Regularized Extreme Learning Machine Based on Remora Optimization Algorithm for Printed Matter Illumination Correction

In the process of printed matter quality inspection, the influence of lighting conditions leads to a large color difference between the captured image and the actual image. Therefore, this paper proposes an illumination correction model based on Remora Optimization Algorithm for Regularized Extreme Learning Machine (ROA-RELM). First, the algorithm extracts the statistical features of the printed matter image using the Grey-Edge framework. Next, the statistical features are used as inputs to the Regularized Extreme Learning Machine (RELM). Then, the remora algorithm (ROA) is used to optimize the weights w and bias b of the input layer of the RELM, which improves the prediction accuracy. At last, the image is corrected by color constancy algorithm based on the output. In order to verify the effectiveness of the algorithm proposed in this paper, the method of color difference detection is used in this paper to evaluate the effect of the illumination correction algorithm on the detection of the color quality of printed matter. The experimental results show that the algorithm mentioned in this paper can effectively improve the presentation of printed matter images, and the angle root-mean-square error is reduced by 11.04% compared with the control group's illumination-corrected model.