Research On The Recognition Technique Of Meter Parameter Symbols

The recognition of meter parameter symbols is to recognize all the symbols on the dial plate except the areas of dial pointer and scales. These meter parameters are scattered in any position of dial plate except the areas of dial pointer and scales. They are not fixed and arranged orderly like automobile registration numbers and ID card numbers. There are various kinds of parameters on the dial plate: accuracy grade, meter name, meter type, unit, measurement range, grade of defending the magnetic field, standard signal scope, produce date, factory serial number etc. contents. Their positions are not regular on the dial plate. Some meter parameter symbols are special signs. So the current OCR software can't identify them. This thesis is to research the segmentation and recognition of meter parameter symbols. These meter parameters symbols have to be recognized and defined at the time of measuring in order to realize the reading, calibration, and management of meters automatically.The main research results and innovative points of this thesis are as follows:1. An algorithm is presented to enhance the contrast of meter image based on single scale Retinex and bilateral filter. This algorithm strengthens the information hidden in the large scale shadow of meter image efficiently.2. Two illumination correction methods based on the simulation of the background of the image are put forward: piece sampling and reformed Top-hat transformation.3. A new scheme of locating dial plate parameters area indirectly is put forward. This design includes two kinds of methods. One is to detect and delete the arc range using the combined algorithm of tracking edge and fan-shaped scan. Another is to choose the paramrters area based on dial information and texture analysis.4. Based on the combined features of characters such as connectivity, character central distance, symmetry etc, the meter accuracy grade and parameter symbols are segmented efficiently.5. Based on the combined features of Hu moment & compactness and support vector machine, the meter accuracy grade and parameter symbols are recognized, the algorithm of this paper improves recognition rate of meter parameters symbols efficiently. It lays the foundation for realizing the reading, calibration, and management of meters automatically.