Past　few　years　have　witnessed　the　compelling　applications　of　the　remote　sensing　satellites　in　our　daily　life,　ranging　from　the　weather　forecast　to　military　surveillance.　Due　to　the　computation　and　power　constraints,　the　LEO　satellites　have　to　download　the　remote　sensing　data　to　the　ground　stations　for　further　processing.　However,　the　long-distance　transmission　and　the　ionospheric　interference　is　problematic　for　supporting　the　latency-sensitive　remote　sensing　services,　such　as　hotspot　detection,　hotspot　tracing.　As　a　remedy,　in　this　paper,　the　space　stations　are　introduced　to　offload　the　computation　task　of　the　remote　sensing　satellites　to　reduce　the　transmission　delay.　With　the　space　station　joining　in,　a　three-tier　intelligent　remote　sensing　satellites　operation　system　is　constructed.　In　order　to　perform　well,　we　study　the　computation　resource　allocation　strategies　in　this　three-tier　system.　We　model　the　resource　management　and　pricing　problems　among　three　players　as　a　Stackelberg　game,　where　the　space　stations　act　as　the　leaders,　the　ground　stations　as　the　followers,　and　the　LEO　satellites　as　the　sub-followers.　For　searching　the　Nash　equilibrium　of　this　game,　we　apply　＇Wolf-PHC＇　algorithm　for　learning　the　optimal　resource　management　strategies.　In　addition,　some　simulation　results　are　presented　to　demonstrate　the　feasibility　and　performance　of　our　architecture　and　algorithm.　©　2019　IEEE.