基于内容的交互式面料图像推荐研究

doi:10.13475/j.fzxb.20240600501

摘要/Abstract

摘要：

针对传统面料推荐算法忽视面料视觉特征对用户兴趣度的影响问题,提出了一种考虑面料颜色、纹理等视觉特征,将用户喜好与面料特征属性关联的交互式面料图像推荐算法。首先利用基于HSV颜色空间的快速颜色量化算法提取图像的主色颜色集及相应主颜色占比,并采用基于ResNet18网络模型的迁移学习算法对面料纹理进行分类。然后根据显性交互评分和隐性兴趣度度量方法,建立交互式特征喜好推荐模型,该模型依据用户的喜好评分计算出用户对每个特征属性的兴趣度值。最后根据模型预测用户对数据库中每张面料的评分,向用户推荐符合其偏好的面料图像。研究结果表明:所有用户对推荐结果中的9块面料的平均评分在5.5分以上,其中有93.3%的用户对推荐结果表示非常感兴趣,说明该推荐模型的性能良好,能较准确地捕捉用户偏好,为用户推荐其需求的面料,为个性化面料推荐算法研究提供了参考。

关键词: 面料设计, HSV颜色空间, 面料图像推荐算法, 面料特征, 视觉特征, 用户喜好

Abstract:

Objective With rapid development of textile industry, the types of fabrics are increasingly diversified. Facing a huge fabric library, it is difficult for the traditional retrieval method to carry out accurate search particularly when the user needs are not defined clearly enough. The popularity of e-commerce platforms has promoted the application of recommendation algorithms. The main idea of recommendation algorithms is to proactively recommend relevant content that users may be interested in by analyzing their preferences. Therefore, this paper proposes a content-based interactive fabric image recommendation algorithm, which considers the color and texture characteristics of the fabric based on the user's interest, aiming to accurately capture the user's preference, recommend the fabric for the user's needs, and provide certain reference value for the research of personalized fabric recommendation algorithms.

Method This research considers the visual features such as fabric color and texture, and links the user's preferences with the characteristics of fabric, and proposes a content-based interactive fabric image recommendation algorithm. First, the main color set of the image is extracted from HSV color space, and the fabric texture is classified using the transfer learning algorithm based on ResNet18 network model. On this basis, an interactive feature preference recommendation model is established. The model calculates the interest value of each feature attribute according to the user's preference score for the fabric. Used to predict user ratings for potential fabrics. The 9 fabrics with the highest rating are recommended to the user, and the performance of the recommendation model is evaluated according to the user's rating for the recommendation results.

Results The obtained data set was divided into training set, verification set and test set according to the ratio of 8∶1∶1. In this paper, the transfer learning algorithm based on the ResNet18 network model was used to classify fabric textures, and the highest accuracy of the test set was 0.936. It is shown that the ResNet18 network model has a good performance for fabric image texture classification. According to the rating preferences of 30 testers for the 9 recommended fabrics, the average score of these 30 users is above 5.5 points, in line with the range of 4-7 points (generally interested), and most of the scores gather in the range of 7-10 points (very interested), among which 19 people have an average score of more than 8 points, and 4 people have an average score of more than 9 points. The percentage of users who are very interested in the recommendation results as a whole is as high as 93.3%, among which 18 users score at least 7 of the 9 recommended fabrics above 8 points, which indicates that the overall performance of the recommendation model proposed in this paper is good.

Conclusion In view of the fact that traditional recommendation algorithms fail to consider the influence of visual features of fabric on users' interest in fabric recommendation, this paper classifies fabric color and texture based on hand-extracted features and higher-order features respectively. Experimental results show that the transfer learning algorithm based on ResNet18 network model has achieved good results in fabric texture classification. The highest accuracy of the test set reached 0.936. Subsequent experiments will continuously optimize the accuracy and reduce the impact of classification errors on the recommendation results. In this paper, an interactive feature preference recommendation model is established by means of interactive scoring. The model calculates the interest value of each feature attribute according to the user's preference rating, so as to predict the user's rating on potential fabrics, and then recommend fabric images with high predictive ratings for users. Users' preferences for the recommendation results can be used as a standard to evaluate the performance of the recommendation model. The experimental results show that most users are satisfied with the recommendation results, which indicates that the overall performance of the recommendation model proposed in this paper is good, and it can capture users' interests and preferences more accurately, providing certain reference value for fabric image recommendation.

