Research And Application Of Digital Recognition In Intelligent Remote View Of Two - Color Water Gauge

Industrial boiler is a special equipment of pressure vessel, and which safe operation is very important. Drum water level is one of the operating parameters to determine normal running, so a reliable method of measurement and monitoring plays a key roles. Method of image remote viewing often has been used for supplementary means to conventional transmitter monitoring, which intelligence improved can reduce the operational personnel workload and avoid the subjective error brought by manual reading.In this paper, based on digital image processing to achieve intelligent remote viewing automatic monitoring for bicolor boiler water gauge. Firstly, smoothing image noise and highlighting region information by the Wiener filter. Secondly, filtered image is preprocessed though using binarization. Projection algorithm is used to segment water level of liquid column and scale line, and digital samples are gotten by positioned liquid line and segmented level digital. Finally, digital sample is identified and the water level meter is calculated according to the liquid level detection algorithm.Digital recognition is studied in three areas includes feature extraction, recognition algorithm and classifier design. A new method of extracting features based on the average cross features is proposed. The digital sample is calculated by curve fitting in spatial characteristics,3spaced vertical projection feature and5interval horizontal projection feature. Fitting function is found at the crossover point of three-dimensional feature space, and combination weight of feature can be gotten by normalized. The minimum distance template matching algorithm is proposed based on improved BP neural network and image matching recognition algorithm, and the method is high efficiency. The digital recognition classifier is designed though combing classification structure, performance, combined classifiers and identify principles. Finally, experimental has been validated based on MATLAB and VC. The results showed that: the average cross-digital feature algorithm is superior to single feature. Recognition rate get to100%based on the minimum distance classifier and secondary classification of BP classifiers, which has good robustness.