Quantitation and grading method for ceramic tile chromatic aberration based on improved fractal encoding network

doi:10.11772/j.issn.1001-9081.2025030361

Journal of Computer Applications ›› 2026, Vol. 46 ›› Issue (3): 959-968.DOI: 10.11772/j.issn.1001-9081.2025030361

• Multimedia computing and computer simulation • Previous Articles Next Articles

Quantitation and grading method for ceramic tile chromatic aberration based on improved fractal encoding network

Songsen YU(), Huang HE, Guopeng XUE, Hengtuo CUI

School of Artificial Intelligence，South China Normal University，Foshan Guangdong 528225，China

Received:2025-04-08 Revised:2025-05-29 Accepted:2025-06-03 Online:2025-06-16 Published:2026-03-10
Contact: Songsen YU
About author:HE Huang， born in 1999， M. S. candidate. His research interests include computer vision.
XUE Guopeng， born in 1999， M. S. candidate. His research interests include computer vision.
CUI Hengtuo， born in 1999， M. S. candidate. His research interests include computer vision.
Supported by:
Key Project of Guangdong Provincial Basic and Applied Basic Research Fund — Provincial and Municipal Joint Fund(2020B1515120089)

基于改进FENet的瓷砖色差量化分级方法

余松森(), 何皇, 薛国鹏, 崔恒拓

华南师范大学人工智能学院，广东佛山 528225

通讯作者: 余松森
作者简介:何皇（1999—），男，江西抚州人，硕士研究生，主要研究方向：计算机视觉
薛国鹏（1999—），男，江西赣州人，硕士研究生，主要研究方向：计算机视觉
崔恒拓（1999—），男，浙江宁波人，硕士研究生，主要研究方向：计算机视觉。
基金资助:
广东省基础与应用基础研究基金省市联合基金重点项目(2020B1515120089)

Abstract

Abstract:

To address the result instability caused by subjective visual estimation dependence in traditional ceramic tile color difference detection methods， a method integrating texture and color features was proposed for quantitation and grading of chromatic aberration in ceramic tiles. A large-scale dataset named TILE-TCR （TILE Texture and Color Recognition）， containing both texture and color labels， was constructed to enhance the model’s ability to represent texture and color features. At the same time， a color difference grading dataset named TILE-CAG （TILE Chromatic Aberration Grade） was established to support the color difference classification task. Based on these datasets， the network structure of Fractal Encoding Network （FENet） was improved by introducing Spatial Pyramid Pooling （SPP） and SE （Squeeze-and-Excitation） modules， thereby extracting multi-task features and focusing on critical regions. Then， a clustering algorithm was employed to determine the thresholds for color difference grading adaptively， thereby implementing objective quantification of color difference grading. Experimental results show that the proposed improved method achieves an accuracy of 92.82% in the ceramic tile texture classification task， representing a 1.84 percentage point improvement compared to the FENet； in the color difference grading task， the accuracy， precision， recall and F1 score of the proposed method exceed 90%. Furthermore， a simulated production line was built for industrial deployment and real-time performance test of the model. On commonly used ceramic tiles， the average detection time of the proposed method is under 3 seconds， meeting the real-time requirements for online inspection of industrial conveyor belts.

Key words: ceramic tile color difference detection, deep learning, multi-task learning, chromatic aberration grading, intelligent detection system, Fractal Encoding Network (FENet)

摘要：

针对瓷砖色差检测中传统方法依赖主观目测导致的结果不稳定问题，提出一种融合纹理与颜色特征的瓷砖色差量化与分级方法。构建包含纹理与颜色双标签的大规模瓷砖数据集TILE-TCR（TILE Texture and Color Recognition），以提升模型对纹理与颜色特征的表征能力；同时，构建色差分级数据集TILE-CAG（TILE Chromatic Aberration Grade），用于支持色差分类任务。在此基础上，改进分形编码网络（FENet）的网络结构，即引入空间金字塔池化（SPP）与SE（Squeeze-and-Excitation）模块，从而实现多任务特征提取与关键区域聚焦。然后，通过聚类算法自适应确定色差分级阈值，从而实现色差分级的客观量化。实验结果表明，所提改进方法在瓷砖纹理分类任务中的准确率达到92.82%，较FENet提升了1.84个百分点；在色差分级任务中所提方法的准确率、精确率、召回率和F1分数均超过90%。此外，还搭建了模拟生产流水线，以完成模型的工业部署与实时性测试。而所提方法在常见规格瓷砖上的平均检测时间低于3 s，满足工业传送带在线检测的实时性要求。

关键词: 瓷砖色差检测, 深度学习, 多任务学习, 色差分级, 智能检测系统, 分形编码网络

CLC Number:

TP391.41

Songsen YU, Huang HE, Guopeng XUE, Hengtuo CUI. Quantitation and grading method for ceramic tile chromatic aberration based on improved fractal encoding network[J]. Journal of Computer Applications, 2026, 46(3): 959-968.

