Existing change detection studies have paid little attention to changes between video images, especially in severe weather such as fog and dust storms. Considering that the video images captured by the Hawk-Eye surveillance system in foggy weather are noisy and have low visibility, it is not suitable for monitoring. This paper focuses on the change detection algorithm for wide-field-of-view video images under foggy weather. The traditional clustering algorithm randomly selects the initial clustering center in change detection, which is easily interfered by noise points and produces unstable results. Therefore, this paper proposes a novel arctangent operator to suppress the noise interference. Inspired by the relationship between pixel distance and assignment probability, an enhanced K-means clustering algorithm (EN-K-means) is further proposed, which initializes the clustering prototype based on the nearest-neighbor relationship of pixels to improve the stability of the clustering results. Experimental results in real-world scenarios show that, compared to other methods, our algorithm demonstrates higher accuracy and stronger robustness.