Detection Algorithm of Takeover Behavior of Automatic Vehicles’ Drivers Based on Deep Learning

In the process of autonomous driving, the frequency of drivers' takeovers can be used to evaluate the performance of autonomous vehicles, but there is no appropriate method to collect data. This study developed an objective and convenient detection algorithm to measure the level of driverless vehicles by counting the times and duration when the driver takes over the steering wheel in a certain mileage. First of all, the position information of the hands and steering wheel was obtained through object detection algorithm based on CornerNetSaccade. Then a compensation and filtering mechanism was designed to reduce missed or incorrect detections. Finally, the correlation between consecutive frames was used to record the times and duration of drivers' takeovers. Experimental results showed that the proposed algorithm can meet the requirements of recording the degree of manual intervention of automatic vehicles, providing an effective data acquisition method for the classification and evaluation of unmanned vehicles.