余松森, 何皇, 薛国鹏, 崔恒拓. 基于改进FENet的瓷砖色差量化分级方法[J]. 《计算机应用》唯一官方网站, 2026, 46(3): 959-968.

Figures/Tables 14

Fig. 1 Examples of TILE-TCR dataset

Tab. 1 Statistics of ceramic tile texture classification dataset

纹理类型	训练集样本数	测试集样本数
chaos（混乱纹理）	666	86
crackle（裂纹纹理）	535	62
mosaic（马赛克纹理）	190	30
shallow（浅表纹理）	689	71
snowflake（雪花纹理）	396	50
spot（斑点纹理）	881	96
stripe（条纹纹理）	1 149	119
watermark（水印纹理）	325	40
wood（木质纹理）	693	78

Tab. 2 Sample numbers of images in ceramic tile color classification dataset

颜色类型	训练集样本数	测试集样本数
0（白色）	621	49
1（浅灰色）	784	66
2（深棕色）	314	22
3（浅蓝色）	656	90
4（红色）	132	32
5（黑色）	105	5
6（木纹色）	268	30
7（深灰色）	542	50
8（黄色）	127	29
9（米色）	879	96
10（深褐色）	363	40
11（浅褐色）	733	123

Fig. 2 Partial examples of TILE-CAG dataset

Fig. 3 Architecture of FENet-SE-SPP-Color

Tab. 3 Description of main module parameters of FENet-SE-SPP-Color

模块	输入尺寸	输出尺寸	参数量说明
ResNet-Conv1	224×224×3	112×112×64	卷积核7×7，stride=2
ResNet-Layer4	14×14×1 024	7×7×2 048	Bottleneck结构
GAL	—	512	注意力加权融合
GlobalPool+SubLinear	7×7×2 048→1×1×2 048	256	GAP+FC
Upsample+SE	7×7×2 048	28×28×256	转置卷积核4×4
SPP	7×7×2 048	1 024	1×1+2×2+4×4池化
FAP	128通道	3维	每通道分形维度提取
Final Linear（纹理）	256	n	n类纹理等级
Final Linear（颜色）	256	m	m类颜色等级

Tab. 4 Ablation experimental results

骨干网络	A	B	C	最大准确率/%	平均准确率/%	标准差（±Std）/%	p-value值（对比基准模型）
FENet18				90.66	89.746	0.633 0	0.000 0
	√			90.66	89.794	0.490 9	0.872 0
		√		90.82	90.127	0.399 5	0.217 3
			√	91.30	89.907	0.560 5	0.565 8
	√	√		90.82	89.905	0.555 2	0.331 9
		√	√	91.46	89.904	0.964 9	0.740 4
	√		√	91.61	90.332	0.610 1	0.104 3
	√	√	√	92.41	90.363	0.888 1	0.135 9
FENet50				90.98	90.085	0.497 6	0.000 0
	√			91.46	90.181	0.600 0	0.705 6
		√		91.30	90.270	0.686 1	0.405 4
			√	91.61	90.678	0.477 0	0.028 8
	√	√		92.56	90.695	0.851 3	0.131 7
		√	√	91.93	90.743	0.820 5	0.127 2
	√		√	91.72	90.483	0.706 4	0.225 9
	√	√	√	92.82	91.099	0.828 1	0.029 6

Tab. 5 Comparison of texture classification accuracy using different methods

方法	骨干网络	DTD	FMD	MINC-2500	GTOS-M	TILE-TCR
DeepTEN	ResNet18	64.60	68.0	77.79	72.12	87.50
DEPNet		67.80	76.1	78.28	79.18	88.26
HistNet		69.89	83.0	80.57	82.95	88.83
FENet		70.85	81.0	80.45	80.53	90.66
FENet-SE-SPP-Color		73.30	84.0	82.59	82.16	92.41
DeepTEN	ResNet50	69.60	80.6	80.40	74.20	87.28
DEPNet		73.20	82.0	82.00	76.07	89.87
HistNet		72.78	82.0	83.22	82.74	90.19
FENet		74.04	86.0	82.83	83.33	90.98
FENet-SE-SPP-Color		73.46	87.0	83.25	84.27	92.82

