Video has become an important channel for users to obtain destination information in recent years. However, so far, the research on video visual content in the tourism field is still relatively basic, and artificial analysis methods are mostly used, which has obvious limitations. In recent years, the field of computer vision has made great progress in video content analysis. Based on this, this study uses computer science video analysis methods to study user-generated videos on Bilibili.com. Deep learning and machine learning methods are used to reduce the dimensionality of video visual content, and then converted into picture data and text data for sentiment analysis. To perceive the emotions implied in user-generated videos, and to mine the scenes and categories that cause strong emotional changes in tourists. The study found that tourists' image cognition of Macau is concentrated in the fields of architecture, culture, catering, and leisure and entertainment. In terms of emotional image, tourists' positive strong emotions mainly come from leisure and entertainment and cultural modules. The collision of local culture and Western culture, the fusion of traditional culture and modern elements are the main elements that stimulate positive strong emotions. Negative strong emotions mainly come from living and shopping.