Research On Surface Damage Detection Of Wind Turbine Blades Based On Pan Tilt Zoom Images

In recent years,wind power generation,as the most favored clean renewable energy,is growing in China.As an important component of wind turbine,the surface damage of wind turbine blade will not only reduce the service life and power generation efficiency of wind turbine,but also increase the monitoring error,safety risk and maintenance cost.Therefore,timely detection of blade surface damage is very important for wind power generation.At present,wind turbine blade surface damage detection methods generally have many problems,such as high cost,low accuracy,unable to quickly and accurately locate the damage location and measure the damage area.In order to solve these problems,this paper studies the automatic identification,location and measurement method of wind turbine blade surface damage based on the electric cradle image acquisition,image processing and deep learning technology.Firstly,aiming at the problem of insufficient datasets of wind turbine blade image,the commonly used data enhancement technology is analyzed,and a multi-scale image mixed data enhancement method is proposed,which is used to extend a lot of data of existing images.Secondly,a method of wind turbine blade surface damage detection based on Mask R-CNN convolutional neural network was proposed.Mask R-CNN convolutional neural network was used to identify,segment and calculate the surface damage area of wind turbine blades.On this basis,aiming at the problem of large measurement errors of some crack damage types,a wind turbine blade surface damage detection method combined edge detector with Mask R-CNN was proposed.Based on the original Mask R-CNN network,the prediction boundary box was used to lock the damaged area,and the edge detector was used to extract the contour in the area and calculate the area in the contour.Experimental results show that this method has good performance in damage identification and can effectively improve the precision of damage region segmentation and measurement accuracy.Finally,in order to accurately locate the position of the damaged area in the whole wind turbine blade,a contour reconstruction curve feature splicing method was proposed to splice adjacent images,and the real position of the damaged area was calculated by image positioning correction method.Experiments show that the method is accurate and reliable.