Intelligent transportation vehicle detection and tracking based on improved YOLOv8-LSTM fusion model

Intelligent transportation vehicle detection and tracking is a key core technology for intelligent road traffic supervision and speeding risk control, and has significant engineering application value. In response to the problems of low accuracy of multi-scale vehicle detection in complex road conditions and large errors in traditional speed measurement methods, research on intelligent transportation vehicle detection and tracking based on the improved YOLOv8 and long short-term memory network hybrid model has been carried out. First, YOLOv8 is optimized in three aspects: cross-scale feature fusion, multi-dimensional attention embedding, and switchable dilated convolution. Then, a gated recurrent unit-long short-term memory hybrid temporal network is constructed to improve the trajectory prediction module of the byte trajectory tracking algorithm. Finally, the detected coordinates and motion trajectory are fused to automatically identify speeding behavior. The results show that the average accuracy of the improved object detection algorithm is 91.78%, and the single-frame inference takes only 13.51ms. The average accuracy of multi-object tracking and the success rate of identity preservation of the improved byte trajectory tracking algorithm are 72.37% and 86.54% respectively. The maximum speed measurement error of the proposed overspeed detection method is only 2.16km/h, and the mean root mean square error of multiple vehicle types is as low as 1.85km/h. The designed method has good application effects and can provide technical support for precise vehicle detection, stable tracking and intelligent speeding control in complex traffic scenarios.