Automatic parking method and system based on neural network

The invention relates to the technical field of automatic parking planning of automatic driving, in particular to an automatic parking method and system based on a neural network, and is used for improving the success rate and accuracy of automatic parking. The method comprises the following steps: acquiring a panoramic all-around view image around a vehicle through a camera; obtaining a parking position through a deep neural network algorithm based on the panoramic looking-around image; planning a parking path based on the parking position and the vehicle state; and a vehicle control system is started to execute parking, whether parking is finished or not is judged every a certain time threshold value, if yes, detection is finished, and if not, parking continues to be executed. The success rate and the accuracy of lineation parking space detection are improved by using a parking detection algorithm of a deep neural network, path planning is performed by using the neural network and referring to driver experi