Robustness of DPTrack under adverse weather conditions

Determine the robustness of DPTrack—a nighttime aerial tracker that generates prompts via a directional kernel encoded from topology-aware features—under adverse weather conditions such as rain and fog in aerial imagery, assessing whether its tracking performance remains reliable beyond low-illumination scenarios.

Background

DPTrack is introduced as a prompt-based nighttime aerial tracker that encodes topology-aware features into a directional kernel to generate precise prompts, aiming to improve object localization in low-light scenarios. The method comprises three key modules: the Dual Particle Perception module (DPP), the Directional Kernel Adaptive Encoder (DKE), and the Kernel-Guided Prompt module (KGP). Extensive experiments across multiple nighttime aerial benchmarks demonstrate notable performance gains in precision and AUC.

Despite its strong results under low-illumination conditions, the paper explicitly acknowledges that DPTrack’s behavior in other adverse weather conditions has not been examined. Weather phenomena such as rain and fog can introduce additional visual challenges (e.g., scattering, glare, occlusions, and motion blur) that may affect tracking reliability, making the assessment of DPTrack’s robustness in these environments an unresolved question.