Interested Population Attribute Analysis System Based On Computer Vision

Interested population attribute statistics data plays a very important guiding role in advertising industry and sales industry.However,the traditional methods to gain this kind of data are inefficient and inaccurate such as consultation and questionnaire.The life experience shows that people will unconsciously stay and observe when encountering things of interest.So,combining computer vision technology,we can collect and analyze the face image of a person to gain interested population attribute statistics data.This process involves two core steps: face detection and face attribute analysis.Recently,face detection and face attribute analysis are popular research direction in the field of computer vision.Firstly,this paper extensively investigates the development of face detection and face attribute analysis,including traditional feature extraction based methods and recently popular convolutional neural network based methods.And then through complete contrast experiments,we measure the advantages and disadvantages of various methods from three aspect,accuracy,robustness and real – time performance.According to the experiment results,convolutional neural network based methods has Breakthrough performance in accuracy and robustness while compared with traditional feature extraction based methods.Moreover,in the GPU parallel computing capabilities support,convolutional neural network based methods can keep high real-time performance.According to experiment results,this paper choose SSD model as a face detection method and AlexNet model as our face attribute analysis method to construct a complete set of interested population attribute analysis system.In terms of overall architecture,this system takes camera video stream as input,and the data is displayed on the web and the mobile side after a series of analysis.On one hand,we use Django as backend framework to reduce system coupling degree by separating data and views.On the other hand,we use IONIC as frontend framework,which can be used to quickly develop responsive,cross-platform mobile applications and perfectly interact with Django framework.Finally,the functional test results show that the system correctly realize the function of all involved and the system runs stably.