Design And Implementation Of Key Character Recognition System In Mobile Video

With the rapid development of the mobile Internet,more and more short video platforms have emerged,such as Douyin short videos and Kuaishou short videos.These platforms have become an indispensable entertainment activity in people’s leisure life.Most of the videos posted on the platforms are people’s daily life and other content,but some people will use video publishing platforms to post some illegal content,or even appear.Criminals,this poses a big challenge to public safety.At the same time,the key personnel concerned by the public security may also appear in a certain video.The system provides an auxiliary means for the public security to find key people,and at the same time can provide protection for public safety.This article mainly uses the convolutional neural network model in the application of face recognition.Compared with the traditional face recognition method,the convolutional neural network does not need to be performed manually in the complex algorithm design.The main content of this article includes the following parts:(1)This article mainly includes two parts: video capture and download and system management module.Video capture and download are mainly to capture the video URL published in the short video platform,and then download the captured video.At present,the mainstream short video platform is only APP.For video capture,data capture software fiddler is needed.Due to the huge number of Douyin videos,it is necessary to use appium automation tools to replace manual sliding videos,which greatly saves labor costs.For the system management module,the interface of the system is designed using the B/S architecture,and the middleware is implemented using serverlet.Analyze and model system functions through UML modeling,design each function in the system,and then design the function points through the design flowchart.(2)Video capture and downloading is first to capture the video URL by using the working principle of network packet capture,and then call the captured URL through the Python service to download the video.For each function in the management module,class diagrams and sequence diagrams are used to realize the function points.The functions realized by the system include base library management,early warning management,alarm management and system management.This system compares and analyzes the captured video with the person information in the base database after face recognition.If the comparison and analysis result matches,the alarm management function is used to give analarm.Among them,there are people in face recognition.Face detection,face feature point extraction,and comparison analysis are three steps.These three steps are implemented in the system through the faster R-CNN algorithm,KD-tree algorithm,and cosine similarity algorithm.(3)After the overall design and realization of the system,the function and non-function of the system were tested.The system uses white box testing and black box testing.The code is tested through code walkthrough and testbed tool,and the various function points of the system are tested and analyzed through black box testing.According to the test results,it can be concluded that the test results meet the expected results,the system is reliable and stable,and meets the design requirements.The system has good market application value and provides great convenience for the public security system to identify key figures in massive videos.