Research On Segmentation Algorithm Of Infrared Polarized Image Based On Tsallis Entropy

In the industrial field,metal components are subjected to cyclically or randomly changing loads.Under the load with a long-term stress level,the load-bearing capacity of the test piece will decrease,and it may eventually fail.Therefore,how to prevent the failure of metal components and avoid accidents is very important.Among them,fatigue failure under the action of cyclic load is the main form of failure of metal components,and fatigue damage detection of metal components is an important topic in the field of industrial non-destructive testing.The traditional method of detecting the fatigue state of metals based on infrared thermal imaging has the advantages of real-time,non-destructive,non-contact,etc.,but the use of temperature as the evaluation of fatigue index ignores the influence of the change of metal surface morphology during the fatigue process on spontaneous radiation,and it is difficult to completely explain fatigue deterioration process.Theories and experiments show that the surface morphology is also an important feature change during the metal fatigue process,and the polarization characteristic of spontaneous radiation is an important characterization of the surface morphology.The introduction of infrared polarization detection can integrate temperature and surface morphology from the aspects of thermal radiation and polarization characteristic changes assess the degree of fatigue damage.Image segmentation is an important part in infrared polarization detection.It is the basis of infrared polarization detection to segment the target area in the infrared polarized image and then extract the infrared polarization characteristics of the target.If the target and the background area cannot be accurately segmented,the extracted features will not accurately reflect the characteristics of the research object,making it difficult to accurately detect the measured quantity.In this paper,relevant researches on target segmentation in infrared polarization detection are carried out.For the accurate segmentation of targets in infrared polarized images,there are currently the following difficulties:?1?Infrared polarized images are susceptible to interference from the environment and other factors during the imaging process,causing random noise,resulting in blurred images,unclear edges,and low visibility;?2?In the process of polarization imaging,multiple original images with different polarization azimuths need to be acquired for the same target.Polarization information images such as degree of polarization,polarization angle,and stokes parameters will be further generated during polarization analysis.When using polarization information reasonably,you need to consider redundancy and complementarity of many polarization information;?3?Traditional image segmentation methods mainly focus on single images,and will be subject to certain restrictions when applied to infrared polarization images with multi-channel imaging.After investigation,this paper adopts an image threshold segmentation method based on Tsallis entropy.Tsallis entropy extends Shannon entropy.It is a non-extensible entropy.It is more suitable for image segmentation with low contrast,high noise,and blurred edges.Non-extensiveness is reflected in the system consisting of two independent subsystems A and B,the addendum of the counterfeit satisfied by the system entropy?+?.It can describe physical processes with long correlation,long-term memory and fractal structure.The image can be regarded as an information system composed of the target and background subsystems,so the characteristics of Tsallis entropy can be applied to the threshold selection process of image segmentation.Using the non-extensive nature of Tsallis entropy,further considering the correlation between the probability distribution of the target class and the background class can improve the reliability of segmentation.The non-extensibility of Tsallis entropy is mainly reflected in theparameter of Tsallis entropy.Different values ofreflect the difference of non-extensibility of the system.In this paper,Tsallis entropy is introduced into the infrared polarized image system.For the problem ofvalue selection,combined with infrared polarized image The characteristics of the system have been studied to select thevalue that reasonably describes the infrared polarization system and accurately segment the infrared polarization image.Based on the advantages of Tsallis entropy mentioned above,this paper introduces threshold segmentation based on Tsallis entropy into infrared polarized image segmentation.Combined with the characteristics of infrared polarized imaging,the following work is done respectively:1.For a single-channel infrared polarized azimuth image,a Tsallis entropy threshold segmentation algorithm based on inter-class variance verification is proposed.By calculating the inter-class variance of Tsallis entropy threshold segmentation results under differentparameters,the validity of the parameteris verified,thereby combining the traditional Otsu inter-class variance threshold segmentation criterion with the Tsallis entropy criterion to improve segmentation accuracy.2.On the basis of single-channel infrared polarized azimuth image,combined with the characteristics of infrared polarized multi-channel imaging,taking advantage of the redundancy and complementarity between multi-channel images,the multi-channel image as a whole,each channel is polarized The azimuth images are mutually verified to optimize the selection of Tsallis entropyparameters to improve segmentation accuracy.A threshold segmentation method for multi-channel infrared polarized image based on Tsallis entropy is proposed.3.Image segmentation is an important part of infrared polarization detection.At the end of this article,the segmentation is applied to infrared polarization detection,and the infrared polarization characteristics of the metal fatigue process are extracted,and the research of this paper is reflected through the comparison of multiple segmentation algorithms Meaning.At the same time,related processes of metal fatigue detection based on infrared polarization characteristics are discussed.Figure[37]Table[12]Reference[60]...