With the acceleration of urbanization, the increasing demand for environmental protection, and the pursuit of high-quality living standards, the study of visual comfort in urban park landscapes has become a key issue in urban planning and management. However, existing research has yet to establish a comprehensive set of evaluation indicators and quantitative methods. To address this gap, this paper proposes a comprehensive evaluation method for park landscape visual perception based on the theory of perceptual ecology. This method integrates semantic segmentation technology with the HSB color model to quantify landscape features across four dimensions-spatial, natural, architectural, and color, thereby expanding the evaluation framework for urban park landscapes. Through comparative analysis with the stepwise linear regression model, the superiority of the Random Forest regression model in predicting visual comfort in park landscapes is validated, and its performance in various visual comfort assessments is explored. The study is based on a dataset of landscape images from ten waterfront urban parks in Hangzhou. Experimental results reveal that factors such as sky openness, water body coverage, tree richness, herbaceous plant richness, grass coverage, and flowering plant richness positively influence visual comfort, while road paving density and color contrast (warm vs. cool) exert a negative impact. The R² value of the Random Forest regression model (0.814) significantly surpasses that of the stepwise linear regression model (0.601), demonstrating high predictive accuracy across different visual rating levels, particularly in extreme ratings (1 and 5). This research highlights the importance of integrating diverse dimensions in park design, emphasizing the need for a balanced approach that maximizes visual comfort. Additionally, it underscores the potential of employing advanced technologies like semantic segmentation and machine learning for objective and comprehensive landscape evaluation. This research can significantly aid urban planners and landscape architects in creating parks that cater to the diverse needs and preferences of the public, ultimately contributing to the development of sustainable and livable cities. This research paves the way for future investigations into the complex interplay between landscape features and human emotions and well-being. By exploring the specific preferences of different demographic groups and incorporating additional technologies such as deep learning and virtual reality, a more nuanced understanding of the relationship between landscape perception and human experience can be achieved. This deeper insight can inform the development of more personalized and contextually relevant urban design solutions, ultimately enhancing the quality of life for urban residents and promoting the overall sustainability of cities.