Research on Water Meter Digital Recognition Method based on DBNET-CRNN Framework

Abstract

With the prominent drawbacks of traditional manual meter reading mode, the automatic recognition of water meter digits with high efficiency and high accuracy using image processing technology has become one of the issues that more and more researchers pay attention to. In order to more accurately and effectively count the groundwater representation numbers in natural scenes, this paper presents a detection and recognition algorithm for water meter digits based on DBNET-CRNN framework. First, we design text detection model recognition text box based on DBNET network. Next, we use a text recognition algorithm based on CRNN network to predict the text box area and identify the final result. In addition, we have independently analyzed the half-word to calibrate the results, which further improves the accuracy of the model. The experimental results show that the model achieves good results in the water meter number recognition task. It can be applied to water meter number detection and recognition in natural scenes, effectively improve the half-word recognition effect, and is superior to other recognition algorithms.