Research on HVPL for visual detection of bicolor water level gauge

To obtain real-time and accurate value of water level for industrial boiler drum, image preprocessing and number identification algorithms are used on FPGA, including distortion correction, DoG filter (Difference of Gaussian filter), the minimum misclassification error threshold method and crack chain code. Especially, the algorithm HVPL is proposed. Segmentation, extraction and number identification are fulfilled for the level gauge image by using above algorithms. Experiment shows that performing the machine recognition for image by running the fast algorithms on logic circuit has advantages of good real-time performance, robustness and higher recognition rate.