In the field of architectural decoration, the importance of color matching is increasingly prominent. It not only affects the aesthetics of space, but also relates to people's emotions and behavior. This study aims to explore the role of CNN in color matching in architectural decoration and evaluate its effectiveness. This study uses Convolutional Neural Network (CNN) as the core algorithm and combines it with K-means clustering algorithm to optimize the color extraction process. By collecting and preprocessing color data of building decoration materials, this study trains a CNN model and validates its performance using historical color matching case data. The experimental results show that the average value of the CNN model on the color beauty index is 0.955, which is better than the support vector machine (SVM)'s 0.814, demonstrating the generalization ability and accuracy of CNN in color matching tasks. CNN also performs well in originality ratings, with an average score of 9.465, higher than SVM's 7.74, demonstrating its ability to generate high-altitude creative color schemes. However, there are limitations to the research, including limitations in dataset size and diversity, as well as the need to improve algorithm speed and ability to handle large-scale datasets. Future research will focus on expanding datasets, optimizing algorithm performance, and exploring the potential applications of intelligent algorithms in other design fields.