Adaptability of collision warning for motor vehicle drivers: can consciousness recognition improve the warning accuracy?

Aiming at the problems of poor risk identification ability and high false alarm rate in current vehicle collision warning technology, the self-adaptability of vehicle driver collision warning is proposed: Can consciousness recognition improve the warning accuracy? Construct an adaptive collision warning model, and quantify the impact of each vehicle's collision with an adaptive weight distribution method. The collision speed, the collision safety factor and the lateral offset related to the collision process are used as the weight distribution index to calibrate the model parameters. The maximum similarity recursive algorithm is used to estimate the characteristics of the current vehicle driving state. The maximum braking deceleration threshold is used to determine the degree of collision risk between the own vehicle and the following vehicle. The collision risk levels of the two workshops are divided into three levels: safety, critical value, and danger. The signal detection principle is adopted to modularize the initial information of the collision warning system, and the driving state assessment threshold adaptation category of the collision warning system is alarmed according to the actual situation. Experiments have verified that when the vehicle collision risk level is the critical value, the DR value of the model in this paper is 0.06 larger than the value of the NCM model, and it takes 3s-6s more time for the driver to take anti-collision and obstacle avoidance measures. Compared with the area under the ROC curve of the NCM model, the local algorithm has high warning accuracy and low false alarm rate.