Development of Comfortable and Energy-Saving Driving System Based on OBDII and YOLOv7 Neural Network

Autonomous driving has been a very hot topic in recent years. In a state of unmanned control, the vehicle itself needs to have the ability to instantly sense changes in the surrounding environment, such as the distance to the car in front or pedestrians, speed limit signs and traffic lights. Identification, etc., and the recent emphasis on carbon emissions and carbon trading in various countries has made energy-saving driving particularly important in the field of autonomous driving. Therefore, this study uses YOLOv7 (You Only Look Once v7) visual detection to detect vehicles and pedestrians, and combined with The vehicle speed information received in OBDII is used to determine how the car owner should react in the current situation, such as: lightly braking, someone in front, speeding, etc., and suggestions are made through voice expression, which can be effective It avoids the problem of distracted driving while looking at the interface, and decelerates early when detecting red lights or pedestrians. This can effectively reduce fuel consumption and improve overall comfort. All information will be transmitted through the embedded development version Nvidia Jetson Xavier NX is used for calculation, and the test set uses some Taiwan standard road signs for verification. The experimental results show that the YOLOv7 model still has good generalization ability even when the signal types in different countries are slightly different, mAP (mean Average Precision) can reach more than 90.1%, and OBDII can also receive vehicle speed information in a timely manner.