Since the image monitoring of engine lubricating oil wear particles is only suitable for micro flows and easy to be disturbed by bubbles， an optical image online monitoring system for oil wear particles suitable for full flow working environment is designed to distinguish bubbles and wear particles. Through this system， a large number of wear particle and bubble images are collected. A moving object extraction algorithm based on background differences and the Otsu method is used to extract wear debris and bubble image samples. The bubbles and particles are identified by the feature extraction based on the histogram of oriented gradients （HOG） and support vector machine （SVM） classification algorithms. The experimental results show that this system can effectively collect images of wear particles and bubbles. Compared with the traditional morphological feature extraction algorithm and the K-nearest neighbor （KNN） classification algorithm， the HOG-SVM algorithm has a higher classification accuracy of 83.8%.