By integrating AIS data with visual data, this study aims to propose a data fusion-based methodology for the robust assessment of vessel collision risk in inland waterways. With the data from the Huangpu River, one case study is created and comparison results show that visual data could greatly overestimate the vessel collision risk while the vessel conflict count could be underestimated from AIS data. The proposed methodology could contribute to a 35.70% reduction in the false detection rate compared to the pure visual data but an increase of 16.84% in the conflict detection accuracy compared to the AIS data. Moreover, results show that the proposed methodology is able to perform much better than the pure AIS data technique during the daytime period while it outperforms the pure visual data technique during the nighttime period. This suggests that the proposed data-fusion methodology could provide a more reliable and accurate estimate for the vessel collision risk in inland waterways. Results also show that higher collision risk is associated with the high tide period and turning area in the Huangpu River. Therefore, special attention should be placed to enhance the navigation safety during this time period for the inland waterway turning area.