A new method of digital number recognition for substation inspection robot

Abstract

A new method of digital number recognition for mechanical digital meters in substation is studied in this paper, which adopts linear SVM based on Histogram of Oriented Gradients (HOG) features. The grids of Histograms of Oriented Gradient descriptors significantly outperform for feature detection of the gray image which has more information than binary image. A new approach with segmentation of region of character image is proposed in this paper, which is important to the further HOG feature detection. SVM classifier is used in the recognition procession and result shows that HOG has better performance on digit classification in the substation inspection robot instrument recognition.