As　air　pollution　problems　continue　to　grow,　accurate　detection　and　timely　warning　of　particulate　matter　(PM)　are　becoming　increasingly　important.　Traditional　methods　rely　on　specialized　equipment　and　are　not　suitable　for　real-time　monitoring.　In　comparison,　machine　perception　and　learning-based　methods　have　technological　advantages　such　as　real-time　monitoring　and　high　accuracy.　Therefore,　a　detailed　review　of　intelligent　PM　detection,　identification　and　early　warning　methods　based　on　machine　perception　and　learning　in　recent　years　was　presented.　First,　an　introduction　to　the　standards　and　sources　of　PM　was　presented.　Then,　various　methods　from　the　detection,　identification,　and　warning　fields　were　comprehensively　summarized,　highlighting　the　characteristics　and　performance　differences　among　them.　Methods　based　on　machine　learning　and　deep　learning　have　made　significant　progress　in　each　of　these　research　areas.　Finally,　the　paper　concluded　by　summarizing　the　main　content　and　suggesting　the　current　challenge　in　the　field　and　future　research　direction.　Future　research　effort　should　continue　to　prioritize　technological　innovation　and　data　quality　to　achieve　improved　air　quality　monitoring　and　management.　©　2024　Beijing　University　of　Technology.　All　rights　reserved.