In　order　to　enable　users　to　quickly　obtain　the　most　informative　frame　in　the　video,　we　proposed　a　sports　video　summarization　technology　that　comprehensively　applies　the　global　optical　flow　and　local　optical　flow　characteristic.　On　the　premise　of　capturing　visual　differences,　we　consider　the　simple　semantic　features　of　sports　video.　In　this　method,　shot　segmentation　is　done　based　on　color　features　and　some　denoise　process　is　added.　Then　we　used　spline　curve　to　obtain　the　global　optical　flow　and　local　optical　flow　of　each　frame,　and　obtain　the　camera　motion　function.　The　significant　difference　of　camera　motion　function　in　one　shot　between　frames　is　used　to　extract　shot　key　frames.　Then,　we　introduce　local　optical　flow　to　solve　the　problem　of　redundant　frames.　©　2022　IEEE.