Confidence-Feature Fusion: A Novel Method for Fog Density Estimation in Object Detection Systems

• DOI: 10.3390/electronics14020219
• Publication Date: 2025-01-07T10:06:34Z
• Authors: Zhiyi Li; Songtao Zhang; Zihan Fu; Fanlei Meng; Lijuan Zhang
• Keywords: Visibility; Robustness; Density estimation

Foggy weather poses significant challenges to outdoor computer vision tasks, such as object detection, by degrading image quality and reducing algorithm reliability. In this paper, we present a novel model for estimating fog density in scenes, aiming enhance detection performance under varying foggy conditions. Using support vector machine (SVM) classification framework, the proposed categorizes unknown images into distinct levels based on both global local fog-relevant features. Key features entropy, contrast, dark channel information are extracted quantify effects of clarity visibility. Moreover, introduce an innovative region selection method tailored without detectable objects, ensuring robust feature extraction. Evaluation synthetic datasets with densities demonstrates accuracy 85.8%, surpassing existing methods terms correlation coefficients robustness. Beyond accurate estimation, approach provides valuable insights impact contributing safer navigation environments.