In　this　research,　computer　vision　in　image　analysis　is　studied,　and　a　model　is　established　to　solve　the　problem　of　unclear　surveillance　cameras.　Learning　the　basic　knowledge　of　image　processing,　using　image　filter　and　image　addition　and　subtraction　multiplication,　convolution,　Prewitt　operator,　Sobel　operator　and　other　technologies　to　process　and　analyze　the　image,　finally　solved　the　problem　of　recognition　efficiency　caused　by　poor　photo　quality　in　the　traffic　system.　In　the　study,　a　matrix　is　obtained　by　scanning　the　image　first,　and　then　Sobel　operator　is　used　to　detect　the　edge　of　the　image,　and　the　digital　edge　of　the　license　plate　number　is　bolded,　so　as　to　scan　the　number　on　the　license　plate　more　clearly.　Another　model　is　also　built,　using　matrix　subtraction　technology　to　judge　whether　the　license　plate　number　information　matches　the　required.　These　models　provide　two　effective　methods　and　bases　for　better　identification　of　license　plate　number.　In　conclusion,　these　methods　can　help　the　traffic　system　to　screen　restricted　vehicles　and　trace　the　suspect　vehicle　which　can　facilitate　the　society.　©　2023　IEEE.