Application of LiDAR Data for Deep Learning Based Near Crash Prediction at Signalized Intersection

Near crash events are often regarded as an excellent surrogate measure for traffic safety research because they include abrupt changes in vehicle kinematics that can lead to deadly accident scenarios. In this paper, we introduced machine learning and deep algorithms predicting near using LiDAR data at a signalized intersection. To predict occurrence, used essential kinematic variables such lateral longitudinal velocity, yaw, tracking status of LiDAR, etc. A hybrid model Convolutional Gated Recurrent Neural Network (CNN + GRU) was introduced, comparative performances were evaluated with multiple classification models Logistic Regression, K Nearest Neighbor, Decision Tree, Random Forest, Adaptive Boost, like Long Short-Term Memory (LSTM). As occur after sudden brake, considered average deceleration energy drop thresholds identify crashes braking time . We looked the next 3 seconds our prediction horizon. All work best 1-second horizon time. The results also reveal gathers greatest information while working flawlessly. comparison existing prediction, has 100% recall, precision, F1-score: accurately capturing all crashes. This performance outperforms previous baseline forecasting provides opportunities improving via Intelligent Transportation Systems (ITS).