用人工智能保护环境，英特尔推能监测珊瑚礁健康状况 AI 项目

对于保护环境，人工智能（AI）能做什么？有人想到了用人工智能来监测和分析珊瑚礁的复原能力。

由英特尔、埃森哲咨询公司和菲律宾一家非营利组织 Sulubaaï 环境基金，推出了名为 Corail 项目，它的目的在于用人工智能搭建平台，来监测和分析珊瑚礁的复原力。自 2019 年 5 月推出以来，这个项目已经收集了 40000 张菲律宾潘加塔兰岛周围珊瑚礁的图像，研究人员利用这些图像实时评估了珊瑚礁的健康状况。

潜水员在水下布置 Corail 项目

如果在菲律宾的试点计划获得成功，那么 Corail 项目就可以用来监测世界上更多的濒危珊瑚礁种群。珊瑚礁不仅可以保护海岸线免受热带风暴袭击，而且还能为十亿人提供经济收入，仅围绕旅游和娱乐这一项，它就能为每年产生 96 亿美元的价值。

具体来说，Corail 项目一开始会部署一个浮标。浮标配备船用级太阳能电池板、电池和信号传输设备（可能是 WiFi 或者 4G）。同时，研究团队还配了一个系在泊绳上的摄像头。浮标上的计算由树莓 Pi 和英特尔神经计算芯片 2 来完成，岸上的 PC 计算芯片则由英特尔 Arria 10 FPGA 来处理图像。

要部署浮标需要搭建一个人造混凝土礁，据英特尔称，他们采用了苏禄礁假体融合了会生长和扩展的活珊瑚碎片，这样这个水下平台不会破坏环境，还为鱼类和海洋生物提供了栖息地。最初这种苏禄礁假体由 Sulubaaï 环境基金会于 2016 年设计，旨在提高岛上珊瑚生态系统的复原力，并恢复被炸药捕捞破坏的礁石。据介绍，任何类型的珊瑚都可以附着在假体上，调查显示，苏禄礁假体固定第一年的成活率为 76%。

通过摄像头拍摄的图像通过芯片上运行的 Azure 应用程序定期把数据从相机传输到岸上的 PC。埃森哲提供的定制机器学习模型 VASP 对海洋生物进行计数和分类，并将数据从 PC 发送到到 Azure 云，从而进行整理形成报告。

埃森哲首席分析官 Athina Kanioura 表示：「数据的价值取在于你能多快地收集到见解并做出决策。凭借对流视频进行实时分析的能力，VASP 使我们能够利用丰富的数据源，在不破坏水下环境的情况下进行手动监视。」

更好地了解珊瑚礁的健康状况可以帮助研究人员监测气候变化趋势。不论是污染，海洋酸化，还是不可持续的鱼类捕捞或水温升高都已经直接影响了水下生态系统。

未来，Corail 项目团队还计划部署优化的 AI 分类器模型和备用电源。他们还考虑使用红外摄像头，该摄像头可以进行夜间拍摄，并调整平台以研究热带鱼的迁移率以及进入人类受限的水下区域。

最后，值得一提的是， Corail 项目并不是人工智能技术首次实验的环境保护项目。去年，英特尔还发布了 TrailGuard AI 项目，这是与非营利组织 Resolve 合作，来遏制非洲象的偷猎活动。此外，英特尔与微软合作的「AI 为地球」计划与数据科学咨询公司 Gramener 合作，可以帮助生态学家了解南极企鹅的数量。
What can artificial intelligence (AI) do to protect the environment? Some people think of using artificial intelligence to monitor and analyze the resilience of coral reefs.
The Corail project, launched by Intel, Accenture consulting and sulubaa, a Philippine non-profit organization, aims to build a platform with artificial intelligence to monitor and analyze the resilience of coral reefs. Since its launch in May 2019, the project has collected 40000 images of coral reefs around pangataran island in the Philippines, which researchers use to assess the health of reefs in real time.
Divers lay out the Corail project underwater
If the pilot project in the Philippines is successful, the Corail project could be used to monitor more endangered reef populations around the world. Coral reefs can not only protect coastlines from tropical storms, but also provide economic income for one billion people. Focusing on tourism and entertainment alone, they can generate a value of 9.6 billion dollars per year.
Specifically, the Corail project will initially deploy a buoy. The buoy is equipped with marine grade solar panels, batteries and signal transmission equipment (may be WiFi or 4G). At the same time, the team also equipped a camera tied to the mooring line. The calculation on the buoy is done by raspberry PI and Intel neural computing chip 2, while the PC Computing chip on the shore is processed by Intel aria 10 FPGA.
To deploy the buoy, a man-made concrete reef needs to be built. According to Intel, they have used Sulu reef prosthesis to integrate live coral fragments that will grow and expand, so that the underwater platform will not damage the environment, but also provide a habitat for fish and marine life. Originally designed by sulubaa environment foundation in 2016, the Sulu reef prosthesis aims to improve the resilience of the island’s coral ecosystem and restore reefs damaged by explosive fishing. According to reports, any type of coral can be attached to the prosthesis. According to the survey, the first year survival rate of Sulu reef prosthesis fixation is 76%.
The image captured by the camera is regularly transferred from the camera to the onshore PC through the azure application running on the chip. The customized machine learning model VASP provided by Accenture counts and classifies marine organisms, and sends data from PC to azure cloud, so as to organize and form a report.
“The value of data is how quickly you can gather insights and make decisions,” said Athina kanioura, chief analysis officer of Accenture. With the ability of real-time analysis of streaming video, VASP enables us to use rich data sources for manual monitoring without damaging the underwater environment. “
A better understanding of the health of coral reefs can help researchers monitor climate change trends. Whether it is pollution, ocean acidification, unsustainable fishing or rising water temperature, it has a direct impact on the underwater ecosystem.
In the future, Corail project team also plans to deploy optimized AI classifier model and standby power supply. They also consider using an infrared camera that can take pictures at night and adjust the platform to study the mobility of tropical fish and access to restricted underwater areas for humans.
Finally, it is worth mentioning that the Corail project is not the first environmental protection project to be experimented with artificial intelligence technology. Last year, Intel also released trailguard AI, a partnership with resolve, a non-profit organization, to curb elephant poaching in Africa. In addition, Intel’s “AI for the earth” program, in collaboration with Microsoft, and data science consulting firm grameer, can help ecologists understand the number of Antarctic penguins.