Digital instrument character recognition method

Abstract

The invention discloses a digital instrument character recognition method. The method comprises the following steps of: (1) preprocessing an acquired digital instrument image; (2) carrying out regionpositioning on the image preprocessed in the step (1) by adopting an improved pixel point statistics method to obtain a character region in the digital instrument image, and completing region positioning; (3) performing character segmentation on the character region in the step (2) through an average segmentation method to segment a single character in the image; then entering the step (4); and (4) identifying a single character according to different numbers of connected domains of the characters and different centroid coordinates, and determining a numerical value of each character. The method has the advantages that an improved pixel point statistics method is adopted to carry out region selection, and character regions under the condition of all '0', all '1' or '0' and '1' combinationare accurately positioned; character recognition is performed by adopting a method based on connected domain quantity statistics and centroid coordinate comparison, and the method is simple in principle, low in calculation overhead, rapid in recognition and easy to implement.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a character recognition technology based on digital meters, which can be applied to automatic acquisition of digital meter data and meter monitoring systems. Background technique [0002] At present, digital meters are used in all aspects of social life, and the research on character recognition methods for digital meters is becoming more and more extensive. The existing digital character recognition methods mainly include thread recognition method, support vector machine method, neural network analysis method, but these methods have some shortcomings. The threading recognition method to recognize the numbers in the digital display instrument has high requirements on the accuracy of character positioning and cutting in the previous step, and it is difficult to realize; the method of support vector machine needs to solve the multi-classification problem, which is relativ...

AI Technical Summary

Problems solved by technology

The existing digital character recognition methods mainly include thread recognition method, support vector machine method, neural network analysis method, but these methods have some shortcomings.

The threading recognition method to recognize the numbers in the digital display instrument has high requirements on the accuracy of character positioning and cutting in the previous step, and it is difficult to realize; the method of support vector machine needs to solve the multi-classification problem, which is relatively complicated; the neural network analysis method The number of neurons in the hidden layer requires some prior knowledge

Images