原文標題： Planet Successfully Runs AI in Space 來源： https://myppt.cc/e1uCtJ 內文： April 7, 2026 Using NVIDIA Jetson, Planet achieves AI-powered object detection on its Pelican satellite SAN FRANCISCO--(BUSINESS WIRE)-- Following the announcement of a strategic initiative with NVIDIA, Planet Labs PBC (NYSE: PL) announced a landmark technical achievement: the successful deployment and execution of AI-driven object detection directly onboard its Pelican-4 satellite, paving the way toward on-orbit compute for rapid insights, a capability Planet calls Planetary Intelligence. http://i.imgur.com/it0WWLz.jpg Imagery captured by Pelican-4 on March 25, 2026 over Alice Springs, Australia, demonstrating the first successful deployment and execution of AI-driven object detection directly onboard Planet spacecraft. This first detection has an 80% detection accuracy on raw imagery. Work is underway to improve model precision and recall as Planet refines its models. On March 25th, 500km over Alice Springs, Australia, Planet’s Pelican-4 captured an image of an airport, then successfully utilized its onboard NVIDIA Jetson Orin module to run an AI model to detect airplanes in moments. This represents one of the first times an Earth imaging satellite has moved beyond simple data capture to onboard AI inference and analysis. “This success is a glimpse into the future of what we call Planetary Intelligence at scale," said Kiruthika Devaraj, VP of Avionics & Spacecraft Technology. "By running AI at the edge on the NVIDIA Jetson platform, we can help reduce the time between 'seeing' a change on Earth and a customer 'acting' on it, while simultaneously minimizing downlink latency and cost. This shift toward integrated AI at the edge is a technological leap that can help differentiate solutions like Planet’s Global Monitoring Service (GMS), providing valuable insights for our customers and enabling rapid resp “This step with NVIDIA can help speed the pace of insight, reducing the time to potential answers from hours to minutes. This can be the critical difference-maker for our customers from disaster response to security and beyond,” said Will Marshall, Planet CEO and Co-Founder. “Bigger picture: this is an exciting milestone towards delivering Planetary Intelligence. We’re moving AI from the internet into the physical realm, effectively connecting the ‘eyes’ of our satellites with an onboard ‘brain’ to create a nervous system for the planet.” This breakthrough can help progress Planet’s Pelican and forthcoming Owl constellations into a near-real-time intelligence network. Leveraging NVIDIA Jetson and high-speed inter-satellite links, Planet is working to effectively close the latency gap. Rather than waiting for data to download and then be analyzed on the ground, customers could receive actionable insights within minutes of capture. The end-to-end process, spanning initial data generation, deep-net object detection, and full geo-rectification, is designed to occur entirely in orbit. By producing GeoTIFF and GeoJSON insights within isolated Docker containers in space, Planet is striving to be able to generate intelligence about events in minutes. While the onboard models are in their early stages and will continue to be refined, the successful integration of AI applications in space paves the way for a new era of flexibility for satellite constellations. By placing this compute power on the satellite itself, Planet aims to unlock the future of Earth intelligence at a planetary scale with near real-time insights. 機翻： Planet公司利用NVIDIA Jetson技術，在其Pelican衛星上實現了人工智慧驅動的目標偵測。 舊金山--(BUSINESS WIRE)-- 繼宣布與 NVIDIA 達成戰略合作之後，Planet Labs PBC (NYSE: PL) 宣布了一項里程碑式的技術成就：在其 Pelican-4 衛星上成功部署和執行了 AI 驅動的目標，為在軌計算以實現快速洞察鋪平了道路，Planet 將這項能力稱為行星智能。 2026年3月25日，Pelican-4衛星在澳洲愛麗絲泉上空拍攝的影像，展示了Planet太空船首次成功部署並執行人工智慧驅動的目標偵測。此次檢測在原始影像上的準確率達到了80%。 Planet公司正在不斷改進其模型，以提高模型的精確度和召回率。 3月25日，在澳洲愛麗絲泉上空500公里處，Planet公司的Pelican-4衛星拍攝到一張機場影像，隨後成功利用其搭載的NVIDIA Jetson Orin模組運行人工智慧模型，瞬間辨識出飛機。這標誌著地球成像衛星首次超越簡單的數據收集，實現了機載人工智慧推理和分析。 「這項成功讓我們得以一窺我們所說的大規模行星智慧的未來，」航空電子與太空船技術副總裁Kiruthika Devaraj表示。 「透過在NVIDIA Jetson平台上運行邊緣人工智慧，我們可以幫助縮短從『觀測』到客戶『採取行動』的時間，同時最大限度地降低下行延遲和成本。這種向邊緣集成人工智能的轉變是一項技術飛躍，有助於Planet全球監測服務（GMS）等解決方案脫穎而出，為我們的客戶提供寶貴的洞察，並在最關鍵的時刻實現快速響應。」 「與英偉達的這項合作將有助於加快洞察速度，將獲得潛在答案的時間從數小時縮短至數分鐘。這對我們的客戶而言至關重要，從災難響應到安全保障，乃至更廣泛的領域，都將產生決定性的影響，」Planet 首席執行官兼聯合創始人威爾·馬歇爾 (Will Marshall) 表示。 “從更宏觀的角度來看，這是實現行星智能道路上一個激動人心的里程碑。我們將人工智能從互聯網帶入物理世界，有效地將衛星的‘眼睛’與星載‘大腦’連接起來，從而為地球構建一個神經系統。” 這項突破有助於推動 Planet 的 Pelican 和即將推出的 Owl 星座發展成為近實時情報網。 Planet 利用 NVIDIA Jetson 和高速星間鏈路，致力於有效縮小延遲差距。客戶無需等待資料下載並在地面進行分析，即可在資料收集後幾分鐘內獲得可操作的洞察。從初始資料產生、深度網路目標偵測到完整的地理校正，整個端對端流程均在軌道上完成。透過在太空中隔離的 Docker 容器內產生 GeoTIFF 和 GeoJSON 格式的洞察數據，Planet 力求在幾分鐘內產生事件情報。 儘管機載模型尚處於早期階段，並將持續改進，但人工智慧應用在太空的成功集成，為衛星星座開啟了靈活運行的新時代。透過將這種運算能力部署在衛星本身，Planet 旨在以近乎即時的洞察，在行星級尺度上開啟地球智慧的未來。 心得： PL發布消息在太空中成功運行AI 在最新一代的Pelican-4衛星上 使用NVIDIA Jetson平台上運行AI 成功辨識出機場的飛機 目前測試準確率達80% 未來圖像直接在太空辨識 將結果傳回地球 ----- Sent from JPTT on my Google Pixel 7 Pro. -- ※ 發信站: 批踢踢實業坊(ptt.cc), 來自: 49.216.133.107 (臺灣) ※ 文章網址: https://webptt.cc/bbs/Stock/M.1775612572.A.10C.html