Skip to Main Content
Taking the universal tools microscope (UTM) as an example, this paper proposes an original solution to the automatic image recognition of ocular optical measuring instruments based on algorithm fusion. An area-array CCD is used as the image collection device and a series of image pre-processing methods are adopted to locate the reticles and digit characters in ocular lens view images. A two-layer image recognition model which bands together the correlation-based template matching and an optimized BP concurrent neural network is established. The method featured as multiple complementary extraction is used in generating eigenvectors of the network. The experiment result shows the processing speed of the automatic reading method is enhanced on the basis of exerting the advantages of the high recognition ratio of neural network.