Pedestrian detection in challenging environments such as low-light conditions and dense crowds remains a critical yet unresolved problem in computer vision. Despite significant advancements in deep learning-based object detection, existing methods—particularly traditional YOLO (You Only Look Once) architectures—often suffer from reduced accuracy and compromised real-time performance under poor visibility, severe occlusions, and complex background clutter.

To address these limitations, this paper introduces an enhanced YOLO-based algorithm specifically optimized for pedestrian detection in low-light and high-density scenarios. The proposed framework incorporates several key innovations:

1. Advanced Preprocessing Techniques: Adaptive image enhancement and noise reduction algorithms to improve input visibility in low-light conditions.

2. Optimized Backbone Network: A modified convolutional neural network (CNN) architecture designed to extract more discriminative features while maintaining computational efficiency.

3. Attention Mechanisms: Integration of spatial and channel attention modules to enhance focus on pedestrian regions, mitigating the effects of occlusions and background distractions.

Extensive experiments conducted on benchmark datasets (such as COCO, CrowdHuman, and custom low-light pedestrian datasets) demonstrate that our approach achieves superior performance in terms of both precision and recall compared to state-of-the-art methods. The system is implemented in Python and rigorously evaluated across varying lighting conditions and crowd densities, confirming its robustness, scalability, and real-time applicability.

The results indicate that the proposed model significantly outperforms baseline YOLO variants and other contemporary detectors, making it a viable solution for surveillance, autonomous driving, and crowd analysis in suboptimal environments.