With the rapid development of intelligent transportation systems (ITS) and autonomous driving technology, precise and real-time traffic signal recognition has become a core technology for ensuring driving safety and improving traffic efficiency at present. However, traditional image processing methods are prone to instability in complex scenarios such as light fluctuations and adverse weather conditions, which leads to deficiencies in practical application. This paper systematically reviews the research progress of real-time traffic signal recognition technology based on deep learning: Clarify the complete evolution of this technology from traditional methods to deep learning, analyze the core challenges in the real-time recognition process and the corresponding solutions, summarize the key research results and compare the performance of mainstream technical solutions. Research and demonstration show that deep learning demonstrates a generation gap advantage in balancing accuracy and instantaneous response, establishing its position as the core evolution path of the industry. The technical paradigms and optimization details discussed in this article not only provide theoretical support for algorithm iteration, but also offer a key reference for the industrial application of intelligent traffic recognition due to its breakthrough in performance balance.
Weilin Chen (Mon,) studied this question.