Key words: fabric design, HSV color space, fabric image recommendation algorithm, fabric feature, visual feature, user preference

中图分类号:

TS941.26

明宇豪, 张宁, 向军, 潘如如. 基于内容的交互式面料图像推荐研究[J]. 纺织学报, 2025, 46(06): 135-142.

MING Yuhao, ZHANG Ning, XIANG Jun, PAN Ruru. Research on content-based interactive fabric image recommendation[J]. Journal of Textile Research, 2025, 46(06): 135-142.

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: http://www.fzxb.org.cn/CN/10.13475/j.fzxb.20240600501

http://www.fzxb.org.cn/CN/Y2025/V46/I06/135

图/表 12

图1

图2

表1

图3

图4

表2

表3

图5

表4

图6

图7

表5

参考文献 19

[1]	KO H, LEE S, PARK Y, et al. A survey of recommendation systems: recommendation models, techniques, and application fields[J]. Electronics, 2022. DOI:10.3390/edetronics11010141.
[2]	GOLDBEG D, NICHOLS D, OKI B M, et al. Using col-laborative filtering to weave an information tapes-try[J]. Communications of the ACM, 1992, 35(12): 61-70.
[3]	BATMAZ Z, YUREKLI A, BILGE A, et al. A review on deep learning for recommender systems: challenges and remedies[J]. Artificial Intelligence Review, 2019, 52: 1-37.
[4]	王宇哲. 基于内容的电影推荐算法研究[J]. 信息系统工程, 2023,12:117-120.
	WANG Yuzhe. Research on content-based movie recommendation algorithm[J]. Information System Engineering, 2023,12:117-120.
[5]	郑志军. 基于内容的推荐算法在专题数据库中的应用研究[J]. 信息与电脑(理论版), 2023, 35(15):116-119.
	ZHENG Zhijun. Research on the application of content based recommendation algorithms in special data-base[J]. China Computer & Communication, 2023, 35(15):116-119.
[6]	苏启琛, 苏洋. 一种基于内容的生鲜产品推荐算法[J]. 电脑知识与技术, 2020, 16(28):189-191, 207.
	SU Qichen, SU Yang. A content-based recommendation algorithm for fresh products[J]. Computer Knowledge and Technology, 2020, 16(28): 189-191, 207.
[7]	刘克礼, 王荣华. 基于协同过滤的学习资源个性化推荐应用[J]. 安徽广播电视大学学报, 2017(4):125-128.
	LIU Keli, WANG Ronghua. Research on application of personalized recommendation of learning resources based on collaborative filtering[J]. Journal of Anhui Radio and Television University, 2017(4):125-128.
[8]	张迪, 姜久雷, 蒋永旺. 基于用户偏好协同过滤算法的个性化学习资源推荐[J]. 中国信息技术教育, 2018,20:107-109.
	ZHANG Di, JIANG Jiulei, JIANG Yongwang. Personalized learning resource recommendation based on user preference collaborative filtering algorithm[J]. Chinese Information Technology Education, 2018,20: 107-109.
[9]	杨宇. 基于深度学习特征的图像推荐系统[D]. 成都: 电子科技大学,2016:1-60.
	YANG Yu. Image recommendation system based on the image features obtained from deep learning[D]. Chengdu: University of Electronic Science and Technology of China,2016:1-60.
[10]	HE Haitao, ZHANG Ruixi, ZHANG Yangsen, et al. AAIN: attentional aggregative interaction network for deep learning based recommender systems[J]. Neurocomputing, 2023.DOI:10.1016/j.neucom.2023.126374.
[11]	李锋, 冷娜. 协同过滤的纺织面料推荐系统[J]. 计算机应用, 2018, 38(S2):115-118.
	LI Feng, LENG Na. Textile fabric recommendation system based on collaborative filtration[J]. Computer Application, 2018, 38(S2):115-118.
[12]	杨海文, 李知霖, 王建萍, 等. 基于协同过滤的针织面料推荐模型优化研究[J]. 针织工业, 2021(9):12-15.
	YANG Haiwen, LI Zhilin, WANG Jianping, et al. Optimization of knitted fabric recommendation model based on collaborative filtering[J]. Knitting Industries, 2021(9):12-15.
[13]	HE Zhenzhen, MA Yunjiao, XIANG Jun, et al. Content-based fabric recommendation system integrating image and text information[J]. Journal of The Textile Institute, 2025, 116(4):613-622.
[14]	龚雪, 袁理, 刘军平, 等. 混合色彩空间与多核学习的色纺织物组织点识别[J]. 纺织学报, 2020, 41(5): 58-65.
	GONG Xue, YUAN Li, LIU Junping, et al. Recognition of colored spun fabric interlacing point based on mixed color space and multiple kernel learning[J]. Journal of Textile Research, 2020, 41(5): 58-65.
[15]	张宁. 基于特征工程的机织面料图像检索方法研究[D]. 无锡: 江南大学,2021:1-50.
	ZHANG Ning. Research on image retrieval methods for woven fabrics based on feature engineering[D]. Wuxi: Jiangnan University,2021:1-50.
[16]	赵文浩, 向军, 张宁, 等. 基于SURF和VLAD特征编码的面料图案检索研究[J]. 纺织学报, 2023, 44(8):110-117.
	ZHAO Wenhao, XIANG Jun, ZHANG Ning, et al. Research on fabric pattern retrieval based on SURF and VLAD feature coding[J]. Journal of Textile Research, 2023, 44(8):110-117.
[17]	WANG Q, HUANG W, XIONG Z, et al. Looking closer at the scene: multiscale representation learning for remote sensing image scene classification[J]. IEEE Transactions on Neural Networks and Learning Systems, 2020, 33(4): 1414-1428.
[18]	WANG Q, YUAN X, SUN M. Collaborative filtering recommendation algorithm based on hybrid user model[C]// Seventh International Conference on Fuzzy Systems and Knowledge Discovery.[S.l.]: IEEE, 2010: 1985-1990.
[19]	王海燕, 刘国联. 基于消费者感性需求的格子图案认知[J]. 纺织学报, 2014, 35(11):151-156.
	WANG Haiyan, LIU Guolian. Study on cognition of grid patterns based on consumer's sensibility demand[J]. Journal of Textile Research, 2014, 35( 11): 151-156.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

