Driving intention recognition, as a prerequisite in the driving decision process, plays an important role to the vehicle. Aiming at the inadequacy of feature extraction using LSTM structure, a driving intent recognition model with Transformer structure is proposed. First, based on the driving simulator and eye tracker, the driver's vehicle and scan path of eye movement in the highway condition are collected. Second, Transformer is introduced to recognize the driving intention. Finally, the test set is used to verify the proposed model, compared with the LSTM network and the standard recurrent neural network. Accuracy rate is 93.78%.Precision is 93.83%, Recall rate is 92.72%.F1 score is 92.73%. The results show that the feature extraction ability of Transformer structure is better than LSTM, and it has certain applicability to identify the process of lane-change intent recognition in long sequence background. Keywords: intelligent transportation; Transformer; lane-change intent recognition;

CICTP