Abstract: To extract key characteristic parameters such as the particle size, morphology and distribution state of cut tobacco inside cigarette rods, a systematic method for cut tobacco particle distribution and an extraction method for particle size characteristics were established. A micro-focus CT technique was employed to scan cigarette samples and a recognition and segmentation model for cut tobacco particles was constructed. Furthermore, based on a spatial matching algorithm for cut tobacco particles among sequential cross-sections, the three-dimensional characteristic length of cut tobacco was obtained by correlating the same cut tobacco entity in adjacent cross-sections layer by layer. Cigarette samples with a diameter of 7.80 mm and a cut tobacco rod length of 65.00 mm were studied and the results showed that: 1) The YOLOv8 model satisfactorily realized the recognition and segmentation of cut tobacco particles in cross-sections. 2) The number of cut tobacco particles at the mouth end (6 mm from the filter) and the lighting end (6 mm from the rod end) was significantly higher than that in the middle rod. The number of cut tobacco particles in different sections fluctuated between 40 and 70. Cut tobacco particles were mainly distributed within 1-3 mm from the cigarette center, while the number was relatively low within 3-4 mm from the center, indicating a sparse distribution near the cigarette paper. 3) The width of cut tobacco particles was concentrated in the range of 0.72-1.39 mm, with the coefficient of variation ranging from 25.77% to 31.11%. The thickness was concentrated within 0.09-0.18 mm, with the coefficient of variation ranging from 15.32% to 26.48%, showing good overall uniformity. 4) Cut tobacco particles with three-dimensional length of 3.00-5.70 mm accounted for a large proportion, while those longer than 11.10 mm only accounted for 10% of the total. 5) Taking cigarette sample 2^# as an example, the measurement deviation between this method and the conventional cut tobacco length detection method was 3.0%, and the distribution trends of cut tobacco particle length obtained by the two methods were in good agreement, which verified the accuracy of the proposed three-dimensional length detection method for cut tobacco particles. This study provides technical support for the calculation of characteristic size of cut tobacco particles.