Calibration Method for Stereovision Measurement of High-Temperature Components Using Two Infrared Cameras

Abstract

The infrared vision measurement method has some advantages over visible vision in the measurement of high-temperature components. Infrared imaging is based on different imaging principles, however, making it unreasonable to adopt a conventional visiblelight camera calibration method directly. The present study proposes a dual infrared-camera calibration program in which we use the infrared imaging principle to measure high-temperature components threedimensionally. We use two ceramic balls as calibration targets against the background of an external hightemperature radiation source. Multiple feature points are generated from the precise movement of these targets. In order to improve the accuracy of the calibration method, we took the following approaches: A highly accurate edge-detection algorithm is realized by using a fuzzy neural network that is self-learning, self-adaptive, and utilizes fuzzy processing. We thus achieve a low signal-to-noise ratio and low contrast in infrared images. The distance between these two ceramic targets is used as a calibration reference to further reduce the temperature effect and to improve calibration accuracy and efficiency. The calibration results we got are an average residual error of 6.4134 µm and a variance of 2.9205 µm.