Research And Implementation Of Pedestrian Detection Method For Service Robots

As humans gradually enter the information society,artificial intelligence-related technologies have made considerable progress,and intelligent products have entered every aspect of daily life.People's high demand for quality of life has made the demand for service robots more and more urgent.Pedestrian detection is a basic requirement of service robots.The service acceptor of service robot is human,and interaction with people is an indispensable requirement.In the process of human-computer interaction,service robots need to understand the environment,detect pedestrians in the environment,and locate the specific location of pedestrians.Specifically,the detection of pedestrians requires the service robot to classify objects detected in the surroundings and discriminate between pedestrians and non-pedestrians.Pedestrian detection is a basic function which can provide great convenience for upper-level planning.Therefore,accurate and real-time detection of pedestrians is the basis for robots to perform human-computer interaction.Only by better solving this problem,service robots can complete various tasks and better serve people.The purpose of this thesis is to design and implement a fast and accurate pedestrian detection system for service robots to provide support for human-machine interaction of service robots.The research content is divided into two parts:pedestrian detection algorithm and robot pedestrian detection system.In the pedestrian detection algorithm research section,a deep learning pedestrian detection framework is adopted,and by improving the convolutional neural network,the computer can detect pedestrians in the video stream information and perform annotation instructions.In the implementation of pedestrian detection system,this thesis builds a pedestrian detection system based on NAO robot and local server.In this system,the robot transmits the collected video information to the local server.The local server performs pedestrian detection through the pre-trained models.In order to ensure the stability of the network communication and prevent data loss,the communication method between the robot and the server is based on TCP/IP.This thesis starts with the theoretical significance and practical value of service robot and pedestrian detection,addresses the purpose of the research,and introduces the research states and trends of pedestrian detection technology at home and abroad.Secondly,in order to improve the stability and efficiency of pedestrian detection systems,by comparing traditional pedestrian detection methods with deep learning methods,based on the existing research results,an end-to-end deep learning pedestrian detection method based on regression is designed and modified.Then,the NAO robot experimental platform is introduced,the working mode of the NAO robot is analyzed,and the computer is programmed to implement real-time video stream image analysis and target detection.Finally,a NAO robot pedestrian detection system is built to verify the pedestrian detection algorithm based on the deep learning framework.Experimental results show that the pedestrian detection system can effectively detect pedestrians in real-time video streams and display annotations,which can provide support for service robots to perform more advanced tasks.