Video Vehicle Detection Based On Background Difference

Video-based vehicle detection is the foundation and key of Intelligent Transportation System ITS(Intelligent Transportation System),and it is also a hotspot in the field of computer vision.With the continuous research of domestic and foreign scholars,many new vehicle detection algorithms have been proposed in recent years,and the problem of vehicle shadow and background disturbance has been solved step by step.However,most of these algorithms ignore the impact of video processing on the detection effect,and the algorithm is more complex.Therefore,this paper proposes a method to improve the detection rate and real-time performance in view of the existing problems of noise and real-time difference in the video detection algorithm based on video.The following two aspects are studied:(1)Aiming at the problem that the noise in the video affects the vehicle detection rate,three kinds of denoising algorithms,such as median filter,Wiener filter and wavelet denoising,are studied in depth.The paper proposes a method to improve the detection rate of the vehicle based on sparse representation of video noise.By K-singular value decomposition(K-Singular Value Decomposition,K-SVD)algorithm is used to train the complete dictionary,decomposing the noise image sparsed on the over-complete dictionary,and then obtain a clean image.Compared with the three kinds of denoising methods.The experimental results show that the denoising method in this paper has better performance in terms of signal-to-noise ratio and vehicle detection rate.(2)In order to solve the problem of how to establish a background model with robust robustness in background difference,three kinds of classical modeling algorithms: mean method,single Gaussian background model and mixed Gaussian background model are deeply analyzed and studied,an optimization method based on hybrid Gaussian multimodal model is proposed.First,all the models obtained in the modeling process are likely to be background models,and the pixels with smaller model weights may contain real background pixels.Therefore,it is necessary to establish a more realistic background model and reduce the computational complexity of the algorithm by discarding the weighting of the model after the model matching and the comparison of the accumulated value and the threshold.Finally,the use of centroid tracking method completes multi-vehicle tracking and statistical traffic flow.The experimental results show that the proposed multi-modal modeling based on hybrid Gauss has a lower computational cost,better detection and anti-jamming capability.