Research On The Extraction Method Of Driver’s Horizon Attention Region Based On Vehicle Maneuvering Behavior

Recently,intelligent driving systems in preventing the occurrence of traffic accidents are more and more widely used,but it also has some problems and challenges,for example,the current intelligent driving systems are driven by image data.Not only do they have explanatory differences with drivers in understanding the external target information,but they also fail to make correct decisions through traffic scene analysis as human drivers do.Therefore,our paper studies the method of extracting the attention area in the driver’s visual field,which is based on the visual attention mechanism in the intelligent driving system,combined with the driver’s real manipulation behavior and deep learning technology.This method is designed to predict the driver’s attention points to extract the key information affecting the driver’s operation.The proposal of this method lay a theoretical foundation and provide corresponding technical support for the future development of intelligent driving vehicles,the traceability of traffic accident causes and the auxiliary driving systems.The main contents of our paper are as follows:(1)The driver’s attention area identification method based on deep network feature visualization,First,this paper builds a driving behavior information network,which can predict the steering wheel angle and vehicle speed through the current traffic scene,so as to establish a one-to-one correspondence between the driver’s field of view information and the vehicle manipulation behavior.Then,the weighted extraction method of driver’s attention area feature map proposed in this paper can determine the attention area that enables the driver to change the steering wheel angle and vehicle speed in the current traffic scene,and make the driver attention area data set required in this article based on the displayed attention area.The establishment of the driver’s attention area data set provides data support for the realization of the driver’s attention area extraction method based on vehicle manipulation behavior proposed In our paper.(2)The prediction method of driver’s attention point based on attention model.According to the human visual attention mechanism and combined with deep learning technology,our paper designs the driver’s visual attention network,which realizes the high-precision positioning and recognition of visual targets by predicting the driver’s attention points.Compared with traditional methods,the driver’s visual attention network has higher target search efficiency and more accurate recognition accuracy.The construction of the driver’s visual attention network provides algorithmic support for the realization of the driver’s attention region extraction method based on vehicle manipulation behavior proposed in our paper.(3)The experimental verification based on driver’s attention area data set.The driver attention point prediction method based on the attention model is trained on the driver attention area dataset produced in our paper.After the training is completed,it is verified from multiple traffic scenes on the test set.The experimental results demonstrate the effectiveness of the proposed method based on vehicle manipulation behavior.