Research on Vehicle Abnormal Behavior Detection Algorithm Based on Deep Learning

This paper proposes an innovative algorithm combining GCN and self-Attention mechanism for vehicle abnormal behavior detection. The algorithm first uses GCN to construct a graph structure of vehicles and road environments, which can effectively capture the complex spatial relationship between vehicles and between vehicles and road environments, thereby providing richer spatiotemporal features for abnormal behavior detection. Then, a self-Attention mechanism is introduced, which is used to adaptively select the key characteristics of the time series data, which can improve the forecasting capability and precision. In this paper, a lot of experiments are carried out on the data set of public transport behavior. The results indicate that the precision of this method is 99%, which is better than the conventional method by 16%. In addition, the proposed model also performs well in F1 value, precision and recall rate, reaching 0.94, 98.2% and 99.1% respectively, showing its high efficiency and reliability in abnormal behavior detection. Compared with the conventional algorithm, this algorithm is more efficient in computation and real time, and the detection delay is less than 150 ms, so it can satisfy the requirement of real time and precision.