Tab. 6 Experimental results of multi-task recognition

纹理任务权重	颜色任务权重	纹理识别准确率/%	颜色识别准确率/%	总体平均准确率/%
0.9	0.1	89.40	77.85	83.62
0.8	0.2	90.19	78.32	84.26
0.7	0.3	90.03	78.96	84.49
0.6	0.4	88.77	82.75	85.76
0.5	0.5	88.77	81.49	85.13
0.4	0.6	88.92	81.01	84.97
0.3	0.7	88.29	81.01	84.65
0.2	0.8	85.92	83.07	84.49
0.1	0.9	76.74	84.18	80.46

Tab. 7 Experimental results of color difference grading threshold

纹理类别	小色差对照组				大色差对照组
纹理类别	纹理差异均值（DT₁）	颜色差异均值（DC₁）	D₁	$ε 1$	纹理差异均值（DT₂）	颜色差异均值（DC₂）	D₂	$ε 2$
chaos	0.165 5	1.121 9	0.643 7	0.616 2	0.436 8	2.760 5	1.598 6	1.604 4
crackle	0.129 3	1.002 8	0.566 0	0.572 6	0.284 5	2.558 5	1.421 5	1.463 4
mosaic	0.052 5	0.570 4	0.311 5	0.271 3	0.123 7	1.226 2	0.674 9	0.744 4
shallow	0.252 1	1.340 4	0.796 2	0.705 6	0.481 3	3.172 6	1.826 9	1.904 8
snowflake	0.250 5	1.209 8	0.730 1	0.627 1	0.529 1	2.812 8	1.671 0	1.676 6
spot	0.113 4	1.229 0	0.671 2	0.624 6	0.285 8	2.923 0	1.604 4	1.741 7
stripe	0.151 5	1.267 7	0.709 6	0.619 3	0.360 1	3.060 6	1.710 3	1.603 6
watermark	0.156 6	1.158 3	0.657 4	0.493 9	0.267 6	2.318 2	1.292 9	1.575 3
wood	0.122 3	1.266 5	0.694 4	0.653 7	0.366 7	3.261 1	1.813 9	1.712 9

Tab. 7 Experimental results of color difference grading threshold

纹理类别	小色差对照组				大色差对照组
纹理类别	纹理差异均值（DT₁）	颜色差异均值（DC₁）	D₁	$ε 1$	纹理差异均值（DT₂）	颜色差异均值（DC₂）	D₂	$ε 2$
chaos	0.165 5	1.121 9	0.643 7	0.616 2	0.436 8	2.760 5	1.598 6	1.604 4
crackle	0.129 3	1.002 8	0.566 0	0.572 6	0.284 5	2.558 5	1.421 5	1.463 4
mosaic	0.052 5	0.570 4	0.311 5	0.271 3	0.123 7	1.226 2	0.674 9	0.744 4
shallow	0.252 1	1.340 4	0.796 2	0.705 6	0.481 3	3.172 6	1.826 9	1.904 8
snowflake	0.250 5	1.209 8	0.730 1	0.627 1	0.529 1	2.812 8	1.671 0	1.676 6
spot	0.113 4	1.229 0	0.671 2	0.624 6	0.285 8	2.923 0	1.604 4	1.741 7
stripe	0.151 5	1.267 7	0.709 6	0.619 3	0.360 1	3.060 6	1.710 3	1.603 6
watermark	0.156 6	1.158 3	0.657 4	0.493 9	0.267 6	2.318 2	1.292 9	1.575 3
wood	0.122 3	1.266 5	0.694 4	0.653 7	0.366 7	3.261 1	1.813 9	1.712 9

Tab. 8 Experimental results of color difference grading

类别	真实样本数	预测正确数	预测错误数	准确率/%	精确率/%	召回率/%	F1-score/%
总体	800	741	59	92.6	92.0	92.6	92.3
无色差（I类）	250	235	15	94.0	92.5	94.0	93.2
小色差（Ⅱ类）	300	276	24	92.0	90.8	92.0	91.4
大色差（Ⅲ类）	250	230	20	92.0	92.7	92.0	92.3

