Weather-Conditioned Small-Target Detection in Low-Resolution Aerial Imagery: A Review of Robust YOLO Strategies

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The number of pixels required for representing small military objects in aerial im-ages is minimal, and down sampling, background complexity and weak texture can reduce the information necessary to accurately localize and classify these ob-jects. The appearance and contrast is further changed by haze, fog and rain, which results in a domain shift that may not be represented by clear weather bench-marks. This review consolidates recent work on aerial small-object detection, ad-verse-weather robustness and YOLO-based real-time detection, with emphasis on their intersection under low spatial resolution. Existing methods are organized into data-centric augmentation, input restoration and attenuation modelling, fea-ture-preserving network design, and training or deployment optimization. Com-parisons emphasize methodological trade-offs rather than raw performance values because datasets, input scales, thresholds and hardware differ. A review-derived reference architecture is synthesized from the literature and the supplied source presentation. The principal gap is the limited standardization of joint evaluation across object size, resolution and weather severity. Research priorities include weather-stratified benchmarks, domain generalization, small-target-aware high-resolution features, uncertainty estimation and edge-efficient deployment.