Risk level assessment method for urban intersections based on improved principal component analysis

The traffic environment of urban intersections has time-varying non cooperative characteristics and multi-source disturbance coupling effects. Therefore, a risk level assessment method for urban intersections based on improved principal component analysis is proposed. Firstly, a dynamic reference point model and an outer rectangular prism are used to characterize the motion state of the target vehicle. The velocity is estimated through position difference, and the heading angle is obtained by fitting a quadratic curve. Secondly, improve principal component analysis to preserve indicator variation information through mean processing, and enhance feature extraction capability through logarithmic centralization. Finally, based on the entropy weighting method and objective weighting of the main components, five risk levels are classified according to the mean and standard deviation of the comprehensive evaluation value. The experimental results indicate that the F1 Score mean of the proposed method is 0.925, and the Spearman rank correlation coefficient remains stable at 0.95.