Research On The Appearance Detection Problem Of The Transformer Verification System

In recent years, the state grid corporation of china is sparing no effort to carry out the construction of automatic system in measuring infrastructure. At present, more than ten provinces and cities have implemented the automatic verification system for low-voltage current transformer which establishes a standard certification center combining warehouse storage, calibration of metering equipment and logistics distribution.Hence, the processes of handling the transformers, such as output and input of warehouse,transmission, wiring, calibration, are all made automatic. However, in actual operation, the transformer in containers may encounter some problems, such as vacancy, reversal,dumping, inconsistent specifications, etc. This system couldn’t detect and solve these unconventional problems which not only affects the working efficiency, but also damages the transformer, such that great direct economic losses are caused.Under the background above, this thesis designs the overall structure and detection algorithms of transformer appearance detection are designed, and realizes the software system on the basis of learning the transformer automatic verification system and the relevant technologies about machine vision appearance detection system. First, the thesis introduces the current transformer automatic verification system and the existing problems,and realizes the function requirements of transformer appearance detection. Second, the overall structure and workflow are both designed. Third, the solution and algorithm design schemes of the transformer have-or-not detection, reversal detection, dumping detection,current ratio character recognition are provided respectively. It uses the hue histogram and SURF image matching algorithm to detect the states of transformer, and apples character recognition based multistep matching algorithm to recognize current ratio characters. The software development for the detection of transformer appearance has finishied in the MATLAB environment, achieving the transformer have-or-not detection, reversal detection, dumping detection and current ratio character recognition.