A Temperature Measurement Method Based on Visible Light Chromaticity Index and k-nearest Neighbors Algorithm

Based on radiation temperature measurement technology, high temperature objects, such as more than 1000 degrees, can be measured both in the visible and infrared band. Low temperature objects, such as those below 100 degrees, can only be measured in the infrared band because of their low emissivity in the visible band. This paper presents a low-temperature measurement method in visible band by using machine learning method, k-Nearest Neighbors Algorithm, to extracted Red, Green and Blue (RGB) color information characteristics of visible images taken by digital camera. The error between the predicted temperature by this method and the real temperature is about 0.5°. This method has simple operation and low prices. The results obtained have demonstrated that the RGB color information characteristics method can provide sufficient temperature information in low-temperature measurement and the method performs better than models using features of brightness of the black-and-white image.