Depth Camera And 2D Laser Fusion SLAM Security Inspection Robot

In recent years,artificial intelligence has been widely used in areas such as target perception,pattern recognition,and unmanned driving.Among them,the service robot mainly uses the laser point cloud to perceive the outside world,and combined with the algorithm of Simultaneous Localization and Mapping(SLAM)to realize intelligent operations such as navigation and obstacle avoidance.However,the traditional SLAM algorithm mainly uses a single sensor for navigation and map reconstruction,which is limited by the disorder,sparseness and density unevenness of the point cloud,and has poor stability.On the basis of ensuring low Research and development(R&D)costs,this paper proposes the SLAM algorithm based on visual fusion laser loopback detection.The main research contents and innovations are as follows1.Analysis of security inspection robot system.The differential wheeled robot kinematics model is analyzed,and the plane motion robot is solved at every moment;the two-dimensional lidar triangulation method and the Xtion camera's spatial depth information calculation principle are reviewed;the robot chassis and the two-dimensional calibration are combined with least square Based on the position parameters of the laser,the transformation matrix of the coordinate system of the sensing information and the robot chassis is solved2.Review of traditional laser and vision SLAM algorithms.Introduce the laser front-end related scan matching method,that is,use the odometer to estimate the pose to construct a likelihood field,project the current frame with a certain angular resolution and angular offset value,and finally find the best matching pose by multi-resolution search pose,And implement non-linear optimization at the back end;the visual SLAM algorithm is analyzed,which uses the ORB(ORiented Brief)features extracted by the front end to achieve map construction and pose estimation,and the back end combines the LM(Levenberg-Marquarelt)algorithm to optimize the camera and points postures,and the word bag vector represents the similarity of the two frames of images to determine whether the robot is looping3.The proposition of loop detection method of vision fusion laser.First,the visual bag of word model is cited as the basis for the judgment of laser loopback detection.Then,when the two frames before and after the laser are too small for the global optimization of the loopback,the visual key frame and the laser key frame are synchronously aligned by time stamp,and the visual current frame detects the loopback key frame through appearance verification and geometric verification.The visual loopback key frame is aligned to the laser frame,so that the laser frame is globally optimized.The visual loopback key frame is aligned to the laser frame,so that the laser frame is globally optimized.Through experiments,it has been proved that in a specific small environment,a single laser can not detect robot loopback,and visual laser fusion can detect a 100%loopback;in a specific large environment,a single laser loopback detection rate is 60%,and visual laser fusion is 100%.4.The completion of the inspection task of the robot in the actual environment.In the robot inspection strategy,the first inspection,the second inspection and the three inspections implemented in the normal mode or the specific mode are designed;the inspection objects such as doors,windows,and indoor lights are defined in the inspection,and the security inspection is implemented object security status recognition.In the experiment,the robot loads an environment map based on visual laser fusion,and the system realizes target detection through the ROS communication mechanism.