Research On Robot's Omni-directional Environment Perception Technology

In recent years,with the rapid development of science and technology,the performance and function of traditional robots can not meet the needs of people's production and life.The development of robots gradually tends to be intelligent and networked.Among them,as the core technology of mobile robot positioning and navigation,environment awareness technology has always been the focus and hot research issue in the field of robot.Nowadays,most of the vision systems used for mobile robot navigation are depth cameras.Although the camera can obtain the depth information of the environment,it has small viewing angle range and high production cost.In the light of this problem,this paper chooses catadioptric panoramic camera as the sensor of the robot to complete the robot environment perception.In this paper,the robot environment perception technology based on catadioptric panoramic camera is studied1)Based on the improved algorithm of concentric circle approximation expansion,the ring panoramic expansion technology is studied.Because the distortion of circular panorama obtained by the folded reflection panorama camera is large and it is not suitable for people to watch,the improved concentric circle approximate expansion algorithm is used to expand.The traditional algorithm of concentric circle approximate expansion is slow to operate and the image distortion is large after expansion.In order to solve this problem,the algorithm of approximate expansion of concentric circles is improved.According to the principle of the refraction panoramic imaging,the pixel values of each circle of the circular panorama are calculated and arranged in a certain order by the mutual transformation of polar coordinate system and rectangular coordinate system.Finally,the circular panorama is expanded by interpolation.The experimental results show that the speed of the improved algorithm is improved obviously,the distortion has been improved and the imaging resolution is improved.2)Robot localization based on catadioptric panoramic camera.Because catadioptric panoramic camera is not a depth camera,it can't obtain the depth information of the environment,so it can't complete the robot's own positioning and environment perception.To solve this problem,this paper proposes a method suitable for catadioptric panoramic camera,that is,for the same object,when the position of the robot changes,the position of the object on the ring panorama and the pixel value change accordingly.This paper uses the law to calculate the position coordinates of the robot itself,and through the experimental verification,the method is effective and reliable.3)Geometric ranging based on catadioptric panoramic camera.In this paper,the relative position relationship between the target and the robot is obtained by using the target feature point coordinates and the robot coordinates,and then the target position in the world coordinate system is obtained.Finally,the actual distance between the target and the robot is calculated by using the mathematical geometric relationship of the target in the coordinate system.Through the experimental verification,the robot based on catadioptric panoramic camera can complete the environmental perception,and has reliability.