Research On Defect Detection Of Wind Turbine Blade Based On Digital Image Processing

Presently,wind power generation is the power generation technology with a very wide range of applications.As the renewable clean energy,wind energy is extensively used all over the world.The wind turbine blade is a crucial part of wind turbine.If there is damage to the wind turbine blade,its consequence will be extremely serious,either economic losses or safety hazards.The conventional wind turbine blade defect detection mainly includes manual testing and acoustic nondestructive testing,etc.It has various drawbacks,such as unsafe,time-consuming and low accuracy.However,the surface defect detection method of wind turbine blade based on image processing is a non-destructive testing technology.It has advantages of safety,simplicity and high accuracy.Under this background,this paper presents the main research contents based on the research on defect detection of wind turbine blade based on digital image processing.According to that Gabor filter and the Log-Gabor filter need to extract features through multiple filter templates,and the number of output images is large.These images each cover different points of information with its own strengths.A Log-Gabor feature extraction method based on image fusion is proposed.An adaptive Log-Gabor filter is proposed to solve the problem that the parameters of Log-Gabor filter need to be selected and the optimal solution is affected by human selection and has limitations.Log-Gabor feature extraction based on image fusion.The first is to combine the Laplacian pyramid decomposition with the regional energy fusion algorithm,a new image fusion method is obtained.Next,the two filters are improved with this new image fusion method,and the output multiple images are fused.The advantages of the improved algorithm based on image fusion are compared in the tests.An improved adaptive Log-Gabor filter,PSO algorithm is used to improve it.However,the PSO algorithm tends to have problem in local optimal solution.The PSO algorithm can be improved with Levy flight strategy,and the improved algorithm is called LPSO algorithm.It can allow PSO algorithm to be free from local optimal solution successfully.Then combine the LPSO algorithm with the LogGabor filter to generate an adaptive Log-Gabor filter.Compare the two adaptive filters through simulation experiments.Finally,the image is processed by morphology.Based on the SVM binary classifier,a HOG+SVM multiple classifier is designed which can judge four defect types.The classifier training was completed,and the feature images generated by different improved algorithms were used to compare the detection accuracy and illustrate the advantages and disadvantages of these different methods.Design GUI user interaction interface,realize the visualization of operation steps,simple and intuitive observation of the results.