Abstract: To improve fresh tobacco leaves grading before tobacco leaf curing, a method of image detection and classification of fresh tobacco leaves was presented based on the machine vision technology for classifying tobacco leaves with different maturities. An automatic detecting and grading device of fresh tobacco leaves was designed. The mechanical structure of the system included four parts, namely, automatic sample grasping structure of tobacco leaves, tobacco leaves’conveyor platform, machine vision detection platform, and sorting platform. The automatic loading of tobacco leaves were detected by a CCD which was on the conveyor belt. The image of tobacco leaves was processed and an integration algorithm of image fast processing was proposed. First of all, the combination method of neighborhood mean and median filtering was used to denoise the noise of the image. The minimum error threshold segmentation method was used to separate the background and the tobacco leaves, and then the image was enhanced for color value getting. Four values (R, G, B color value and hue value) of the sample tobacco leaves were calculated and used to represent tobacco leaves’characteristics. Four characteristic values of the sample tobacco leaves were used to build sample library and then was remembered by the device. The unknown sample was tested and graded with reference to the sample library. The results showed that the classification accuracy was determined by the color difference of adjacent classified tobacco leaves. The greater the difference, the more accurate the classification was. The average speed of the device detection was 2-3 seconds per piece, which could meet the requirements of the on spot detection.