Accurate　modeling　of　fluid-particle　interactions　in　geotechnical　systems,　particularly　those　involving　irregular　particles,　presents　significant　challenges　in　computational　mechanics,　necessitating　a　versatile　Eulerian-Lagrangian　framework　capable　of　handling　diverse　particle　geometries.　This　paper　presents　a　novel　point　cloud-based　semi-resolved　CFD-DEM　coupling　method　to　ensure　accurate　void　fraction　calculations　in　fluid　fields　surrounding　particles　of　arbitrary　shapes　under　arbitrary　grid　configurations.　The　method　integrates　high-precision　particle　volume　sampling　via　point　cloud　technology　with　a　Gaussian　kernel　function　for　fluid　velocity　reconstruction,　enabling　accurate　void　fraction　calculations　and　robust　flow　field　characterizations.　Numerical　tests　validate　the　method＇s　effectiveness,　demonstrating　high-quality　particle　volume　mapping　and　smooth,　continuous　void　fraction　distributions.　In　benchmark　cases　such　as　single　particle　sedimentation,　three-phase　flow　dam　breaks,　and　wave　impacts　on　rock　piles,　the　method　shows　excellent　agreement　with　experimental　data,　particularly　in　capturing　fluid-particle　interaction　forces　and　interface　dynamics.　The　computational　efficiency　is　significantly　enhanced　through　the　incorporation　of　k-d　tree　data　structures　for　range　searching.　It　is　indicated　that　the　proposed　semi-resolved　CFD-DEM　coupling　method　provides　a　new　insight　and　tool　for　accurate　simulation　of　fluid-particle　coupling　systems　in　complex　geotechnical　engineering　problems.