Research On Automatical Temperature Interpretation Algorithms Based On Temperature Sensitive Paint Image

Temperature sensitive paint(TSP)is a special material that can change its color with the change of temperature and the color-temperature matching relation is definite.Therefore,the temperature can be identified by judging the paint colors,which forms a contact-type temperature measurement method.This method is especially suitable for measuring large area temperature field and has been widely applied for surface temperature measurement of aero engine components.At present,TSP temperature interpretation is mainly based on manual vision.Since the test results are influenced by operator’s personal experience and proficiency,the test repeatability and accuracy can hardly be guaranteed and the efficiency needs to be improved as well.Therefore,this dissertation explores an automatic TSP interpretation,where algorithms based on image processing and machine vision were developed to enhance TSP images and interpret temperature distribution.The research in this dissertation includes:(1)Research on preprocessing algorithms for TSP imagesThe TSP color data is crucial for the temperature interpretation and its accuracy directly affects the final interpretation results.In this dissertation,different TSP color shift phenomena caused by the light condition,paint color degradation and surface sand-blasting are investigated.The feature difference between color shift images and original images is analyzed,and color restoration algorithms for each issue are proposed.(2)Research on clustering segmentation algorithm for TSP imagesIsotherm is an important indication of TSP temperature interpretation.According to the characteristics of TSP images,the feature information of isotherms is fuzzy,which reduces the accuracy and efficiency of isotherm extraction.Related work shows that the isotherm quality can be improved by accurate segmentation of TSP image effectively.Therefore,the application of clustering algorithm for TSP image segmentation is investigated.Then the corresponding improvements for TSP images are studied.(3)Research on TSP isotherm extraction algorithmsIn order to solve the problem of isotherm lost,break,and rugged,the reasons for the above problems are analyzed firstly,and then the corresponding solutions are studied.Specifically,this research involves the issues of contrast lost caused by image dimension-reduction and poor isotherm quality caused by the limitations of edge feature.Finally,based on the study of edge extraction,an isotherm detection algorithm is presented.(4)Research on isotherm recognition algorithm for TSP imagesThe traditional isotherm recognition algorithms are based on pixel values(color)feature.However,it’s usually influenced by some objective factors such as illumination conditions,which causes temperature interpretation error.To deal with above problem,a stable feature and the corresponding isotherm recognition algorithm are presented.(5)Development of automatic TSP interpretation systemBased on C++ and QT platform,a software platform of automatic TSP interpretation system is developed,which integrates the image preprocessing algorithm,isotherm detection algorithm,segmentation algorithm and temperature interpretation algorithm.The framework and functions are designed according to the practical engineering requirements.It has now acquired software copyright and has been applied successfully.The contributions of this dissertation are as follows:(1)A series of preprocessing algorithms are proposed to deal with the problem of color shift under different conditions.The hot gas flow in engine makes the paint color degraded,which causes errors in the color-temperature matching stage.According to the features of TSP images,current contrast enhancement algorithms cannot protect the hue feature effectively.In this dissertation,the problem of color degradation is transformed into a haze removal problem,and a restoration algorithm based on atmospheric scattering model is proposed.This algorithm can restore the lost contrast and saturation information effectively.Furthermore,the hue can be well protected.In addition,based on Retinex theory,a TSP color correction algorithm is proposed to fix the color shift caused by the light source change,which improves the robustness and simplifies the test process.(2)An isotherm extraction method is proposed by combining cluster segmentation and edge detection.This dissertation presents two segmentation algorithms for TSP images based on FCM clustering algorithm and K-means clustering algorithm,which enhances the feature of isotherms greatly.Then an isotherm extraction algorithm is proposed by combining the contrast-preserving decolorization,the edge feature of Laplacian operator and Canny operator,which reduces isotherm lost and improves isotherm quality.(3)A temperature interpretation algorithm for TSP isotherms based on the maximum mean square error channel sequence is proposed.The traditional automatic interpretation algorithms are based on the pixel value of TSP images.In this dissertation,it is found that the feature of maximum mean square channel sequence exhibits the excellent stability.Therefore,we take the sequence as the identification feature of the isotherm and propose a TSP temperature interpretation method with high robustness.