A Machine Vision-Based Character Recognition System for Suspension Insulator Iron Caps.

Character recognition on the surfaces of iron caps is pivotal in the automated production of suspension insulators. This article proposes a machine vision-based character recognition system for the iron caps of suspension insulators in order to solve the problem of recognizing curved characters with 3-D structures and severe surface reflections. In this system, first, the 3-D character extraction system (3D-CES) converts 3-D characters on curved surfaces to flat 2-D characters. Then, the 2-D characters are segmented into individual characters using an improved drop-fall (IDF) segmentation algorithm. Finally, the individual characters are recognized using a pruned Visual Geometry Group 16 (VGG-16) network. This article tested the proposed system on real iron cap data, and a recognition accuracy of 98.3% was achieved. The performance of the proposed system is higher than existing deep-learning-based text character recognition methods by 2.9%.