<p>With the growth of economic level, people's pursuit of visual landscape is gradually increasing. However, existing visual landscape evaluation methods have a high degree of subjectivity. Therefore, the paper constructs a visual landscape evaluation method for scenic plant communities built on K-means clustering and beauty evaluation model. This method investigates the plant communities in scenic areas, uses K-means algorithm to cluster and analyze the color and texture of plants, uses whale optimization algorithm to improve the initial clustering center of K-means, and uses isolated forest and K-nearest neighbor algorithm to detect abnormal clustering points. This study combines the Analytic Hierarchy Process and the Scenic Beauty Estimation Model to construct an indicator evaluation system and conduct quantitative analysis of subjective factors. The experiment showed that the maximum clustering accuracy of the improved K-means was 97.4%, and the performance only decreased by 0.2% in the test set, indicating good generalization. In spring, there was a difference in the color attributes of plant communities, with the green color scheme accounting for the largest proportion, reaching 61.5% in plot 1, while other color schemes were all below 20%. The visual landscape in spring was optimal, with the beauty values of plant communities 1–4 being 1.682, 1.635, 1.651, and 1.596, respectively. Therefore, research methods can effectively evaluate the visual landscape beauty of plant communities and improve the objectivity of scenic spot beauty evaluation.</p>