Research On On-line Detection Technology Of Optical Component Damage Based On Degradation Model Restoration And Convolutional Neural Network

With the development of computer technology,computer vision began to shine in the field of vision.In the field of large optical devices,the traditional off-line detection technology needs to analyze the components in optical device after disassembly.Therefore,the off-line detection technology cannot guarantee the efficiency and will affect the cost and service life of the device.Therefore,the research of on-line damage detection of optical components has become an urgent problem.In the research of on-line detection of optical components,many scholars at home and abroad have also proposed a variety of research methods based on traditional image processing methods in recent years.With the development of deep learning in recent years,convolutional neural networks have become a popular tool in the field of computer vision.In this context,an online damage detection method for optical elements based on degenerate model recovery and convolutional neural network is proposed.In this paper,convolutional neural network is combined with image restoration technology to detect damaged images.Firstly,the image is modeled by the degradation model restoration technique in this paper.Then,according to the prior knowledge in the image,the restoration algorithm is designed.It is applied to the restoration of the degenerate model of the damaged image.The experimental results are observed and compared with traditional image processing methods.Secondly,due to the particularity of large optical systems,the cost of acquiring optical damage images is extremely high.In order to facilitate the subsequent training of classifiers,a data set making scheme is designed in this paper.The data were divided into small areas,and then the damage points in the area were detected accurately.Then a large number of pseudo-samples are expanded based on small image regions,which can be used as training samples to train the classifier.Finally,this paper designed a classifier based on convolutional neural network for classification detection,and then trained and tested it with the prepared data set,and compared it with other machine learning and deep learning algorithms to observe and analyze the results of the classification.Subsequently,the image data set after restoration processing and the image data set without restoration are respectively trained and detected by this network,which proves that the restoration algorithm is effective for damage image recognition and detection.