Tunnel　LED　lighting　system　is　an　important　infrastructure　to　ensure　the　safety　and　comfort　of　driving　in　highway　tunnels.　There　had　been　a　series　of　researches　about　analyzing　these　LED　lamps　working　state　by　detection　of　missing　LED　lamps　in　highway　tunnels.　In　this　paper,　an　automatic　tracking　method　for　the　salient　line　segments　in　highway　tunnels,　the　left　and　right　critical　boundaries　between　the　top　dark　area　of　the　tunnel　and　the　white　side　wall,　is　proposed　for　vehicle-mounted　video　processing.　This　method　includes　the　following　two　aspects.　Firstly,　a　three-phase　algorithm　was　proposed　to　detect　the　salient　line　segments　by　analyzing　the　geometric　structure　of　the　tunnel　interior　environment,　after　the　line　features　were　extracted　by　the　LSD　operator.　In　this　step　of　processing,　the　set　of　parallel　line　segments　converging　on　the　vanishing　point　in　front　of　the　vehicle　and　the　front-vehicle　vanishing　point　itself　should　be　computed　as　the　intermediate　results.　Secondly,　the　visual　features　were　extracted　in　the　extended　rectangles,　which　can　be　formed　easily　by　extending　the　line　segment　outward　with　external　equal　distance　amplification　in　the　initial　frame,　and　tracked　in　continue　frame　with　more　efficiency　and　more　tolerance　to　larger　displacement.　The　salient　line　segments　extracted　in　the　initial　frame　can　be　tracked　in　the　next　frame,　using　the　corresponding　relationship　embedding　with　the　displacement　of　the　same　feature　points　in　the　two　consecutive　frames.　The　experimental　results　show　that　the　satisfactory　results　for　tracking　the　salient　line　segments　are　obtained,　and　meet　the　actual　needs　of　reducing　the　disturbance　of　vanishing　point　estimation　in　continuous　on-board　video　and　improving　the　recovery　performance　of　geometric　information　in　road　tunnels.　©　2023　Technical　Committee　on　Control　Theory,　Chinese　Association　of　Automation.