As a traditional Chinese handicraft, the pattern design of lacquer ware has important cultural value and artistic charm. Traditional lacquerware pattern design mainly relies on manual production, which has problems such as low efficiency and poor precision. The development of modern computer vision technology provides new ideas for the automatic identification and optimization of lacquerware patterns. It is of great significance to inherit and develop lacquer art by realizing automatic detection and classification of lacquer patterns through image recognition technology, and optimizing and innovating the geometric form of the patterns by combining with intelligent algorithms. Objective: Aiming at the problems of low efficiency and lack of innovation in traditional lacquerware pattern design, this study proposes a method of optimizing the geometric form of lacquerware patterns based on image recognition technology to realize automatic detection, classification and innovative design of lacquerware patterns. Methods: Improved Canny algorithm is used for lacquerware pattern contour detection, and the detection accuracy is improved by bilateral filtering, multi-directional Sobel template and interactive threshold detection based on Otsu, etc.; HOG feature extraction and SVM classifier are used to realize automatic detection of lacquerware patterns; optimization and innovation of the geometric form of the lacquerware patterns are carried out based on BP-GA algorithm; and virtual reality technology is used to carry out the Perceptual imagery evaluation experiment. Results: The experimental results show that the lacquerware pattern detection model proposed in this paper achieves 94.80% mAP in the recognition of 7 pattern types, among which the best effect is achieved in the detection of plant patterns, with 96.09% mAP among classes; in the pattern classification task, the accuracy of the model in this paper reaches 94.73%, with 94.14% F1 score; the optimized lacquerware patterns are more effective in the recognition of the sensual imagery in the categories of “hip” and “chic”; and the optimized lacquerware patterns are more effective in the recognition of the sensual imagery in the categories of “new” and “chic”. Conclusion: The method in this paper effectively improves the recognition accuracy and classification performance of lacquerware patterns, and the optimized pattern geometry of the BP-GA algorithm is more in line with the modern aesthetic demand, which provides technical support for the digital protection and innovative development of traditional lacquerware art.
