An online machine vision defect detection method based on two-stage image acquisition structures is proposed for the online detection of the hard resin lenses. With the two-stage image acquisition structures, this method is able to increase the image processing speed by improving the image acquisition speed and reducing the image data amount. With the usage of image processing tools, the defects are rapidly identified according to the characteristics of filling degree and position, and the sizes of defects are determined based on the different measurement accuracies of the two-image acquisition stages. Then the lenses are classified according to the obtained identification and sizes. The experimental results show that this method can meet the requirement of online real-time detection for resin lenses and has a good classification result.