Autonomous Robot Navigation System Based On Embedded Vision

With the mobile robot gradually step into people's life,the robot navigation as a key technology of mobile robot has attracted more and more attention from researchers.Due to the complexity of the mobile robot working environment,it needs a more efficient sensor to perceive the surrounding environment.Therefore,it is the first choice for the robot navigation sensor to collect the environmental information in real time.At present,due to more comprehensive perception of the surrounding environment information,the visual robot navigation technology has become a hot topic in the field of robot navigation.In this paper,a set of robot vision navigation system is designed based on embedded processor and RGBD depth camera as the sensor to collect the surrounding environment information.The specific work of this paper is as follows:First of all,in view of the limited computing power of the embedded platform,this paper uses the fast ORB feature as the image feature point extraction algorithm to estimate the robot's position and posture under the premise of rapid extraction of feature points,and the random sampling consistency algorithm is used to eliminate the mismatch problem in ORB feature matching.Secondly,according to the accumulated error caused during the movement of the robot,the robot loopback detection algorithm based on the visual dictionary is introduced to calculate the similarity between the current frame image and the map key frame,and to detect whether the robot has returned to a certain position in the past.Meanwhile,bundle adjustment is applied to eliminate the cumulative error of motion process.Finally,aiming at the problem that the 3D point cloud map is difficult to be directly applied to the path planning of mobile robot,this paper proposes an algorithm based on the projection of the surface.This method can reduce the complexity of the point cloud map by extracting the ground plane in the 3D point cloud map and projecting the complex 3D point cloud map onto the robot motion plane,and then build the raster map based on the projected ground plane cloud map.Finally,the A star algorithm is used to realize the path navigation of the robot.The experimental results show that this system improves the density of sparse point cloud map and increases the description of the motion environment of the mobile robot,and accurately extract the motion plane of the robot to solve the problem that the point cloud map can not be directly applied to the path planning.