Automatic recognition method for lightning arrester gauge reading

Abstract

The invention discloses an automatic recognition method for lightning arrester gauge reading, which is characterized in that recognition for a lightning arrester gauge is divided into region segmentation and reading recognition; firstly, a gauge dial is segmented into multiple regions by using a connected region detection algorithm and a rectangle fitting algorithm; then the regions are detected by using prior knowledge, and connected regions representing a pointer region and a digital region is reserved; then minimum area rectangle fitting is performed on the two connected regions to acquire two rectangles with a deflection angle; rotation correction is performed on an image according to the angle, then the pointer region and the digital region are segmented from the corrected image; and finally, reading recognition is performed on the pointer region and the digital region by respectively using an angle method and a convolutional neural network method. According to the automatic recognition method, pointer reading recognition and digital reading recognition can be accomplished simultaneously, effective correction can be performed on the image, and the reading accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of image recognition, in particular to an automatic recognition method for instrument readings of lightning arresters. Background technique [0002] The arrester is an electrical device specially used to limit the lightning overvoltage or operating overvoltage in the substation. It contains a valve whose resistance value changes with the voltage. Under the rated voltage, the resistance value of the valve plate is very large, which is equivalent to an insulator. At this time, the current value flowing through the valve plate is small and stable; when the voltage at both ends exceeds the threshold value, the valve plate is turned on, and there will be a large The current passes through the valve plate and then pours into the ground, thus avoiding the impact of large current on other parallel equipment, so as to complete the overvoltage protection; when the voltage recovers, the state of the valve plate returns...

AI Technical Summary

Problems solved by technology

The BP neural network method can continuously learn and modify parameters at all levels during the training process, so as to achieve an excellent classification effect on training samples; but it relies too much on the selection of input feature vectors

In addition, the digital instruments currently being studied do not need to be divided into areas, and the focus is on classification and identification; while the digital areas in the arrester instrument are small and need to be divided before processing, so the existing methods need to be improved.

Images