Research On Infrared Image Segmentation Method Of Aircraft Skin Damage

Aircraft skin is one of the most important components of an aircraft.The presence or absence of skin damage directly determines whether an aircraft can fly safely.Infrared thermal imaging non-destructive detection technology can not only detect the damage on the surface of the aircraft skin,but also detect the damage in the interior of the composite material skin without contacting or damaging the aircraft skin.In order to further identify and judge the skin surface damage and internal defects,and provide a basis for subsequent fault diagnosis,it is of great significance to segment the damage image of aircraft skin.Firstly,the infrared thermal imaging detection system is built to collect the infrared images of the damaged specimen of aircraft skin.For the low contrast of the infrared image,the adaptive histogram equalization with limited contrast is adopted to enhance the infrared images;for the high noise and low noise ratio of the infrared image,the wavelet denoising is adopted to denoise the imagesSecondly,I-Ching divination evolutionary algorithm(IDEA)is studied and two schemes are proposed to improve the I-Ching selection operator : improved roulette selection and tournament selection.According to the experiments,it is determined to apply the improved roulette selection to improve the IDEA to optimize Otsu,and apply the tournament selection to improve the IDEA to optimize the maximum entropy.Furthermore,an infrared image segmentation method based on Otsu and roulette selection improved IDEA is proposed,and an infrared image segmentation method based on maximum entropy and tournament selection improved IDEA is proposed.The above two methods are used to process the infrared images at the same time and the infrared images at different times,then the infrared image with better segmentation effect is selected to study the influence of population size on convergence speed,so as to determine the population size suitable for the image.Compared with other optimization methods,the experimental results show that these two methods are superior to other algorithms in terms of convergence speed and time consumption,which indicates that these two methods can effectively improve the speed ofthreshold solution and better segment the damaged images.Then,because of the emperor penguin optimization algorithm has the problem of slow convergence and is prone to fall into a local optimum,an improved emperor penguin optimization algorithm(IEPO)is proposed.In IEPO,the emperor penguin population is firstly initialized by cubic map chaotic initialization,which is helpful to avoid the algorithm from falling into the local optimum.Then Gaussian mutation was introduced into the position updating formula of emperor penguin to increase the diversity of emperor penguins.Furthermore,Levy flight is performed to improve the randomness of the emperor penguin population and prevent the emperor penguin individual from falling into the local optimum.Due to the problems of large amount of calculation and slow running speed of the two-dimensional Tsallis entropy based on the gray level-local entropy(GLLE)histogram,the improved emperor penguin optimization algorithm is used to optimize it to increase the speed of threshold selection.Experiments show that compared with the other three methods,the proposed method has better segmentation effect on infrared images and shorter running time.