Vehicle abnormal driving behavior identification method based on small sample learning

The invention discloses a vehicle abnormal driving behavior identification method based on small sample learning, and relates to the technical field of environment perception of safe driving. The method comprises the following steps: acquiring driving behavior data of a target vehicle; the driving behavior data comprises time, a vehicle number, a vehicle speed, a vehicle acceleration and a vehicle position; preprocessing the driving behavior data to obtain preprocessed data; according to the preprocessed data and an abnormal driving behavior recognition model, determining whether the target vehicle has an abnormal driving behavior or not and an abnormal driving category when the abnormal driving behavior exists; the abnormal driving behavior recognition model is constructed according to a twin neural network; the twin neural network comprises an input layer, a basic network layer and a Lambda layer which are connected in sequence. The method effectively improves the recognition accuracy of the abnormal drivin