基于北斗数据的东黄海捕捞渔船行为模式量化与分析方法研究

Since ancient times, the East China Sea and Yellow Sea provide people with abundant animal protein. They have a number of famous traditional fishing grounds. But in recent years due to high intensity of fishing, fisheries resources in the East China Sea and Yellow Sea are decreased and the sustainable development of resources is affected. The bottom of the sea floor is mainly sand and clay. Because of its small shear strength, it is affected by fishing vessel easily. China has long been on total control of fishing vessel number, tonnage, power to implement and the management of the fishing moratorium, prohibit fishing area. The management system plays an important role in the protection of resources. With the application of Beidou and remote sensing technology in marine fisheries, there is still room for further management. The project relies on the Beidou system. The trawl, gillnet, stow net, and seiner fishing in the East China Sea and Yellow Sea are as research objects. The identification of fishing vessels is studied based on the BP neural network and the statistical characteristic curve of speed. The navigation speed threshold division and hidden Markov algorithm are used to study the status of fishing vessels. The results are validated by remote sensing. The cluster analysis method is used to study on influence factors of fishing vessels. And quantitative attribute to voyage data are as the foundation. Based on the data statistics and the longest common subsequence distance method, the project studies the relationship between the optimal efficiency and the effective distance. The cumulative value is calculated with grid technique to quantify fishing vessels intensity. It is in order to reveal the temporal and spatial distribution of fishing effort, the sources of fishing pressure in the sea area, and whereabouts the fishing effort of the zoning fishing vessels. It will provide scientific basis for fishery management and conservation, and also provide theoretical basis for the research of fishing vessel behavior.

自古以来，东黄海为人们提供着丰富的动物蛋白，有着多个闻名的传统渔场，而近年来因高强度捕捞，出现资源衰退，影响了渔业可持续发展。我国的渔船总量“双控”，以及休渔期、禁渔区制度对资源保护起到重要作用，但随着北斗与遥感技术在海洋渔业中的应用，管理方式仍有深化的空间。本项目依靠北斗系统，以东黄海拖网、刺网、张网、围网渔船为研究对象，采用BP神经网络和航速统计特征曲线，研究渔船类型的识别，采用航速航向阈值划分和隐马尔可夫算法，研究渔船状态的判断，并辅以海上调查和遥感结果验证；利用聚类分析法，以航次属性量化数据为基础，研究渔船捕捞影响因素；利用空间分析与最长公共子序列距离方法研究互助渔船关系，阐明生产高效最佳距离、救援有效距离；基于格网累计值技术量化渔船捕捞强度，揭示捕捞努力量时空规律、海域捕捞压力渔船来源、区划渔船捕捞努力量去向。为渔业管理和资源养护提供科学依据，也为捕捞渔船行为研究提供理论基础。

海洋渔业已成为我国自主北斗导航系统最大的行业用户之一，已有近 6 万艘海洋机动捕捞渔船安装了北斗终端。北斗终端能获取并记录包括渔船位置、航速、航向等在内的数据信息，船位数据挖掘结果可为渔船管理和渔业资源保护等提供时空高精度的决策信息。项目开展了渔船作业类型识别、状态判断、航次提取、捕捞追溯、捕捞努力量评估等研究。在渔船类型识别方面，自定义了10层CNN模型及使用迁移学习和模型微调方法调整后的VGG-16模型，实现刺网与拖网作业类型识别精度达94.3%。在渔船捕捞作业航次量化与分析方面，基于DBSCAN提取定置刺网渔船网次，通过航次特征数据挖掘，提取到2.5万余艘渔船的39.98万个航次数据，分析不同船长、类型、作业期、地区等条件下航次和平均离岸距离的特点，渔船的航次时长、航程和平均离岸距离均由江苏省向北、向南降低。在渔船行为模式研究方面，开展了流刺网网次和方向提取方法研究，使用阈值综合判别的方法判断捕捞作业的状态，通过航速、空间距离、时间间隔和航向变化的阈值判别作业时收网状态的船位点，根据收网状态的起始点判定放网状态的起始点，应用时空统计方法，对浙江省流刺网渔业的捕捞行为时空分布进行了分析。开展了张网渔船捕捞作业行为量化与分析研究，设计了张网渔船状态判断、确定网位、放网时长提取和捕捞努力量计算方法，采用阈值划分和密度聚类算法提取网位，计算各航次的捕捞时长;划分地理格网并累加其范围内的捕捞时长，以各格网平均每平方公里累计捕捞时长作为帆张网捕捞强度的量化依据，以设定航速阈值、距离阈值和时间阈值来提取渔运船转载信息的方法。开展了捕捞渔船异地停靠信息提取与分析研究，分析了捕捞渔船出海时间与航程量化分析。研究结果可应用于禁渔区、协定区、休渔期内的非法捕捞行为识别，渔船捕捞作业行为分析等，随着北斗导航全球布局的实现，其在海洋渔业领域的应用前景将更为广阔。