Abstract The human impact on the coastal ecosystems is a global environmental concern. Due to the growing urbanization, industrialization, and transportation, this impact on the living and non-living components of the coastal area is expected to further increase in the coming years. Artificial intelligence based automation of the coastal monitoring, including data collection, analysis and decision making, provides real-time insights and opportunities for large-scale coastal management and governance. In this paper, a framework for autonomous litter surveying and human activity monitoring for governance intelligence in coastal eco-cyber-physical systems (ecoCystem) is presented. A large dataset of more than 20,000 images focused on smart coastal management is collected to model the real world scenarios. A combination of various artificial intelligence based methods are used for automatic detection and classification of various litter in the coastal environment. Furthermore, the proposed framework is capable of autonomous monitoring of humans activities and detection of illegal entry of vehicles and boats to the beach area. The accuracy of the proposed autonomous system is 87% for correct classification of fully visible litter and 95% for fully visible vehicles. The experimental results show that the application of computer vision and machine learning for autonomous litter classification shows promising results for increasing the speed and scale of litter surveying in the coastal area. Further training of the artificial intelligence models is necessary for increasing the accuracy of the proposed framework and real-world deployment in the coastal environment. The proposed human activity monitoring system can be used for autonomous coastal law enforcement and real-time and active protection of the coastal zones.

中文翻译：

---

沿海生态信息物理系统治理智能的自主垃圾调查和人类活动监测

摘要 人类对沿海生态系统的影响是一个全球性的环境问题。由于城市化、工业化和交通运输的不断发展，预计未来几年对沿海地区生物和非生物部分的影响将进一步增加。基于人工智能的沿海监测自动化，包括数据收集、分析和决策制定，为大规模沿海管理和治理提供实时洞察和机会。在本文中，提出了一个用于沿海生态网络物理系统（ecoCystem）中治理智能的自主垃圾调查和人类活动监测框架。收集了一个包含 20,000 多张专注于智能海岸管理的图像的大型数据集，以模拟现实世界的场景。结合各种基于人工智能的方法用于沿海环境中各种垃圾的自动检测和分类。此外，拟议的框架能够自动监测人类活动并检测非法进入海滩区域的车辆和船只。所提出的自主系统对于完全可见垃圾的正确分类准确率为 87%，对于完全可见的车辆准确率为 95%。实验结果表明，计算机视觉和机器学习在自主垃圾分类中的应用在提高沿海地区垃圾调查的速度和规模方面取得了可喜的成果。人工智能模型的进一步训练对于提高拟议框架和沿海环境中实际部署的准确性是必要的。拟议的人类活动监测系统可用于沿海自主执法和沿海地区的实时主动保护。