Considerable　errors　exist　in　speed　data　which　obtained　from　various　detectors,　and　its　accuracy　directly　affects　the　results　of　various　traffic　analysis　models.　This　research　implemented　12　groups　of　field　experiments　on　roads　of　different　classes　by　using　a　test　vehicle　with　GPS　modules.　It　attempts　to　verify　the　accuracy　discrepancy　of　travel　speed　data　collected　by　floating　car　system,　microwave　detectors,　and　loop　detectors.　Analysis　shows　that　the　average　travel　speed　errors　of　floating　cars　on　road　section　of　urban　expressways　and　arteries　are　13.6%　and　27.8%,　and　the　collected　data　and　the　actual　speed　have　the　same　distribution.　Compared　with　actual　speed,　the　errors　of　microwave　detector　on　expressways　and　loop　detectors　on　arteries　are　30%　and　56%.　Although　there　exist　a　certain　degree　of　error,　basically,　it　can　characterize　the　driving　conditions　on　the　road　section　when　the　road　section　speed　of　the　microwave　detector　is　used　to　denote　the　travel　speed.　Study　results　provide　suitable　data　source　for　different　traffic　models　that　have　different　accuracy　requirements.　Besides,　it　provides　a　theoretical　basis　for　improving　data　accuracy　and　completeness　by　using　the　redundant　information　of　multi-source　data.