Research on Digital Instrument Segmentation Based on Improved Maximum Entropy Algorithm

In the visual recognition technology of digital instruments and meters by electric inspection robot, when extracting the state information contained in the unit image of the device, the foreground part containing the state information must be segmented from the background part of the image containing various complex features by the image segmentation algorithm. In the process of threshold segmentation for the unit image of the device with various problems, such as voice interference, small-scale spot or shadow interference, uneven overall illumination, low overall image contrast and so on. Although image enhancement algorithms can suppress various interference factors caused by illumination, but which cannot be completely eliminated. This paper first analyzes and compares Otsu threshold segmentation algorithm, Bernsen threshold segmentation algorithm, Niblack threshold segmentation algorithm and Sauvola threshold segmentation algorithm. Then, based on the digital instrument image features processed by image enhancement algorithm, this paper proposes parameter entropy measure to find the best threshold of image segmentation. Finally, the proposed method is compared with other algorithms.