Research And Implementation Of The Broken-down Video's Image Identification And Diagnosis Based On Cnn

In digital image processing,the image quality will be affected by the hardware device,image processing method,the environment change,noise and other factors which will distort and degrade the image.This will not only cause the loss of information,but also affect the subsequent image processing.It requires the detection and recognition of video image quality.The traditional method of image quality detection is used for image feature extraction and analysis based on artificial experience.In view of the fact that there are many reasons for the quality of video image abnormally,and these anomalies show diversity and complexity,The reliability and stability is poor to use a single or a few characteristics to detect the video quality in the process of the practical application.Convolution neural network model is a multi-level system structure,which with high fault tolerance,parallel computation ability and high computing performance,has been widely used in the fields of pattern recognition and intelligent image analysis.It use the local receptive field and weights of sharing mechanism to reduce the complexity of network model and the number of weights,which can realize the extraction of image features and learning.The detection and classification of convolution neural networks applied to the quality of the video image,has a good application prospect.In this paper we constructs the fault image detection and recognition model based on convolution neural networks,realize the definition and the classification of anomaly image.We mainly do the following work,have a deep the research of basic theory of convolution neural networks and related applications,especially the BP algorithm,the activation function and pooling,through the analysis of the existing several significant application model of character recognition system like ImageNet model and face detection model,we summarized the network effects in the process of setting up the data set,the network layers and other factors on the model.We design and achieve the convolution neural network structure of fault pattern recognition,produce the feature model,and by expanding the data set,we adjust the network layers and the parameters to optimize the model measures.Our experimental study show that,the convolution neural network model proposed in this paper can be distinguished on the abnormal image and normal image which achieved good results.