Accurate vehicle density estimation is crucial for effective traffic management and urban planning. This paper introduces a novel adaptive vehicle density estimation framework that integrates an improved YOLOv8 object detection model with Kernel Density Estimation (KDE). The proposed system aims to provide high-precision, real-time vehicle density analysis to enhance intelligent transportation systems. The core of our approach is the improved YOLOv8 model, specifically optimized for vehicle detection. Enhancements include the incorporation of spatial attention mechanisms and fine-tuning on extensive vehicle datasets, leading to superior detection accuracy and reduced false positives. This robust detection capability is critical for accurately identifying and localizing vehicles in diverse traffic conditions. Once vehicles are detected, their positions are mapped onto a grid covering the region of interest. KDE is then applied to these positions to generate a smooth and continuous vehicle density map. This statistical method effectively captures the spatial distribution of vehicles, offering a more precise density estimation compared to traditional grid-based methods. The resulting density map is crucial for understanding traffic flow and congestion patterns. The system is designed to be adaptive, continuously learning from new data to refine its parameters and improve estimation accuracy. By deploying the algorithm on edge computing devices, the system achieves real-time processing capabilities, enabling immediate responses to changing traffic conditions. Experimental results on real-world traffic datasets demonstrate the proposed framework’s effectiveness, showcasing significant improvements in both accuracy and processing speed over existing methods. The integration of an improved YOLOv8 model with KDE not only enhances vehicle detection and density estimation but also provides a scalable and efficient solution for modern traffic management systems.