Image registration between visible and infrared images for electrical equipment inspection robots based on quadrilateral features

摘要

Visible and infrared images are essential for electrical inspection robots to detect equipment or components. Here how to combine these two kinds of information is a critical problem. Image registration is one of the necessary processing procedures. While used in practice, traditional intensity-based or feature-based registration methods appear many deficiencies such as low accuracy, low reliability or low speed. In this paper, we propose an efficient registration method based on the quadrilateral features of electrical equipment. Contour and straight line detection technology is utilized to reconstruct the quadrilateral structural information of the monitoring object from original images. Experimental results indicate that our proposed registration method can align visible and infrared images fast with high accuracy and be less likely to be influenced by image quality or other irrelevant objects.