Fig. 4 Visualization of color difference detection grading results

Tab. 9 Color difference detection time analysis in industrial deployment

瓷砖尺寸

（mm×mm）

最大

检测时间/ms

平均

检测时间/ms

Fig. 5 Color difference detection system

References 24

[1]	刘策，刘公言，林振国，等. 手持式色差仪测定不同观赏动物的被毛的色差值［J］. 当代畜牧， 2020（4）： 24-26.
	LIU C， LIU G Y， LIN Z G， et al. Measurement of the chromatic values of the coats of different ornamental animals by hand-held colorimeter［J］. Contemporary Animal Husbandry， 2020（4）： 24-26.
[2]	BERNS R S， BILLMEYER JR P F W， SALTZMAN M. Billmeyer and Saltzman’s principles of color technology ［M］. 4th ed. New Jersey： John Wiley & Sons， Inc.， 2019.
[3]	杨坤. 基于CCD的瓷砖色差与尺寸检测研究［D］. 武汉：华中科技大学， 2007.
	YANG K. Research of color difference and dimension measurement of ceramic tile based on CCD ［D］. Wuhan： Huazhong University of Science and Technology， 2007.
[4]	刘利. 基于图像处理的瓷砖色差在线检测方法研究［D］. 长春：长春理工大学， 2019.
	LIU L. Research on online detection method of ceramic tile color difference based on image processing［D］. Changchun： Changchun University of Science and Technology， 2019.
[5]	QIAN W， YU X， XIE W， et al. Color-separation framework for ceramic-tiles based on few labeled data by HSV and SVM ［J］. Journal of Physics： Conference Series， 2021， 2044： No.012105.
[6]	吴庆涛，曹再辉，施进发. 基于改进颜色直方图和灰度共生矩阵的图像检索［J］. 图学学报， 2017， 38（4）： 543-548.
	WU Q T， CAO Z H， SHI J F. Image retrieval based on improved color histogram and gray level co-occurrence matrix ［J］. Journal of Graphics， 2017， 38（4）： 543-548.
[7]	唐海艳. 基于颜色直方图特征的印刷品颜色缺陷检测［D］. 昆明：昆明理工大学， 2020.
	TANG H Y. Detection on prints color difference based on color histogram features ［D］. Kunming： Kunming University of Science and Technology， 2020.
[8]	NADLER E O， DARRAGH-FORD E， DESIKAN B S， et al. Divergences in color perception between deep neural networks and humans ［J］. Cognition， 2023， 241： No.105621.
[9]	CHEN H， WANG Z， YANG Y， et al. Learning a deep color difference metric for photographic images ［C］// Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2023： 22242-22251.
[10]	REN J， HU X， ZHU L， et al. Deep texture-aware features for camouflaged object detection ［J］. IEEE Transactions on Circuits and Systems for Video Technology， 2023， 33（3）： 1157-1167.
[11]	WU Q， FANG Z， SONG Z， et al. A color extraction algorithm by segmentation ［J］. Scientific Reports， 2023， 13： No.21261.
[12]	QIANG H， ZHANG X， HOU J. Attention-based color difference perception for photographic images ［J］. Applied Sciences， 2025， 15（5）： No.2704.
[13]	XU Y， LI F， CHEN Z， et al. Encoding spatial distribution of convolutional features for texture representation ［C］// Proceedings of the 35th International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2021： 22732-22744.
[14]	HE K， ZHANG X， REN S， et al. Spatial pyramid pooling in deep convolutional networks for visual recognition ［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2015， 37（9）： 1904-1916.
[15]	HU J， SHEN L， SUN G. Squeeze-and-excitation networks ［C］// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2018： 7132-7141.
[16]	MO H， XU B， OUYANG W， et al. Color segmentation of multi-colored fabrics using self-organizing-map based clustering algorithm ［J］. Textile Research Journal， 2016， 87（3）： 369-380.
[17]	DAI Z M， LI D S， XIE H. The analysis for evaluation of ceramic tile’s color difference based on CIELAB color space ［J］. Advanced Materials Research， 2012， 490/491/492/493/494/495： 3726-3732.
[18]	International Organization for Standardization. . Graphic technology — process control for the production of half-tone colour separations， proof and production prints part 2： offset lithographic processes ［S］. Geneva： ISO， 2013.
[19]	XUE J， ZHANG H， DANA K. Deep texture manifold for ground terrain recognition ［C］// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2018： 558-567.
[20]	BELL S， UPCHURCH P， SNAVELY N， et al. Material recognition in the wild with the materials in context database ［C］// Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2015： 3479-3487.
[21]	CIMPOI M， MAJI S， KOKKINOS I， et al. Describing textures in the wild ［C］// Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2014： 3606-3613.
[22]	SHARAN L， ROSENHOLTZ R， ADELSON E. Material perception： What can you see in a brief glance？［J］. Journal of Vision， 2009， 9（8）： No.784.
[23]	ZHANG H， XUE J， DANA K. Deep TEN： texture encoding network ［C］// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2017： 2896-2905.
[24]	PEEPLES J， XU W， ZARE A. Histogram layers for texture analysis ［J］. IEEE Transactions on Artificial Intelligence， 2022， 3（4）： 541-552.