色系	h	s	v
黑色	[0, 360]	[0, 1]	[0, 0.2]
灰色	[0, 360]	[0, 0.2]	[0.2, 0.8]
白色	[0, 360]	[0, 0.2]	[0.8, 1]
红色	(315, 360]∪[0, 20]	[0.2, 1]	[0.2, 1]
黄色	(20, 45]	[0.2, 1]	[0.2, 1]
绿色	(45, 90]	[0.2, 1]	[0.2, 1]
青色	(90, 160]	[0.2, 1]	[0.2, 1]
蓝色	(160, 270]	[0.2, 1]	[0.2, 1]
紫色	(270, 315]	[0.2, 1]	[0.2, 1]

面料序号	C₁, C₂,…, C₉	T₁, T₂,…, T₆
i₁	0,1,…,0	1,0,…,0
i₂	1,0,…,1	0,1,…,0
︙	︙	︙
i_k	1,1,…,0	0,0,…,0

用户序号	C₁, C₂,…, C₉	T₁, T₂,…, T₆
u₁	7.53, 3.61,…, 5.87	2.83, 9.52,…, 7.30
u₂	2.33, 8.56,…, 0	0, 6.58,…, 8.23
︙	︙	︙
u_n	1.30, 2.71,…, 7.82	3.55, 2.01,…, 9.34

数据集编号	图像数量/张
数据集编号	训练集	验证集	测试集
Q₁	736	92	92
Q₂	624	78	78
Q₃	496	62	62
Q₄	672	84	84
Q₅	640	80	80
Q₆	328	41	41

用户编号	平均评分	高评分面料数	用户编号	平均评分	高评分面料数
u₁	8.34	7	u₁₆	8.66	8
u₂	8.13	6	u₁₇	7.55	6
u₃	7.38	6	u₁₈	5.62	3
u₄	8.00	7	u₁₉	7.89	6
u₅	9.50	9	u₂₀	8.51	7
u₆	7.44	6	u₂₁	7.21	4
u₇	7.60	7	u₂₂	8.11	5
u₈	7.35	7	u₂₃	9.13	8
u₉	6.37	3	u₂₄	8.80	8
u₁₀	8.67	9	u₂₅	7.26	5
u₁₁	8.10	7	u₂₆	8.50	7
u₁₂	8.73	7	u₂₇	8.36	7
u₁₃	7.38	7	u₂₈	8.92	8
u₁₄	8.73	6	u₂₉	8.15	6
u₁₅	9.80	9	u₃₀	9.42	9