Design And Implementation Of Rescue Robot Based On SLAM And Machine Vision

Rescue robots can enter dangerous rescue scenes and carry out search and rescue missions on behalf of rescuers.It is a popular field of robot research in recent years.Among them,SLAM and machine vision are the key technologies for rescue robots to participate in rescue missions.This article analyzes the overall requirements of the robot by referring to the Robocup Rescue obstacle crossing and search and rescue project competition rules,and builds a ROS-based rescue robot system and completes the main projects of SLAM and machine vision required by the competition.The thesis first elaborated on the status quo of rescue robots,SLAM and machine vision.Under the Ubuntu operating system,it explained the reasons for choosing ROS as the core system of the robot from several aspects.According to the requirements of the competition and the rescue environment,the robot structure and the peripheral hardware it carries have been considered in many aspects,and the most suitable equipment for the rescue robot is selected to carry out the subject research.Then,according to the map construction requirements of the game,the SLAM technology was studied.Analyze the basic principles of EKF-SLAM,particle filter SLAM and graph optimization SLAM.Explains the mainstream SLAM map types,and uses the currently popular Hector-SLAM algorithm and Gmapping-SLAM algorithm to carry out related simulation experiments,and finally builds a corresponding venue for experiments to verify the feasibility of the algorithm.The last chapter deals with machine vision,briefly introduces machine vision technology and its components.It should be noted that ROS robots usually need to convert image data in ROS into Open CV format images for processing.The camera is calibrated using Zhang Zhengyou's calibration method to calculate the internal and external parameters of the camera to prevent image distortion.Explain the information type of each area of the QR code and use the Zbar component to realize the QR code recognition;mark the location of the QR code in the experimental environment on the map and display the main information contained in the QR code during the SLAM process,different from the traditional target position calculation,According to the rules of the game,this article puts forward a method to mark the target on the map without using the depth camera to obtain the depth information of the target position,which saves the cost to a certain extent.;use the SIFT-based algorithm The feature matching method realizes the identification of dangerous goods signs;after the image is processed by grayscale and Gaussian filtering,the grayscale difference is used to determine the motion state of the object and finally the moving object is recognized.