Real-time detection of rear-moving objects is crucial for safe navigation in traffic management as rear-end collisions are one of the common types of accidents. Existing algorithms have successfully been implemented on static or front-moving objects. There are challenges such as shadows, similarly colored objects, partial occlusion by road infrastructure, varying lighting conditions, objects appearing small, etc. associated with rear-moving objects. To address the challenges, we propose a method that integrates background subtraction with YOLO object detection models for detecting rear-moving objects in traffic scenes. The background subtraction technique creates a binary mask that distinguishes between background and foreground objects by isolating dynamic elements. Morphological operations are then applied to refine the mask and eliminate noise. The continuous update of foreground and background models by a Mixture of Gaussians provides a robust method for real-time object detection. We collected and use video frames on pre-trained Yolov5s, Yolov7, Yolov7-tiny, and Yolov8s, and each of the models detect moving objects in the traffic scene. Yolov7-tiny achieves a Frame Per Second (FPS) of 123. The integrated approach enhances detection accuracy and reliability in complex traffic environments, making it suitable for real-time applications.