Design And Research Of The Cracks Determination System For Aircraft Inlet Protective Grid

In order to solve the problems of insufficient automation,low efficiency,low reliability and difficulty in meeting the needs of the rapid development of China’s military equipment in existing crack detection methods for aircraft inlet protective grid,this paper designed and studied an aircraft inlet protective grid crack detection system based on machine vision and deep learning.Machine vision image processing and deep learning target detection algorithms are applied to crack detection of aircraft inlet protective grid.Then,the overall scheme of the detection system is designed and analyzed,according to the structural characteristics of the protective grid,an aircraft protective grid crack detection device is proposed,and an image processing method for grid segmentation of the protective grid is designed,and based on the self-made protective grid crack data has carried out a research on the detection of cracks in the aircraft protective grid.At the same time,the entire protective grid crack detection system is further built based on Labview software,the experiments are carried out on the detection device to verify the stability and feasibility of the protective grid crack detection system.The details are as follows:(1)First,the working principle and structure of the aircraft inlet protective grid are introduced.By analyzing the existing methods of crack detection of the protective grid,a crack detection method of the aircraft inlet protective grid based on machine vision and deep learning is proposed.The overall scheme of the detection system is designed and analyzed.According to the structural characteristics of the protective grid,the design of the crack detection device for the protective grid is carried out.According to the characteristics of the grid image of the protective grid,an image processing method for grid segmentation of the protective grid is designed.The research and implementation steps of the deep learning target detection algorithm on the crack of the protective grid are presented.Finally,according to the working principles and performance parameters of the camera,lens and light source,the key hardware selection of the detection device is completed,and the overall layout of the crack detection system is completed.(2)Secondly,in order to collect the grid image of the protective grid with cracks,an image processing method of grid segmentation is proposed: first,the image filtering process is performed on the protective grid image collected by the camera to highlight the detailed characteristics of grid cracks,and binarize the protective grid image,analyze and compare the three convex hull algorithms of Jarvis March,Graham and Andrew,and select the Graham convex hull algorithm to process the binarized image and the contour information of the complete grid in combination with the characteristics of the protective grid image.At the same time,the analysis of various filtering algorithms is introduced,and the particle filtering method is selected to filter the noise in the binarized image to obtain the center coordinates of each grid,and the grid area image is obtained through pixel calculation and image segmentation.The protective grid crack image is obtained by screening,which is used as a sample self-made protective grid crack data set,and provide image input for deep learning crack detection.(3)Then,analyze and evaluate the structural characteristics,detection process and detection performance of the deep learning target detection algorithm.Combined with the requirements of the crack detection system,the Faster-RCNN algorithm is used as the crack detection algorithm of this system,and the method are used.ZF-Net,VGG-16 and Res Net-101 convolutional neural networks are used as Faster-RCNN feature extraction networks to carry out research on crack detection of aircraft inlet protective grilles in the self-made data set.Further comparison experiments are carried out under different lighting factors to verify the feasibility of the algorithm in this chapter.(4)Finally,analyze the software requirements and development methods of the crack detection system,determine the Labview software as the system software development platform,combine machine vision image processing technology and deep learning crack detection algorithm,and complete the construction of the detection system in a modular manner.Experiments were carried out on the detection device to verify the reliability and feasibility of the crack detection system.