The Design And Implementation Of Pornographic Picture Detection System Based On Convolutional Neural Network

In recent years,with the rapid development of computer technology and network technology,the number of Internet users in China continues to grow,has reached 800 million people,the Internet users are showing a trend of low-age development,the number of young Internet users is increasing.A large number of pornographic pictures on the Internet have brought great negative effects on teenagers,affecting their healthy growth.At the same time,some lawless elements,using the characteristics of pornographic pictures easy to spread,induced users to download trojans,viruses and other malicious software,destroyed the healthy environment of the Internet,to the network supervision caused great difficulties.With the rapid development of social networks,massive image information is uploaded to the network space.Large-scale image information brings greater challenges to the detection of pornographic pictures.The massive pictures in the cyberspace depend on the way of manual review,which is difficult to complete.With the development of artificial intelligence technology,researchers propose the use of artificial intelligence technology to automatically identify images,replacing the traditional manual review.ln this thesis,combined with the artificial intelligence technology,designs and realizes a pornographic picture detection system,system from the function module can be divided into access control module,image detection module,web pages detection module and the historical record module,the system can be directly within the image detection or according to the site automatically detect web pages on images to return the result.ln pornographic picture recognition system model,adopt the method of the migration study,will be applied to other image recognition task the depth of the convolution neural network model,the migration to the pornographic picture recognition tasks,and using the existing data to fine-tune the model,and then get pornographic picture recognition model training,save training resources and time,improves the image recognition rate.This thesis firstly introduces the project background and research status of the system,clarifies the main work of this thesis,and then describes the technology and related theories used in the system development;Then introduced the system demand design and the overall design plan,elaborated the main module detailed design and the realization.At the same time,the problem analysis,model design,model training and experimental analysis of pornographic picture recognition model are introduced in detail.This system is independently developed by trend technology company in order to meet the needs of the government and enterprises in the field of bad information management and control.At present,this system can effectively carry out automatic examination and detection of network pictures and web pages after actual operation and test.The system runs stably and meets the expected demand.