Design of Machine Vision Imaging Color Correction Method Based on BP Neural Network

In the process of using machine vision system instead of human eyes to detect the product color, the color of the image collected by the machine vision imaging equipment has a certain degree of nonlinear distortion compared with the actual color of the object. Therefore, before the color difference detection, it is necessary to correct the color. At present, a variety of BP neural networks based on algorithm optimization can be used for color correction, with excellent performance, but the calculation cost is too high, the performance evaluation index is not unified with the product detection index, and only for a small number of or specific colors, these characteristics make it difficult to carry out real-time color difference detection in industrial automation detection. Based on the above problems, this paper uses BP neural network to carry out color correction under the data set of collecting Pantone color card as a sample, and mainly studies the influence of the full color domain data set and its divided color domain data set on the correction results in RGB and lab color spaces. The experimental results show that under the sample and model design method, the BP neural network based on RGB full color gamut model has higher accuracy and more stable performance for color correction of machine vision imaging, which is suitable for real-time color correction and color difference detection in industrial automation detection.