Complex traffic environments object detection is a key area in intelligent transportation and autonomous driving technology research. It has a direct influence on target recognition accuracy and provides credible data support to environmental understanding and decision-making. This paper reviews recent advances in YOLO-based object detection methods in traffic scenarios, including challenges of long-range detection, occlusion, truncation, illumination, and weather. Over the last few years, many of these models have been optimized and trained for the above purposes with methods such as attention mechanisms, feature fusion, and improved loss functions. Detection accuracy, robustness, and real-time response under some conditions have all been enhanced using these methods. This paper adopts literature review and analysis techniques to conclude the performance of general YOLO models after some optimizations in some complex environments, and discusses future possibilities for model improvement. The findings are expected to help with intelligent traffic system development and serve as a suitable reference for future practical use.
Zebin Luo (Tue,) studied this question.