Abstract: The morphological evolution of rivers is closely related to regional socio-economic development and ecological environment. Taking the lower Minjiang River as the research object, this study employed neural network algorithms to interpret river information from remote sensing images from 1990 to 2021. Changes in water body, sandbars, and floodplain areas, as well as channel meandering, were analyzed to explore the characteristics and causes of river channel evolution. The results indicate that the constructed neural network algorithm is effective for remote sensing water body extraction, with an average accuracy of 95% and average recall rate of 90%. Over the past 30 years, sandbars and floodplain areas in the lower Minjiang River have generally decreased. The river channel widened from 1990 to 2000, contracted from 2000 to 2011, and has since stabilized. Decreases in sandbars in the Minqing-Minhou sections resulted in river channel widening, while occupation of sandbars in the Beigang section led to channel contraction. Reductions in floodplain areas in the Nangang section corresponded to channel widening, while the Maowei section exhibited relatively stable channel morphology. Factors influencing the morphology of the lower Minjiang River include reservoir operation, sand mining activities, geological factors, engineering construction, and tidal effects. This study can provide a reference for similar studies on river evolution and for planning and conservation efforts, morphology evolution and protection of other tidal rivers.