As　a　distributed　parameter　system,　traffic　flow　model　of　freeway　traffic　is　determined　by　the　traffic　state　on-road　and　boundary　flows　form　on-ramp　or　off-ramp　sections.　The　existing　studies　for　traffic　estimation　mainly　focus　on　the　traffic　parameter,　namely　density　(or　vehicles)　of　mainline　traffic.　In　this　paper,　Kalman　filtering　for　simultaneous　traffic　counts　and　off-ramp　flow　estimation　is　proposed　with　the　linearization　of　the　speed　density　observation　equation.　The　state-space　model　is　formulated　by　using　a　Markov　chain　to　describe　the　vehicles＇　lane-change　movements.　Numerical　studies　are　carried　out　to　investigate　the　performance　of　the　developed　approach.