Quantitation and grading method for ceramic tile chromatic aberration based on improved fractal encoding network

基于改进FENet的瓷砖色差量化分级方法

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 14

References 24

Related Articles 15

Recommended Articles

Metrics

[1]	Min CHEN, Xiaolin QIN, Shaohan LI, Hao YANG, Taohong LI. Review of deep learning applications in severe convective weather prediction [J]. Journal of Computer Applications, 2026, 46(3): 980-992.
[2]	Jian ZHANG, Jianbo YU, Jian TANG. Municipal solid waste incineration state recognition method based on multilayer preprocessing [J]. Journal of Computer Applications, 2026, 46(3): 940-949.
[3]	Yongwei JIANG, Xiaoqing CHEN, Linjie FU. Elastic medical image registration model with high-frequency preservation based on spectrum decomposition [J]. Journal of Computer Applications, 2026, 46(3): 924-932.
[4]	Haoqian JIANG, Dong ZHANG, Guanyu LI, Heng CHEN. SetaCRS： Conversational recommender system with structure-enhanced hierarchical task-oriented prompting strategy [J]. Journal of Computer Applications, 2026, 46(2): 368-377.
[5]	Jincheng FU, Shiyou YANG. Short-term wind power prediction using hybrid model based on Bayesian optimization and feature fusion [J]. Journal of Computer Applications, 2026, 46(2): 652-658.
[6]	Ming LI, Mengqi WANG, Aili ZHANG, Hua REN, Yuqiang DOU. Image steganography method based on conditional generative adversarial networks and hybrid attention mechanism [J]. Journal of Computer Applications, 2026, 46(2): 475-484.
[7]	Jie CAO, Lingfeng XIE, Bingjin WANG, Changhe ZHANG, Zidong YU, Chao DENG. Adolescent scoliosis screening method considering class imbalance and background diversity [J]. Journal of Computer Applications, 2026, 46(2): 630-639.
[8]	Zeyi GUO, Fenglian LI, Lichun XU. Double decision mechanism-based deep symbolic regression algorithm [J]. Journal of Computer Applications, 2026, 46(2): 406-415.
[9]	Xiaolei CHEN, Zhiwei ZHENG, Xue HUANG, Zhenbin QU. Panoramic video super-resolution network combining spherical alignment and adaptive geometric correction [J]. Journal of Computer Applications, 2026, 46(2): 528-535.
[10]	Quanjie LIU, Zhaoyi GU, Chunyuan WANG. Unsafe driving behavior detection under complex lighting conditions [J]. Journal of Computer Applications, 2026, 46(2): 613-619.
[11]	Jinjiao LIN, Canshun ZHANG, Shuya CHEN, Tianxin WANG, Jian LIAN, Yonghui XU. Vehicle insurance fraud detection method based on improved graph attention network [J]. Journal of Computer Applications, 2026, 46(2): 437-444.
[12]	Feng HAN, Yongfeng BU, Haoxiang LIANG, Shuwen HUANG, Zhaoyang ZHANG, Shijie SUN. Vehicle trajectory anomaly detection based on multi-level spatio-temporal interaction dependency [J]. Journal of Computer Applications, 2026, 46(2): 604-612.
[13]	Kejian CUI, Zhiming WANG, Zhaowen QIU. Method for retinal vessel segmentation and coronary artery disease prediction using optical coherence tomography angiography [J]. Journal of Computer Applications, 2026, 46(2): 640-651.
[14]	Xiaoyong BIAN, Peiyang YUAN, Qiren HU. Dual-coding space-frequency mixing method for infrared small target detection [J]. Journal of Computer Applications, 2026, 46(1): 252-259.
[15]	Zhihui ZAN, Yajing WANG, Ke LI, Zhixiang YANG, Guangyu YANG. Multi-feature fusion speech emotion recognition method based on SAA-CNN-BiLSTM network [J]. Journal of Computer Applications, 2026, 46(1): 69-76.