Research And Implementation Of Anti-Terrorism Features Recognition Based On Image Processing And Convolutional Neural Network

At present,China’s anti-terrorism information platform is still in it’s preliminary stage,The provinces and cities have conducted anti-terrorism information platform construction,but the functional modules of the platform are still in a blank state.Therefore,completing the functional modules as soon as possible is the most important task in today’s anti-terrorism work.Today,many years of anti-terrorism work have accumulated a large amount of data.Finding the value from a large amount of data is the main task of this paper.Combining computer technology with surveillance videos to complete the functional design of the key position control module,it can get detection of sensitive features early and allow staff to allocate the police force reasonably and financial resources,make preventive measures in advance,adjust the control of different key positions flexibly,reduce the workload of the staff and improve work efficiency.This paper mainly studies the application of convolutional neural networks in the identification of terror sensitive features.The terror sensitive features of the study are mainly sensitive clothing,The following types of sensitive features have been organized: Xingyue logo,Crape,Beard,Ricek’s turban,and Burka clothing.Today’s network monitoring system is becoming more and more mature,and monitoring has spread all over the streets,providing a lot of video information for the public security department.Therefore,this paper will use convolutional neural network to conduct sensitivity features training,so that the trained identification model replaces the manpower to search sensitive features.Convolutional neural network mainly abstracts and reduces the dimension of input samples to achieve the purpose of feature extraction.The theory is to extract the features in the terror sensitive feature pictures through repeated learning of input data,continuously adjust the sensitivity of sensitive features,and show the classification results of sensitive features in the hidden layer.At the same time,a certain neuron in the convolutional neural network is abnormal,and other neurons can still preserve the corresponding feature weights.Therefore,the network has a feature like strong anti-interference.In view of the fact that there are vague conditions in the video image that are influenced by various external factors,this paper will provide in-depth understanding of the sample data provided by the anti-terrorism department and summarize the existing problems,and select appropriate image processing technology to carry out purposeful processing of sample data to achieve a good picture pre-processing effect,in order to identify the sensitive feature effectively.The experimental part of this paper will be implemented through two software simulations,image denoising and image enhancement will be implemented throughMatlab software for experimental analysis,deep learning experiment process will be carried out in tensorflow environment,using the Syder tool for coding experiments.At the end of the experiment,two sets of data tests are performed,one is the identification of the correct and the error set sample,the other is the identification of the sample after the image pre-processing.The experimental results prove that pre-processing technique of terror sensitive features selected in this paper is effective,and the classification result of the sensitive feature recognition model is good.