Table 2 Specific applications of intelligent techniques and optimisation algorithms in the field of colour difference detection and assessment

SVM, PCA and GA

This research focused on developing a colour difference evaluation model for dyed fabrics using SVM, in which the researchers used Principal Components Analysis to select several independent indicators from training and test data, and optimised its parameters through a GA

The proposed model was compared to the traditional Naive Bayesian algorithm and was found to provide a 9% improvement in prediction accuracy and a 0.0985 reduction in relative error

Zhang and Yang (2014)

Improved LSSVR algorithm and GM (1,1)

This research aimed to achieve colour constancy of textured textiles by combining an improved LSSVR algorithm and GM(1,1)

The results showed an increase in prediction accuracy was achieved when the improved LSSVR algorithm was combined with the improved GM(1,1) model

Zhang et al. (2017)

ANN

This research aimed to detect the colour difference of fabrics by extracting the colour eigenvalues of fabric images

The accuracy of the colour difference detection exceeded 93%

Li et al. (2015)

Bagging-PSO-ELM

This research aimed to develop an accurate light correction model for dyed fabrics by using a hybrid of an ensemble ELM mechanism based on Bagging and PSO, which is Bagging-PSO-ELM, to avoid light variations that can lead to severe colour difference evaluation errors

The results indicated that the best predictive performance can be obtained with Bagging-PSO-ELM

Zhou et al. (2016)

T–S FNN algorithm

This research focused on developing a mapping relationship between RGB space and CIE L*a*b* space using the T–S FNN algorithm to simplify the colour space conversion process. The block method was also used to identify colour differences in dyed fabrics with a wide range of viewing angles

The predicted results were in accordance with the spectrophotometer measurements, demonstrating the effectiveness and feasibility of the T–S FNN algorithm for the real-time colour detection process

Li et al. (2017)

DE-OSELM, SVR and ELM algorithms

This research proposed a DE-OSELM algorithm based on a rotating forest framework for solving textile colour correction

The developed algorithm reduced the mean squared error (MSE) by 10.3% and 7.8% compared to textile colour correction algorithms based on support vector regression (SVR) and extreme learning machine (ELM) algorithms, respectively, indicating strong robustness and excellent prediction capability

Zhou et al., (2019a, 2019b, 2019c)

DE-WOA-ELM (WOA optimized by the ELM of DE)

This research proposed a colour-difference classification approach using DE-WOA-ELM

The average classification precision of the dataset was enhanced by 2.15%, 11.06%, 12.11% and 0.47% in comparison with ELM, SVM, BPNN and KELM, respectively

Zhou et al., (2019a, 2019b, 2019c)

RF-PSO-SLSSVR, RF, PSO, and SLSSVR

This research proposed an RF-based ensemble PSO and SLSSVR algorithm to improve the accuracy of light correction models. Artificial intelligence techniques used: Random Forest (RF), Particle Swarm Optimization (PSO), and Sparse Least Square Support Vector Regression (SLSSVR)

The RF-PSO-SLSSVR algorithm reduced the root mean square error (RMSE) of angular by 13.6% and 10.6% compared to the conventional SVR and ELM algorithms

Zhou et al., (2019a, 2019b, 2019c)

DE-GOA-KELM, which is a KELM using an improved GOA

This research investigated the classification of colour differences in dyed fabrics. Artificial intelligence techniques used: DE-GOA-KELM, which is a KELM using an improved GOA

The proposed DE-GOA-KELM model had an average classification accuracy of 98.89%, while the accuracy of the KELM was only 91.08%

Li et al. (2020)

Improved K-means algorithm

This research focused on developing a colour difference online detection algorithm for fabrics

The results showed that the developed algorithm was able to meet the real-time requirements of the system, and the detection speed increased from 0.2 to 0.8 m/s in line with the sampling

Xie et al. (2020)

YOLO (you only look once) convolutional neural network

This research proposed a real-time inspection system using the YOLO convolutional neural network to detect common defects and colour differences in warp-knitted fabrics

The proposed method performs well in real-time and is highly accurate, meeting the fabric inspection requirements of warp knitting factories

Xie et al. (2021)

The SCA-MPA-RVFL algorithm

This research evaluated the colour constancy of dyed fabrics by using the SCA-MPA-RVFL algorithm to eliminate the effect of light variation on the colour difference classification of dyed fabrics

The results of the study indicated that the SCA-MPA-RVFL algorithm was effective in restoring the images of dyed fabrics close to those under standard illumination, thus evaluating the colour constancy of dyed fabrics well

Liu and Yang (2021)

DPS-DE-RELM

This research proposed a new optimization technique called DPS-DE-RELM for developing colour difference classification models for dyed fabrics

The proposed method showed faster convergence, high classification accuracy, and robustness. The maximum classification accuracy achieved was 98.87%

Zhou et al., (2021a, 2021b)

FWSVR and PSO

This research proposed an approach for predicting recipes for factory dyeing

The results showed a small average colour difference between the desired colour and the re-created colour; the Mean Absolute Percentage Error (MAPE) in the predicted concentration was less than 5%

Li et al. (2022)

improving the HGS-optimized RVFL algorithm using GWO

This research aimed to classify the colour differences in dyed fabrics by improving the HGS-optimized RVFL algorithm using GWO

The results showed that the proposed algorithm had good stability and fast convergence

Zhang and Zhou (2022)