Research On The Technologies Of Precise Guidance And Accurate Positioning In Vision-guided AGV

As an ideal logistics automation transport device—AGV(Aautomatic Gguided Vehicle)have a wide range of applications in the manufacturing,logistics,warehousing and other industries.With the expansion of the scope of application and in order to meet the demand of production tasks,AGV need improve their guidance and positioning accuracy,and can sense its global position in the production environment and optimize the scheduling.This paper summarizes the existing technology of vision-guided AGV,researches the accurate estimate of path in visual guidance technology,precise parking location technology and multi-sensor information fusion technology of global positioning.This article focuses on the factors which affect the accuracy of guidance,sets up visual sensor model in which the camera installs obliquely,and performs the precise calibration;In the image preprocessing stage,this article does joint correction for the image of the three kinds of distortion,and enhances the image effectively through the improvement of rapid weighted median filter;Through the adaptive multi-threshold segmentation to get binary image,this article rasie the center path contour feature extraction method to simplify path model with the point set;Through the slope variance method to realize the path model classification,this article uses RANSAC algorithm to eliminate noise points from the center feature points,improves the accuracy of the path fitting and the robustness of parameter estimation when the road is path damage,pollution,and distraction.In order to solve the problem of AGV parking ability and accuracy,the article establishes the visual real-time measurement model,and realize the transformation of the image coordinate system for the feature point coordinates to the physical location of the actual scene;According to the magnification of the pixels in the measure model as multiwindow static partition,and according to the linear relationship between image brightness and intensity of illumination,the image window can be divided dynamically,the article combines with the result of two kinds of classified methods to realize the image window of dynamic division;To extract the center coordinates of characteristic symbol,extracting the point set of the center in the horizontal and vertical direction with the row and column scanning ways,and by the least squares fitting method,the article fits out of the intersection of two straight lines as a marker of ellipse center,and uses different control strategies in different window,realizes accurate parking location for AGV.Then,paper aimed at the problem of global positioning ability of vision-guided AGV,puts forward the method using its own visual sensor to extract feature as the global position,using inertial measurement unit as an external sensors to obtain real-time speed,angle and displacement of AGV;During the path without features mark,AGV locate itself by the Inertial Measurement Unit(IMU)sensor;During the path with the characteristic symbol,AGV obtain the body posture accurately by the information fusion algorithm of EKF which combines visual data and IMU information,eliminate the cumulative error,and achieves the function of AGV global localization with combination of the known topology map.Finnally,paper based with the technologies of vision guidance and global positioning,developes the omnidirectional vision-guided AGV and does a large number of experiments with the AGV,experimental results show that the proposed visual precision guidance technology and accurate positioning method is feasible and effective.