Application of Improved ViBe Algorithm in Vehicle Detection
Dongming Xu, Guogang Han · 2021 4th International Conference on Artificial Intelligence and Pattern Recognition · 2021
Background modeling technology, as the mainstream algorithm of vehicle detection, is widely used in the field of intelligent transportation. In ViBe algorithm, the "ghost" phenomenon in the model initialization stage, the update mechanism in the background model update stage and the flicker point interference caused by leaf disturbance, camera jitter and other factors in road scenes are improved. According to the similarity of road gray distribution, a judgment strategy of pixel gray value is introduced to solve the "ghost" phenomenon caused by foreground objects in the image when ViBe algorithm is modeling the first frame. The fixed update strategy in ViBe algorithm is changed to an adaptive update strategy to better adapt to the changes of road scenes. At the same time, the scintillation point determination mechanism is introduced to solve the leaf disturbance and camera jitter problems. The algorithm was verified by the standard data set Cnt 2014, and the results show that the detection effect is better than the classical algorithm.