Design Of Digital Image De-Rain System Based On Vision Sensor And Deep Neural Network

De-rain of digital images is an important issue of common concern in the field of computer vision and digital image processing.In rainy days,this image is widely used in security video surveillance,intelligent transportation,and even military applications because images after raining can give a clearer image of surveillance cameras.The traditional method of rain removal is mainly to hard code the raindrop itself through various methods.The human experience is to find the characteristics of the raindrop in the whole image to achieve the purpose of rain.Such methods have the disadvantages of large subjective factors,complex algorithm design,poor real-time performance of the algorithm,and inefficient use of existing computing resources.Deep neural network is a new technology that has been developing very rapidly in recent years.It frees researchers from traditional algorithm design and feature extraction,allowing them to focus on researching better,more efficient models and how to Effective use of computing resources.This paper first introduces the development of deep neural networks and briefly explains the working principles and application prospects of neural networks.Then the whole structure of the digital image dewatering system is discussed,and the design ideas of each module of the system are elaborated.The algorithm design of the digital image de-raining system is introduced in detail.It is divided into three parts to explain: Part I.This article establishes a data set for the de-raining task.This data set is crawled through the network,data cleaning and normalization,artificial the three steps of adding rain achieved a better simulation of the actual situation.In the experiment,the model trained using this data set also achieved good results when the actual image was tested.In the second part,the design framework of this paper was introduced in detail.The idea,based on the current stateof-the-art technology of image dewatering,proposes an improved method and designs a new architecture to remove raindrops and rainlines based on the background image as much as possible.This article mainly discusses the matters needing attention in the training of the network designed by this article,as well as the details of the parameter settings.The algorithm design of the digital image de-raining system is introduced in detail.It is divided into three parts to explain: Part I.This article establishes a data set for the de-raining task.This data set is crawled through the network,data cleaning and normalization,artificial The three steps of adding rain achieved a better simulation of the actual situation.In the experiment,the model trained using this data set also achieved good results when the actual image was tested.In the second part,the design framework of this paper was introduced in detail.The idea,based on the current stateof-the-art technology of image dewatering,proposes an improved method and designs a new architecture to remove raindrops and rainlines based on the background image as much as possible.This article mainly discusses the matters needing attention in the training of the network designed by this article,as well as the details of the parameter settings.Finally,we summarizes the main work of this paper in the design process,the current problems of the system and the future improvements to the system.