对工业化循环水养殖模式下的养殖鱼类提供可行的饥饿程度量化策略，还能够为水产养殖中智能化投喂装备的研发提供理论依据。

本项目以加州鲈等游泳型鱼类为研究对象，针对工业化循环水养殖模式下的鱼群饥饿程度精准量化难题进行研究，旨在构建适用于工业循环水养殖模式的游泳型鱼类整体饥饿程度表征模型。本项目研究内容包括：1）对具备表征鱼群饥饿程度的全局和局部突发行为特征进行了研究，研发了适用于该类行为（包括全局突发行为和局部突发行为）的快速监测方法，其平均准确率可分别达97.20%和89.89%；2）结合实际生产环境，建立了鱼群个体水下身份识别方法，实现了可利用极少标记样本对活鱼个体进行较高精度识别的功能，促进了鱼群突发行为中个体身份的精准量化；3）耦合鱼群摄食行为及其关联水体波动时空特征，研究了鱼群整体饥饿程度量化方法，构建了鱼群整体饥饿程度表征模型，同步引入多重指标对所构建模型进行了验证和参数优化；模型最终平均准确率可达98.31%，优于当前相关主流方法。本项目以现场测试数据为基础，采用理论分析与模型模拟相结合的研究方法开展研究工作。项目研究成果不仅将对工业化循环水养殖鱼类提供可行的饥饿程度量化策略，还能够为水产养殖中智能化投饲技术与装备的研发提供理论依据。