Regression prediction model of fabric brightness based on light and shadow reconstruction of layered images

The traditional fabric colour matching process has problems such as long cycle, low efficiency, and high cost. To address these limitations and improve the accuracy of fabric brightness prediction, this research combines the theory of light and shadow reconstruction of layered images, multiple regression (MR) prediction, and a neural network to build a fabric brightness regression prediction model. The primary objectives include enhancing the visual representation of fabric colours and establishing a high-precision predictive framework. First, the layering process of the fabric image was determined by the light and shadow reconstruction method, and then a series of factors affecting the brightness value were set as independent variables and the brightness value as the dependent variable on the basis of the layered image by combining the MR model to build the brightness prediction model. Finally, the prediction model was optimised by incorporating a backpropagation neural network. This hybrid method ensures interpretability and nonlinear adaptability through MR prediction and neural networks. Testing the model performance, it is found that the final fabric brightness regression prediction model has better prediction performance, and its error performance is better than that of the traditional multivariate regression model. In practical applications, the proposed model can obtain 95.3 user satisfaction and 97.8 merchant satisfaction, and the difference between its predicted and actual brightness values is within 3°. These results demonstrate the model’s potential for industrial adoption in digital textile design. Future research will extend the model to dynamic lighting conditions and multi-material fabric interactions.