Application Of Image Processing Technology On Automatic Water Meter Calibration Device

The record of water meter is all to depend on the manual methods at present. Heavy workload is consumed, and drawback in management is inevitable. Water meter can be automatically read by applying image processing and artificial intelligence technology. To realize intelligent processing, an improved method based on color image was applied for detecting and extracting automatically the pointers and digital display area of water meter.The content of the research work in this thesis is described as follows:1. Target location: A new location algorithm based on color information was proposed. Possible targets were firstly extracted by training. A new labeling algorithm was used to delete large-area interference regions. And then region labeling was used to choose the interesting targets. According to the dial-center-point, the interested pointer subimage and digital display area subimage were extracted in the last step of the location.2. Pointer detection and recognition: The threshold of every pointer subimage was calculated by Ptile combined with color features. According to association of pointer indication, improved template matching method based on the several characteristic parameters of the pointers was used to recognize the indicating values of the pointer.3. Digit extraction and interpretaion: Aimed at the problem brought by unknown illumination condition, fast two-dimensional Otsu algorithm was used to segment characters. To avoid the difficulty of constructing template because of different stroke width, a new template matching method was proposed based on key skeleton features of character. The deformity character was divided into two sub-images, and was extended to the size of one full character. No need to reconstruct template image, identical method associated with height ratio was used to recognize deformity character.4. Recogniton system design: Automatic water meter calibration system was operated in a VC++ environment. The system was designed to achieve indication detection and recognition automatically for water meter images.The experimental results have showed that the proposed method is more effective and stable than traditional method in practice . It is of better adaptability for low contrast image.