Research On Automatic Calibration System Of Cold-Water Meter Based On Machine Vision

The accuracy of the water meter is of great significance for energy saving and consumption reduction,our country requires that the water meter must conduct quality inspection before entering the market.At present,most of the domestic water meter verification stations still adopt manual verification as the dominant method.Long-term repetitive work and the influence of personal subjective factors make the efficiency and accuracy of water meter verification unsatisfactory.Therefore,developing an automatic water meter verification system to replace manual verification has become an urgent demand of water meter verification industry.In this paper,taking the cold-water meter verification of a domestic water meter verification station as the research object,a set of automatic verification system of cold-water meters based on machine vision is designed,which achieves the automatic verification of five different calibers of coldwater meters.Firstly,according to the performance indicators of the verification system,the hardware platform and software platform of the system are designed respectively.Through the contrast experiment,the selection of visual devices is completed,and a lighting scheme for surface image acquisition is designed,which ensures the stability of illumination environment in the process of image acquisition and completes the hardware vision platform for cold water meter verification.Then,the pointer and wheel recognition in the cold-water meter are studied respectively.The recognition of pointer reading adopts Hough gradient circle detection to locate the area of sub-dial,through the least square method to fit the axis of pointer rotation,and three concentric circles are created in turn with the axis of pointer rotation as the center.By calculating the intersection point of the pointer contour and the concentric circle,the slope of the pointer rotation can be obtained and converted into the corresponding rotation scale value,thus achieving the accurate interpretation of the pointer readings of different calibers of cold-water meters.The recognition of characterwheel reading,based on the transfer learning method,uses the pre-trained convolution neural network to train and classify the created data set,which achieves the accurate recognition of the character-wheel and effectively solves the problem of misrecognition of mixed numbers in the wheel.Finally,a set of automatic cold-water meter verification software based on machine vision is developed on Microsoft Visual Studio 2010 platform to realize human-computer interaction and historical data preservation and analysis.The automatic water meter verification system has been successfully applied in a water meter verification station.Field experiments prove the feasibility of the proposed algorithm,which is obviously superior to the traditional manual verification method in verification efficiency,and has good application and popularization value.