Abstract: Destination image perception is essentially a further manifestation of tourists’ spatiotemporal behavior patterns. Addressing the lack of empirical studies on this topic in small and medium-sized cities, this study takes Yibin City as a case area. Based on user-generated content (UGC) including tourist reviews and travel notes from Ctrip and Mafengwo between 2017 and 2024, a "spatiotemporal behavior-image perception" analytical framework is constructed. By integrating network text analysis, k-core decomposition, and the TF-IDF algorithm, this study examines how tourists’ behavioral characteristics are manifested in the expression of destination image perception, focusing on their representation patterns and perceptual differences. The results show that tourists are more likely to visit natural scenic spots in spring and summer while preferring cultural attractions in autumn and winter, which corresponds to high satisfaction with factors such as "landscape" and "historical culture." Core scenic areas receive more positive reviews due to higher visitation frequency, whereas remote destinations tend to receive more negative feedback due to poor transportation and service management. Semantic network analysis reveals that the central place-name nodes closely align with tourists’ travel paths. However, the absence of direct connections between these nodes indicates insufficient linkage among core attractions. The study suggests enhancing the unique characteristics of key scenic spots, improving service facilities, and promoting coordinated development among attractions to optimize travel routes and improve image perception experiences.