<p>Preventing traffic crashes presents a formidable challenge due to the intricate interplay between drivers and other participants within a complex urban infrastructure. In recent years, increasing studies on road safety involved computer vision and machine learning to detect visual features from street view imagery (SVI) and explore their impacts on crashes, though the recent progress is poorly understood. This paper conducted a comprehensive review of existing literature to investigate how SVI has been used in traffic crashes and road safety studies, utilizing a broad database collection including Scopus, Web of Science, and Transport Research International Documentation. We categorized SVI-generated features into two types of factors, explored their relationship with traffic crashes, and examined the prevalent detection models. Our review demonstrated that SVI plays an important role in capturing road design and driving environment factors, which significantly influence the frequency and risk of traffic crashes. These findings underscore the significant impact of these street visual factors on road safety. Through a systematic review of recent progress, we also identified challenges and future research opportunities for SVI applications in traffic crash study, such as the potential use of large language models.</p>