Intelligent traffic flow analysis, as an emerging technology, provides effective solutions for urban traffic management through big data analysis and artificial intelligence algorithms. The Yolov4 algorithm is used to incorporate the Deep SORT multi-target tracking algorithm and multi-scale feature fusion detection model in this study. In addition, the improved Yolov4 algorithm model with multi-scale feature fusion based on the tracking algorithm is formed by combining the CLAHE algorithm in the input module of the original Yolov4 algorithm and by replacing the backbone network and optimising the neck network. The model presented in this paper has a better training effect on the vehicle flow evaluation index. The tracking algorithm employed in this paper is now 0.268 more accurate than the KCF algorithm, and it takes only 5.65ms to process the video images. In addition, the mAP, Rank-1, and Rank-5 of this paper’s model for monitoring small vehicles are 82.81%, 86.79%, and 95.43% respectively, which are better than the base model. The improved Yolov4 traffic monitoring algorithm proposed in this paper balances counting accuracy and running speed, and has superior vehicle traffic monitoring performance than the comparison algorithm.