Abstract: In natural environments, the maturity of tobacco leaves in the field is difficult to recognize due to the influence of light. To solve this problem, a fuzzy illumination-based method for identifying the maturity of tobacco leaves in the field is proposed. Firstly, a convolutional neural network segmentation model is used to extract regions of interest in tobacco leaves; Secondly, a segmented model of the tobacco leaf area is constructed to build a fuzzy nonlinear relationship between lighting and tobacco leaf color information. The role of this segmented model is to eliminate the influence of lighting. Next, the prior knowledge of colors at different maturity levels is calculated. Based on the prior knowledge, establishing a fuzzy relationship between yellow and green in the natural environment infers the color attributes of tobacco leaf pixel points, and calculates the yellow area. Finally, a membership probability relationship between the yellow area and the maturity of fresh tobacco leaves is constructed to calculate the maturity of tobacco leaves. For fresh tobacco leaves from the middle and upper leaves of Yunyan87 and Zhongchuan208 in Sichuan region, this research method achieves maturity classification accuracy of 82.0%, 77.0%, 75.0%, and 71%, respectively, which is generally better than ELM, SVM, and BP neural networks. The results show that the proposed fuzzy illumination-based method for determining the maturity of tobacco leaves in the field could effectively overcome the influence of lighting, accurately determine the maturity of tobacco leaves in different field environments, and provide a theoretical basis for the visual system of tobacco intelligent collection equipment.