DEEP LEARNING BASED HUMAN OBJECT DETECTION IN FOG ENVIRONMENT FOR DEFENSE APPLICATION

--Foggy weather conditions greatly affect the performance of vision-based surveillance systems used in sensitive defence areas such as borders, military bases, and ammunition depots. When fog is present, visibility becomes very low because fog particles scatter and absorb light. As a result, cameras capture unclear images with poor contrast and reduced details. Due to this, traditional computer vision systems often fail to detect human intruders accurately, which can create serious security risks in defence environments. To address this problem, this work proposes a deep learning-based framework for detecting humans in foggy environments, specially designed for defence surveillance applications. The proposed system combines a lightweight image enhancement module with an improved YOLO-based object detection model. In the first stage, the enhancement module improves the quality of foggy images by increasing contrast and restoring important visual details. This step helps make the objects in the scene clearer and easier to detect. After the image enhancement stage, the improved images are processed by the YOLO-based detection model, which identifies humans in the scene by generating bounding boxes around detected objects. The use of YOLO allows the system to perform detection quickly while maintaining good accuracy, making it suitable for real-time surveillance applications. The model is trained and tested using fog-specific datasets such as Foggy Cityscapes and RTTS, alongwith additional low-visibility surveillance videos. Experimental results showthat the proposed system improves mean average precision (mAP) compared with standard detection models, while still maintaining near real-time performance. Overall, the results show that the system can effectively support continuous monitoring and early intruder detection in defence environments under foggy conditions. In the future, the system can be further improved by integrating thermal or infrared sensors, which can help detect humans even in verydense fog. The system can also be deployed on embedded devices to support large-scale defence surveillance operations.