Research On Face Detection And Recognition Method Of Intelligent Security Inspection Robot

With the advancement and development of science and technology in the world today,artificial intelligence technology has continuously penetrated into the fields of agriculture,service industry,aerospace,medical and health,national defense and military,and has brought about a new-round of scientific and technological revolution.With the support of artificial intelligence technology,the intelligent security robot industry is also developing rapidly.my country's current security industry has a huge gap in personnel,and with the current aging of the population in my country,the traditional physical-defense&human-defense model can no longer meet the growing security needs of our society,so smart security robots are used to replace traditional security.The model has become the current development trend in the security field in my country,and face detection and recognition technology is an important technology to ensure that smart security robots successfully perform securi ty inspection tasks.Research on the application of face detection and recognition technology on smart security robots Efficient application has high engineering significance.Therefore,based on the intelligent security inspection robot,this paper has la unched the research on the related methods of face detection and recognition.The main research work is as follows:(1)The paper introduces the background and significance of the research of this subject,proposes current problems and solutions,and also introduces the current research status of intelligent inspection security robots at home and abroad and the current research status of face detection and recognition technology at home and abroad.Then introduced the research content and organization chart of the thesis.(2)The paper studies a fast face detection method based on the improvement of the MTCNN(Multi-task Cascaded Convolutional Networks)framework.Aiming at the problem of insufficient real-time face detection of the on-board computer of the intelligent security inspection robot,the MTCNN target detection framework is improved: first,by adjusting the input image pyramid According to the requirements of subsequent face recognition tasks,adjust the network structure to ensure that the smallest face image size can be recognized,avoid detecting too small face areas that cannot complete the recognition task,and waste computing resources;Secondly,by improving the output of the network,the task of detecting key points of the face is removed,and only the face classification result and the four coordinate positions of the face area frame are retained,thereby increasing the speed of the face detection network and realizing the face detection network on the machine.Good real-time performance on the loading computer.The improved network is trained on the public face datasets of Celeb A(Celeb Faces Attribute)and Wider Face(Web Image Dataset for Event Recognition).The experimental results show that the improved MTCNN face detection algorithm and other well-known face detection algorithms have improved real-time performance and accuracy.Comparison The original MTCNN algorithm has better performance in terms of accuracy and real-time.(3)The paper studies the deep learning face recognition method base d on the joint attention mechanism.It first studies and compares several loss functions,and finally designs the joint loss function;then analyzes the benefits of introducing the attention mechanism in computer vision,and designs The joint attention mec hanism network of the channel domain attention network and the spatial domain attention network is combined,and then the joint attention mechanism network and the feature extraction network are combined to design a face recognition model.Finally,trainin g on the LFW(Labeled Faces in the Wild)and YTF(You Tube Faces)public data sets,the experimental results show that the introduction of the joint attention mechanism can make the model pay more attention to important features in the face recognition proc ess,ignore the unimportant features,and can better improve the Performance of face recognition model.(4)The paper introduces the intelligent security inspection robot vision experiment platform built in the laboratory,including the hardware compositio n,software architecture and wireless bridge communication mechanism of the intelligent security inspection robot,and introduces the structure of the intelligent security inspection robot in detail.The principle of motion control,and a detailed introduc tion to the algorithm of synchronous positioning and map construction of the intelligent security inspection robot and the principle of path planning are also given.Finally,the paper carried out the autonomous patrol experiment of the intelligent securit y inspection robot based on the real scene of the laboratory.The experimental results show that the intelligent security inspection robot can complete the inspection task well in the real scene,detect the passing personnel in time and doing accurate identification.