The invention provides a vehicle travel trajectory identification method based on bayonet data, which adopts a spectral clustering method to realize automatic separation of travel chain from the time-space characteristic of bayonet data, thereby identifying the starting point, the end point and the passing point of a single travel trajectory. The method particualy includes generating a complete travel chain of the vehicle on the day; establishing a similarity matrix S; calculating an adjacency matrix W and a degree matrix D; constructing a Laplace matrix L and further determining a characteristic matrix F; using the DBSCASN clustering algorithm to perform cluster analysis, and obtaining the clustering results; determining the cluster partition result of trajectory points to generate sub-trajectories. This method takes into account both the time and space attributes of vehicles, and is more comprehensive than the method based on travel time to partition the trajectory. No preset threshold is needed, and the vehicle trajectory recognition is completely driven by the measured data, which improves the accuracy of trajectory recognition. The analysis results can be used in vehicle travel behavior recognition, travel characteristics analysis, traffic flow OD estimation and so on.