Pointer type instrument reading identification method based on deep learning

Abstract

The invention discloses a pointer type instrument reading identification method based on deep learning. The method comprises the following steps: identifying a dial area in a to-be-identified instrument picture through a first deep network model; intercepting an area where a dial plate of the instrument to be identified is located from the picture; identifying a circle center coordinate, each scale line number coordinate, a semi-pointer tip coordinate and coordinates of two scale line numbers located at two sides of the pointer in the dial plate through a second depth network model; respectively intercepting the areas where the two scale line numbers are located from the intercepted picture of the dial area of the instrument to be identified; and identifying numerical values of the two scale line numbers through a third depth network model, and reading by adopting an angle method. According to the pointer type instrument reading identification method based on deep learning, the readingnumbers of multiple types of pointer type instruments can be identified at the same time, and the reading efficiency is higher.

Description

technical field [0001] The invention relates to the technical field of intelligent identification of instruments and meters, in particular to a method for recognizing readings of pointer instruments based on deep learning. Background technique [0002] As a measuring instrument, the pointer instrument has the characteristics of simple structure, low price, dustproof and waterproof, and strong anti-electromagnetic interference ability. It is widely used in power system, petrochemical, aerospace and other fields. At present, most pointer instruments basically rely on human eyes for reading and reading. In the process of human eye recognition, the accuracy is not high, the workload is heavy, and the efficiency is low. In addition, it is not suitable to use human eyes in high temperature, high pressure, and high radiation environments such as substations. Therefore, it is necessary to study the use of machines instead of labor to identify the readings of pointer instruments. ...

AI Technical Summary

Problems solved by technology

In meter positioning, Hough transform circle detection and feature matching methods are inefficient and slow

The central projection method, subtraction method and region growing method in the extraction of dial elements have high requirements on the acquisition of instrument images, are easily affected by the shooting environment, and have poor robustness, while the Hough transform and least squares method have a large amount of calculation and high error , resulting in slower fetching of pointers

Among the methods for interpreting instrument representation numbers, the error of the angle method is larger, and the calculation amount of the distance method is larger

In addition, the inventor also found that the current meter reading recognition methods are only for reading recognition of one type of meter, and cannot recognize multiple types of meters at the same time.

Images