<p>The contradiction between the irrational use of urban landscape land and the public’s high demands for environmental health is becoming increasingly apparent. With the advancement of artificial intelligence, there is unprecedented potential to optimize urban landscape land use rates to mitigate environmental health risks and enhance public physical and mental well-being. This study enhances the SegFormer model by incorporating a Convolutional Block Attention Module and proposes a land use optimization method based on the improved model. Research results show that the feature extraction accuracy of this model can reach 97.4%, and the feature extraction time only takes 12.3&#xa0;ms. This model has a better feature extraction effect on images. The effectiveness of the land use optimization method based on this model was then analyzed. Results indicated that after implementation of this optimization approach, the city’s land resource utilization rate increased by 23.3%, ecological environment quality improved by 32.8%, and urban environmental health risks decreased by 33.1%. In summary, the proposed optimization method for urban landscape land use could enhance the effectiveness of such land use, improve the quality of the urban ecological environment, and promote public physical and mental health.</p>