主页 > 国内 >

尼基塔第三季

Chinese-made infrared telescope with domestic detectors achieves astronomical imaging, reducing foreign reliance_我的网站

怒火攻心

一 |

Ground-based infrared telescope installed at Saishiteng Mountain, Lenghu township, Northwest China's Qinghai Province Photo: Screenshot from gdstc.gd.gov.cn
Chinese scientists on Wednesday released the first complete set of observational images captured by China's new-generation ground-based infrared telescope, filling a gap in the country's near-infrared astronomical observation capabilities and reducing reliance on foreign infrared detector technology, China News Service reported on Wednesday.
The 80-centimeter near-infrared telescope was installed at Saishiteng Mountain, Lenghu township, Northwest China's Qinghai Province in October 2024. Jointly developed by Sun Yat-sen University and several partner institutions, the telescope is China's first to achieve scientific imaging in the 1.2-micrometer J band and 2.2-micrometer K band using domestically developed scientific-grade InGaAs and HgCdTe detectors.
The telescope enables unattended remote operations, and can lock onto any target in the sky within 15 seconds. Its infrared capability allows it to see through interstellar dust, enabling it to observe cold and distant celestial objects that are difficult for optical telescopes to capture.
According to Ma Bin, a professor at the School of Physics and Astronomy, Sun Yat-sen University who led the research team that released the infrared telescope's observational images, all three of the telescope's core astronomical observation capabilities were fully validated during operation.
It rapidly responded to and tracked multiple gamma-ray bursts and supernova events, recording the evolution of their infrared afterglows following the explosions. It continuously monitored brown dwarfs, precisely capturing periodic variations in their atmospheric brightness. Using image-stacking technology, it also observed active galaxies billions of light-years away.
Since its deployment in Lenghu, the system has supported infrared observations for more than 10 Chinese research institutions, accumulating valuable experience for the construction of future large-scale infrared telescopes.
Lenghu township, where the world-class astronomical observation base takes shape, is hailed as "the most Mars-like place on the Earth." The astronomical observation base was said to be an astronomical observation site comparable to the top ones in the world by a Nature article published on August 18, 2021, according to the Chinese Academy of Sciences.
Once prospered in 1950s due to its rich petroleum resources, Lenghu is now becoming a premier astronomical observation site and Mars-themed tourist destination, according to Xinhua News Agency.
Global Times
。 “画面显示现场搭设有金属管脚手架，脚手架上有一名佩戴红色安全帽的人员，安全防护多处缺失。脚手架下方地面堆放有钢管、扣件等物料，左下角可见两名佩戴白色安全帽的人员站立，背景可见混凝土墙体结构。”这是在杭州一工地抓拍的现场画面，但就在快门按下的瞬间，一个看不见的“安全监管员”完成了对这张照片的“审阅”——脚手架搭设不合规、高处作业未系跨安全带、高处作业吊篮不规范，三条隐患同步弹出，整改指令随即下发。

二 | 这位“安全监管员”，就是浙江蓝宸数联科技有限公司打造的AI建安智检平台。“过去，工地安全巡检主要靠人，特别是对于一些隐患识别，还得看老师傅的经验。但人总有局限，不能时时刻刻盯着，像外立面、深基坑、高边坡这些‘难以一眼看清的角落’，往往是事故的高发区。”蓝宸数联董事长魏自金曾在城建领域深耕十余年，对此深有感触。

三 | 而如今人工智能的应用，让巡检效率有了质的提升。

四 | AI建安智检平台将工地划分为巡检网格，联动球机、无人机、机器狗、执法记录仪等设备，开展空天地一体主动巡检，依托自主研发的AI智能体驱动，可以实现从自动抓拍、AI识别、分级派单到整改复核、线上闭环的全流程数字化管控。目前，平台已覆盖14大类、46大重点隐患类型、140余类安全隐患，语义描述准确率达98.21%，平均AI隐患识别准确率不低于86%，处于行业领先水平。

五 | 这些成效斐然的数字背后，其实有本不断更新的“教科书”在支撑——工地安全隐患高质量数据集，它让AI不仅能“看到”，更能“看懂”。“数据集不是数据的简单搜集，而是带有成效逻辑的完整链条，包括采集、标注、分析、闭环四个环节。

六 | ”魏自金解释，以基坑为例，拍照只是第一步，接下来要从土堆的高度、坡度、湿度、行业规范等多个维度进行综合分析，判断是否存在滑坡风险。发现隐患后还要派单、整改、复查，直到确认闭环。而这精准研判、举一反三的能力，来自日复一日的高强度训练。记者了解到，蓝宸数联的数据来源非常广泛，监管数据、项目数据、施工企业数据都在不断“喂”给AI建安大模型。不仅如此，团队还会主动“投喂”一些错误数据，倒逼模型练就一双“火眼金睛”。“同样的隐患，我们识别得更快更准，靠的就是数据集更全面、训练更充分。”目前，蓝宸数联高质量行业数据集已入选杭州市第一批城乡建设领域“人工智能+”高质量数据集建设名单。值得关注的是，蓝宸数联还将“AI建安智检系统”与“工地安全隐患数据集”挂牌杭州数据交易所，推动企业数据资产化、赋能智慧城市建设高质量发展。深耕城建领域智能化转型，是蓝宸数联立足行业痛点、把握数字风口的精准布局，早在2020年，魏自金便坚定投身城建AI视觉研发，在他看来，城建领域项目体量庞大、施工场景繁杂、从业人员众多、工况变化频繁，是打磨人工智能算法、锤炼AI技术能力与业务综合能力的“绝佳试炼场”。“能啃下这块‘硬骨头’，后续再将技术、模型、体系复制落地到其他领域，便能事半功倍。”魏自金表示，蓝宸数联即将把业务拓展至交通、能源、水务等板块，先攻克通用类隐患，再逐步深入各行业的专属场景。在他的畅想中，未来不同领域的智能体将打破行业壁垒，相互配合、协同作业。