In　the　industrial　production　line,　the　operation　of　packing　small　packages　that　are　densely　arranged　in　vertical　rows　and　move　quickly　is　still　mainly　done　manually.　In　this　situation,　it　is　a　typical　and　difficult　problem　for　robot　to　put　small　packages　with　various　quantity　and　arrangement　requirements　into　large　packages　in　real　time.　In　order　to　solve　this　problem,　this　paper　proposes　a　real-time　detection　method　based　on　the　improved　YOLOv5　deep　neural　network　for　identifying　the　precise　quantity　and　abnormality　of　dense　vertical　bagged　small　packages　on　the　assembly　line.　This　method　can　quickly　detect　the　position,　category,　quantity,　and　rotation　angle　of　small　packages.　Combining　depth　information　with　bounding　box　rotation　angle　information,　it　can　timely　detect　abnormal　placement　of　small　packages　on　the　assembly　line.　Then　the　misjudgment　of　the　number　of　small　packages　caused　by　blurred　images　is　corrected　by　amending　the　width　of　the　object　with　low　reliability.　This　method　effectively　identifies　abnormal　situations　in　the　industrial　production　line　and　provides　an　effective　solution,　which　is　crucial　for　robots　to　autonomously　complete　packaging　tasks　in　real-time　according　to　various　quantities　and　arrangement　requirements.　©　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.　2024.