交通事件之確認與通報為交通管理之重要議題，往往耗費人力資源以及相互溝通確認時間。雖然人工智慧日益蓬勃發展，但卻鮮少研究透過人工智慧處理交通事件辨析之議題。因此，本研究旨在發展以人工智慧為基礎的影像偵測模式，透過交通事件之自動化偵測，以提升交通管理效率。具體而言，本研究建構單次多重目標檢測器之深度神經網路，以偵測交通事件。為了測試本方法之效果，本研究以高雄市一個交叉路口作為實測場域。實測分析結果顯示，本研究提出之交通事件單次多重目標檢測器模式可以成功地偵測交通事件，並可同時偵測車速與車輛數。再者，本研究示範如何將人工智慧為基礎之交通事件影像偵測技術應用於實體交通環境上。本開創性研究提供清楚的指引，讓交通管理單位可以透過人工智慧技術去偵測交通事件，對於交通管理具有高價值貢獻。

The detection and reporting of traffic incidents is of major importance to all those involved in traffic management, as it results in increased labor and prolonged periods of complex communication. However, few studies have explored how these pressures might be relieved through the use of artificial intelligence (AI). Accordingly, this study aims to develop an AI-based image-detection model, the Single Shot MultiBox Detector (SSD) with a deep neural network, which will detect and report traffic incidents automatically, and thus enhance the efficiency of traffic management. This study used the field case of a real intersection in Kaohsiung City, Taiwan, to test the effectiveness of the proposed approach, and the results indicated that the proposed traffic-incident SSD model was successful not only in identifying traffic incidents, but also in monitoring key background traffic parameters such as the numbers and speeds of vehicles on the road. This pioneering research also demonstrates how AI-based image-detection technology for traffic incidents could be installed in the physical environment, and provides clear and valuable guidance to traffic managers interested in utilizing AI technology in their field.