Automatic Object Detection in Atmospheric Turbulence-Affected Environments

Atmospheric turbulence significantly hinders the interpretation and analysis of surveillance imagery, complicating tasks such as object classification and scene tracking. Such turbulence also diminishes the effectiveness of conventional methods used for detecting and tracking targets. While deep learning-based object detection methods perform well in normal conditions, they cannot be directly applied to sequences affected by atmospheric distortion. To address this challenge, we propose a novel unified architecture that learns to compensate for distorted features, thereby enhancing object detection and classification in turbulence-affected environments. Our approach employs a new 3D extension, 3DMAMBA, of an existing Mamba-based method in combination with an object detection model based on LWDETR. The proposed system is benchmarked against existing mitigation and detection methods and is shown to achieve superior performance in terms of mean Average Precision (mAP), outperforming non-combined approaches.