Research on Color Correction Algorithm for Aged Tobacco Leaves Based on Deep Learning

Tobacco leaf maturation is a critical process in the storage of raw materials within the tobacco industry, as the degree of maturation directly impacts the quality of cigarettes. During this process, regular color observations and grading are required. However, these assessments can be easily affected by environmental lighting, leading to inaccurate judgments. To address this issue of lighting interference in tobacco leaf color assessment, a deep learning-based method is proposed for detecting color under varying lighting conditions during the maturation process. This method constructs a color correction model (CCBS) based on SqueezeNet, obtains real lighting values using a standard color card, and employs a diagonal model to eliminate lighting interference, producing tobacco leaf images under standardized lighting conditions. To evaluate the performance of the proposed algorithm, a series of experiments were conducted using angular error as a metric. Experimental results show that the proposed method achieves the lowest value across all evaluation metrics and achieves high-precision detection without increasing the model size. This method offers strong support for real-time acquisition of the true color of tobacco leaves and for accurately assessing their maturation degree.