With　the　development　of　the　Internet　and　video　editing　technologies,　there　are　a　large　number　of　near-duplicate　videos　on　the　Internet　today.　This　can　cause　a　lot　of　trouble　in　video　content　retrieval　and　copyright　protection.　It　is　time-consuming　to　manually　classify　a　large　number　of　near-duplicate　videos　A　method　is　proposed　here　to　automatically　recognize　and　classify　near-duplicate　videos　based　on　temporal　and　spatial　key　points.　This　method　extracts　key　frame,　proportion　of　video　segment,　average　gray　level　and　average　segmentation　ratio　as　the　video　key　information,　which　is　used　to　identify　the　approximate　video.　For　near-duplicate　video,　this　method　has　a　good　effect.　©　2020,　Springer　Nature　Singapore　Pte　Ltd.