In order to segment the wear particle area accurately for calculating the ferrographic abrasive content in the oil, and finally obtain equipment oil products and fault information, taking the wear particle image of the image-visual online ferrography sensor as the object, the region-based geometric active contour models based on curve evolution theory and level set method: LBF model and IR model were introduced. By comparing the segmentation effects of two models, it is found that the IR model has higher segmentation accuracy and faster convergence speed. The influence of model parameters on the segmentation results was further analyzed, and the segmentation parameter values corresponding to the optimal segmentation effect and the shortest computing time under different wear particle concentration conditions were obtained, which provided a basis for the adaptive selection of wear particle image segmentation parameters in online monitoring. The experiment shows that the segmentation model adopted has higher segmentation accuracy and convergence speed for the wear particle image, which provides a guarantee for the subsequent fast and accurate calculation of the ferrographic wear particle content in the oil.