To evaluate quantitatively the uniformity of image illumination for an On-line Visual Ferrograph (OLVF)imaging system, a new mathematical model of image illumination was presented. The view field of an object space was represented by image space parameters and optical magnification, then it was discretized into a micro area. By taking Lambert-cosine law as reference, the image illumination model based on the aperture of entrance pupil was established, the image illumination distribution was calculated reliably and the uniformity of image illumination was evaluated successfully. On the basis of Matlab simulation analysis, the number of single Light Emission Diodes(LEDs) in the circular ring array source of the OLVF imaging system was obtained and the optimized values of focal length and optical magnification were determined. The light loss of imaging system and the illumination distribution of wear debris deposition region were analyzed and calculated under the condition of the oil cavity filled with lubricant oil. The relationship between the absorptivity of lubricant oil and the peak value of CCD image illumination was investigated. The experimental results show when the luminous intensity of LED arrays is known, the imaging plane illumination non-uniformity is more than 5.60% by simulating calculation and that is from 8% to 9% by a practical test, which satisfies the demand by indicator less than 10%. Finally, this imaging system was practically used to capture the reflected ferrograph. The results show that the characteristics of wear debris has been clearly distinguished from the reflected ferrograph, and it is favorable to the ferrograph segmentation and image extraction. It concludes that the proposed model evaluates quantitatively the uniformity of image illumination, optimizes systematical structure and improves the imaging performance of the system.