Image Processing System For Quality Monitoring Of The Water Meter

This paper deals with the design and implementation of an image processing system for water meter quality monitoring. For some large-scale water meter manufacturers, tens of thousands of meters are produced every day. The inspection of the water meter is a labor-intensive heavy workload. So it is necessary to develop a quality monitoring system with higher degree of automation and reliable calibration results to realize the automatic inspection of the water meter and to achieve a computerized management of the calibrated data. This is of great value to improve the efficiency and cost-effectiveness of the water meter manufacturing enterprises.In this paper, based on the deeply study of the process in the quality verification of the water meter, we proposed the use of image recognition method to obtain water meter readings in order to achieve the automation of the verification. Specifically, the main contents are included as follows,1. Design an automatic calibration system for examination of water meter and improve the existing water meter calibration device by study the process of the verification for the water meter. Design a dual-use calibration control system of both computer control and manual control. Write code for the auto-approval procedure and the information management procedure. Realize the automation of calibration and information of the management.2. Design an image processing system of the water meter. For the problems in the actual collection of water meters, such as the image geometric distortion and that the brightness changes greatly, we employ the geometric transformation and positioning to determine the various recognized sub-field image. And then with the image segmentation based on edge detection and the corresponding noise removal method, we extract the character image pixels as the edge pixels.3. Achieve the water meter image recognition. For the character recognition in the water meter image recognition, we study the specialty and recognition difficulty of the coded characters and the wheel characters. Considering the diversity of the coded character, we build the BP network to identify the coded character under a variety of fonts; According to continuously rotating properties of the mechanical meter and the associated rotating model between the wheel characters, we use the relevant template matching method to identify the incomplete wheel character. Meanwhile we give a quantitative description about the last moment of the corresponding character wheel. Finally, we determine the corresponding number for each character and obtain the reading value of the water meter by the associated rotating models between the character wheel and the pointer.