据中国核建消息，6月3日，中核华兴在南京发布了“核建安全监护哨兵”与“核建安巡机器犬”两款核电建造安全智能监控机器人。经专家评审，该技术填补了核电建造领域智能安全监控的技术空白，已具备工程化应用条件。

According to information from China Nuclear Construction, on June 3rd, China Nuclear Huaxing in Nanjing released two types of intelligent safety monitoring robots for nuclear power construction - "Nuclear Construction Safety Guardian Sentinel" and "Nuclear Construction Patrol Robot Dog". After expert review, this technology has filled the technical gap in intelligent safety monitoring in the field of nuclear power construction and has met the conditions for engineering application.

破局启航：以“智械”作答

Breakthrough and Set Sail: Answer with "Mechanical Intelligence"

针对核电工程建设中复杂的安全管控需求，中核华兴程小华劳模创新工作室研发了这两款机器人。“核建安全监护哨兵”是一款便携式定点监控设备，重约七公斤，集成了高清摄像头、红外热成像模块和气体传感器，能实时识别风险隐患、监测人员状态与环境数据。“核建安巡机器犬”基于四足机器人平台，具备强大地形适应能力，可自主巡检，实现厘米级定位与数据实时回传。两者协同，构建起全天候的安全监控网。

To meet the complex safety control requirements in nuclear power plant construction, the Zhonghe Huaxing Cheng Xiaohua Model Innovation Studio has developed these two robots. "Nuclear Construction Safety Guardian Sentinel" is a portable fixed-point monitoring device, weighing approximately seven kilograms. It integrates high-definition cameras, infrared thermal imaging modules and gas sensors, and can promptly identify potential risks, monitor personnel status and environmental data. "Nuclear Construction Safety Patrol Robot Dog" is based on a quadruped robot platform and has strong terrain adaptability. It can conduct autonomous inspections and achieve centimeter-level positioning and real-time data transmission. Together, they form a round-the-clock safety monitoring network.

蝶变升级：给摄像头装上“大脑”“嘴巴”和“鼻子”

Transformation and Upgrade: Equipping Cameras with "Brains", "Mouths" and "Noses"

研发初期，“前端拍摄、后端计算”的模式因数据延迟被放弃。团队转而采用边缘AI计算系统，让算法在设备本地部署。通过采用YOLO与TSM模型，自建标注体系并采集数万张核电环境样本进行训练，构建了专属安全数据库。在白龙核电项目实测获得认可后，设备进一步集成了声光报警、热成像、气体传感器、传呼装置和可插拔电池等功能。同时，后台“安全驾驶舱”作为“中枢大脑”，实现了数据的集中管理与全流程追溯。

In the early stage of development, the "front-end shooting, back-end computing" model was abandoned due to data latency. The team then switched to an edge AI computing system, deploying the algorithms locally on the devices. By adopting the YOLO and TSM models, establishing their own annotation system and collecting tens of thousands of nuclear power environment samples for training, they constructed an exclusive safety database. After being recognized in the Bai Long Nuclear Power Project, the equipment was further integrated with functions such as sound and light alarms, thermal imaging, gas sensors, paging devices, and replaceable batteries. Meanwhile, the "safety cockpit" at the back end, acting as the "central brain", achieved centralized management of data and full-process traceability.

淬火成钢：让AI“看懂”核电工地

Quenching into steel: Making AI "understand" nuclear power plant construction sites

让AI精准识别核电建设现场行为是核心挑战。针对现有数据库的不足，团队进行了大量细致的样本采集与数据标注工作，以区分不同作业场景的细微差别，提升隐患识别的针对性与适配性。同时，为确保便携性，团队经过反复优化，将“哨兵”设备的重量成功控制在七公斤左右。

Making AI accurately identify behaviors at nuclear power plant construction sites is a core challenge. To address the shortcomings of the existing database, the team conducted extensive and meticulous sample collection and data annotation work to distinguish the subtle differences in various operation scenarios, thereby enhancing the specificity and adaptability of hazard identification. At the same time, to ensure portability, the team repeatedly optimized and successfully controlled the weight of the "Sentinel" device to be around seven kilograms.

向新而行：构建“三位一体”核建安全智能监控体系

Moving towards the future: Establishing a "trinity" nuclear security intelligent monitoring system

此次发布的机器人是该工作室在核电智能建造领域的又一项创新。下一步，团队将继续研发无人机智能监控装备，旨在构建“空中——地面——定点”三位一体的安全智能监控体系，结合AI技术开发个性化解决方案，逐步实现核电作业全场景、全流程的安全智能管控。

The robot released this time is another innovation of this studio in the field of intelligent construction of nuclear power plants. In the next step, the team will continue to develop intelligent monitoring equipment for unmanned aircraft, aiming to build a three-in-one safety intelligent monitoring system of "air - ground - fixed point", combine AI technology to develop personalized solutions, and gradually achieve safe and intelligent control of all scenarios and entire processes of nuclear power operations.