Digital electric meter reading self-adaptive identification method based on deep learning

Abstract

The invention discloses a digital electric meter reading self-adaptive identification method based on deep learning, comprising the following steps: acquiring a mechanical electric meter, a nixie tubemeter and a non-intelligent electric meter to establish an electric meter image data set, generating a standard data set of a Pascal Voc system, and adopting a Faster-RCNN algorithm to train the generated standard data set to complete the number identification and frame selection work of the electric meter, using the convolutional neural network for predicting category scores of a series of candidate frames on the feature map by using a convolution kernel, and using the category with the highest score as an output result to complete adaptive identification. The method can achieve the adaptivepositioning of digital regions of different types of electric meters, and achieves the recognition of characters in the digital regions through a character recognition algorithm after image preprocessing.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to an adaptive recognition method for digital electric meter readings based on deep learning. Background technique [0002] At present, there are still many infrastructures such as base stations and computer rooms of communication operators using mechanical meters. It takes a lot of manpower to copy the meter line codes every month, which is inefficient and error-prone. The existing technology object detection network relies on the area proposal algorithm to assume The location of the object, since the meter image is uploaded by the meter reader using a mobile phone to take pictures of the meter, there may be situations such as blurred photos and too dark light, which have a great impact on recognition. Contents of the invention [0003] The technical problem to be solved by the present invention is to provide an adaptive recognition method for digital electric...

AI Technical Summary

Problems solved by technology

[0002] At present, there are still many infrastructures such as base stations and computer rooms of communication operators using mechanical meters. It takes a lot of manpower to copy the meter line codes every month, which is inefficient and error-prone. The existing technology object detection network relies on the area proposal algorithm to assume The location of the object, since the meter image is uploaded by the meter reader using a mobile phone to take pictures of the meter, there may be situations such as blurred photos and too dark light, which have a great impact on recognition

Images