Research On Method Of Temperature Measurement Using Image Based On Grnn

With the rapid development of computer technology and digital image processing technology, the temperature detection skill based on digital image processing has been widespread concerned at home and abroad, which is an important branch of non-contact temperature measurement research areas. The most important issue of image thermometry is to establish a mathematical model between the image characteristic value and the temperature. There are many ways can be divided into two categories:one is the mechanism modeling built on the Plank radiation law; the other one is the gray box model based on the observation values of input and output. Scholars made a lot of research on different method and issues and put forward a number of algorithms.This article aims to improve the accuracy of temperature measurement when using neural networks. On the basic study of temperature measurement methods using image, proposed a method of temperature measurement using image based on generalized regression neural network (GRNN).The main study contains the following points:Firstly, analyze the feasibility of application GRNN to temperature measurement using image, including the basic mathematical theory and implementation platform. On this basis, given the establishment steps of temperature measurement using image based on GRNN;Secondly, it is need to extract the image feature value of the furnace on the modeling process, so the method of image segmentation is also studied in this article. Focus on the OTSU image thresholding segmentation method, and because it is slow speed when dealing with high-pixel image segmentation using this method, so OTSU image segmentation method based on particle swarm optimization algorithm is proposed in this article, and verify the validity of the method;Finally, design and complete the experiment of image temperature measurement based on GRNN. Include designing the experimental devices and precautions, getting the furnace image, pre-processing the images and extracting the feature value from images, establishing the temperature model and forecasting the furnace temperature. The prediction result is compared with the BP neural network at last.The result indicated that it is high precision and speed to use the method of temperature measurement using image based on GRNN.This method has the important application value and prospect regarding at high temperature measurement filed.