| As the government vigorously promotes the construction of urbanization, the gas demand is continuing to grow. It is essential to change the existing gas meter on issues like the meter reading methods and billing fees. The existing impulse type reading and IC card reading have been far from meeting the needs of convenient settlement of natural gas. But, there have been many different ways of remote meter reading systems, but these are structured by changing the internal structure of the gas meter. On the one hand, it causes a waste of resource and increased costs. On the other hand, the various types of meter reading system lead to a lack of product versatility. And the reliability and flexibility of the meter reading technology are lack of adequate technical support, which is not widely used in the market.Machine vision has the characteristics of high efficiency and highly automated, which can achieve high precision and speed. It has become a new hotspot to implement a meter-reading method of direct reading based on machine vision. This system not only avoids the waste of resources, but also can ensure the reliability and safety of meter-reading method. Under the premise of without changing the structure of the original gas meter, designed a gas meter remote meter reading system based on machine vision. The system collects gas meter image through image acquisition equipment. The main work of this thesis includes the following aspects:(1) This thesis analyzed the research status of character recognition technique and direct-reading remote smart meter. The functional requirements of meter reading system was analyzed, in view of this analysis, we completed the overall design of the meter reading system.(2) The pre-processing technology of the gas meter image was analyzed, every single character image of gas meter reading with high quality were obtained, which provided a good data source for the next character recognition.(3) A combination of multiple feature extraction methods was used to extract character data for reducing the complexity of systems, and then using these characteristic data to represent each character image. The rough set attribute reduction algorithm optimized by genetic algorithm was studied, and a large amount of redundant data in the feature data were deleted, which further reduced the complexity of the system.(4) The character recognition algorithm based on neural network optimized by genetic algorithm was studied to improve recognition rate of the character image for the gas meter, so that the recognition system had good fault tolerance, strong self-organizing ability and interference suppression.(5) The design of hardware circuit and software procedure for the gas meter remote meter reading system was completed, and the system was tested. Test results show that the system had the characteristics of simple structure, stability, high recognition rate, safety and convenience.Finally, this thesis summarized the work in this research, and prospected the future work. |
PDF Full Text Request