Vehicle driving scene discrimination method for automatic driving task

The invention discloses a vehicle driving scene discrimination method for an automatic driving task. The method comprises the following steps: acquiring a vehicle driving image; inputting the vehicle driving image into an ASR-Net module, wherein the ASR-Net module is used for identifying a driving scene of the vehicle in the vehicle driving image; the identified driving scene is input into an LD-SR module; and sending an identification result to the client. According to the scheme provided by the invention, the driving environment of the vehicle can be more accurately judged by the sensing layer through accurate judgment on the typical vehicle driving scene, and the decision-making layer can efficiently make different driving decisions according to different environment characteristics, so that obstacles are reasonably avoided, and safe driving of the automatic driving vehicle is ensured. Meanwhile, under the condition that a driver exists, accurate vehicle scene judgment can also assist the driver in control