Objective: To solve the problem of appearance defects of Baoyu-flavor-slices, such as edge damage, internal porosity and uneven wrapper thickness in the production process. Methods: An on-line detection method based on machine vision was proposed. The Baoyu-Flavor-Slices were arranged into a single layer array by mechanical carding device. After image acquisition, image segmentation, gray value stretching and contour edge extraction were selected for image processing. Edge damage detection was completed by using the feature of outer contour roundness, abnormal wrapper thickness detection was accomplished by measuring wrapper thickness, by calculating the porosity area, the porosity detection was completed. Results: The detection rate of this on-line detection method was 100% on edge integrity defect, 100% on abnormal thickness defect and 98.65% on porosity defect. Conclusion: The method has feasible application and can realize the online detection of defective products of Baoyu-flavor-slices.