Research On Facial Recognition System Of UAV Intelligent Housekeeper

In recent years,with Alpha Go's victory over world Go champions and successful road shows of driverless cars,and the application of text,fingerprint and voice recognition technologies,artificial intelligence has been highly concerned by people,forming a trend of"artificial intelligence +"and rapidly penetrating into various fields.But the multi-functional humanoid robot housekeeper,which can be applied to multi-scene,needs a long research and development process to put into practice.In this paper,we try to replace humanoid robots with unmanned aerial vehicles(UAVs),which are popular in recent years,and focus on information collection,communication,processing and storage,to build an open and light intelligent housekeeping system,so as to avoid the complicated and long development of mechanical systems.The main research and development work completed in this paper is as follows.In this paper,a design scheme of light intelligent housekeeper system based on small UAV is proposed,and the development of ground control subsystem,voice technology subsystem,face recognition subsystem and Internet of Things subsystem of intelligent housekeeper system is realized in cooperation with other members of the research group.It makes the intelligent housekeeping system initially have "brain”,"eyes","ears","nose","sensory"and"wings".It can listen,speak,see and fly.It also expands these basic functions and achieves more specific top-level application functions.This paper is mainly responsible for the research of"brain"and "eyes"inintelligent housekeeping system,and independently completes the development of face recognition subsystem.At the same time,it undertakes the development of computer vision of UAV ground control subsystem.In this paper,we extract Haar-Like features,use Adaboost training method to obtain strong classifiers,and cascade strong classifiers to achieve a set of face recognition system.In order to improve the recognition rate,the paper first preprocesses the photos with gray scale,adaptive Wiener filtering and histogram equalization,and the face detection rate reaches over 93%.In order to carry out face detection in order to recognize the identity,the paper establishes a face image database for the laboratory members,and tests the performance of EngineFace,Fisherface,LBP and FaceNet based on convolutional neural network on the face database and AT&T face database.Finally,the FaceNet algorithm is selected to realize the system's face recognition function and recognition accuracy.More than 90%.In order to recognize facial expressions,the paper collects and establishes a database of facial expressions,further extracts Gabor features of each facial expression photo,and reduces the dimension through PCA.Finally,a one-to-many SVM classifier is constructed to classify facial expressions,which realizes the facial expression recognition function of intelligent housekeeping system.The accuracy rate of facial expression recognition is more than 80%.Based on VS2017 platform and OPENCV,this paper develops a face recognition software system for UAV intelligent housekeeper,which realizes the basic functions of face recognition,face detection and expression recognition,and further realizes the top-level application functions of attendance management,book borrowing management,message management and entertainment.Experiments show that the software system developed in this paper works stably and basically meets the requirements of intelligent housekeeping system based on UAV.