Measurement Of Temperatures Using Color-Single-Channel Based On Neural Network

Temperature is a common physical quantity in the scientific research, industrial production and daily life. Temperature measurement becomes one of the significant subjects of scientific research. It has important practical significance to research on improving the method of temperature measurement.Firstly, this thesis introduces the actualities and research method of temperature measurement using color and the theory of this method. And then, this thesis proves that it exists a mapping relationship between the color and the temperature of the high-temperature object, and it exists a quantitative relationship between each R,G,B color-channel and the temperature. Against the problem such as the complex nonlinear relationship between the image color and the temperature of high object. According to the advantage of the neural network, this thesis discusses the superiority of temperature measurement using color based on neural network.Secondly, using MATLAB to process the image acquisition samples into digital data, and obtains the pigment-value of R,G,B, furthermore, it realizes the program of BP neural network, and choose Levenberg-Marquardt (LM) neural network as the experimental carrier network by further emulation.Finally, through modeling,training and simulating of LM neural network, this thesis obtains simulation results of each R,G,B color-single-channel.The experimental results show that all of the R,G,B color-single-channels have good measuring accuracy. Further compared the differences of the measurement results among each color-single-channel, and find that the precision of green single channel is the highest and the measuring range is the widest. This thesis realized the improvement of the temperature measurement using color and the optimization of the model.