Due　to　high-population　density,　frequent　close　contact,　possible　poor　ventilation,　university　classrooms　are　vulnerable　for　transmission　of　respiratory　infectious　diseases.　Close　contact　and　long-range　airborne　are　possibly　main　routes　for　SARS-CoV-2　transmission.　In　this　study,　taking　a　university　classroom　in　Beijing　for　example,　close　contact　behaviors　of　students　were　collected　through　a　depth-detection　device,　which　could　detect　depth　to　each　pixel　of　the　image,　based　on　semi-supervised　learning.　Finally,　＞23　h　of　video　data　were　obtained.　Using　Computational　Fluid　Dynamics,　the　relationship　between　viral　exposure　and　close　contact　behaviors　(e.g.　interpersonal　distance,　relative　facial　orientations,　and　relative　positions)　was　established.　A　multi-route　transmission　model　(short-range　airborne,　mucous　deposition,　and　long-range　airborne)　of　infectious　diseases　considering　real　close　contact　behaviors　was　developed.　In　the　case　of　Omicron,　the　risk　of　infection　in　university　classrooms　and　the　efficacy　of　different　interventions　were　assessed　based　on　dose-response　model.　The　average　interpersonal　distance　in　university　classrooms　is　0.9　m　(95　%　CI,　0.5　m-1.4　m),　with　the　highest　proportion　of　face-to-back　contact　at　87.0　%.　The　risk　of　infection　of　susceptible　students　per　45-min　lesson　was　1　%.　The　relative　contributions　of　short-range　airborne　and　long-range　airborne　transmission　were　40.5　%　and　59.5　%,　respectively,　and　the　mucous　deposition　was　basically　negligible.　When　all　students　are　wearing　N95　respirators,　the　infection　risk　could　be　reduced　by　96　%,　the　relative　contribution　of　long-range　airborne　transmission　increases　to　95.6　%.　When　the　fresh　air　per　capita　in　the　classroom　is　24　m3/h/person,　the　virus　exposure　could　be