Research On Visual Positioning Algorithm Of AGV Loading And Unloading System

Traditional loading and unloading systems often have low productivity and low precision because of the artificial restrictions.With serious demand for automation and intelligence,more automated and intelligent ways to replace traditional inefficient methods have been proposed.Nowadays,the research of domestic universities and research institutions in this area is also endless in order to respond to the call of national smart manufacturing.Under such a big background,it's meaningful to study the visual positioning algorithm to make the AGV loading and unloading system automatic.Firstly,this paper introduces the workflow of the automatic loading and unloading system and the composition of the mechanical structure,in this way,the system design specifications and the requirements are clarified.Then the core components of the machine vision system and the motion control system are selected,and the hardware of the system is completed as well based on the actual site requirements.In order to reduce the influence of different noises on the image,an adaptive Gaussian filter algorithm is used to filter the noises after comparing different noise reduction algorithms.And then,it is also necessary to enhance the target region of the image and suppress the background by using image enhancement technology for obtaining a high-quality image.Next,an improved image segmentation algorithm based on region growing is proposed to segment the image.The algorithm uses the histogram statistics method to obtain the best threshold firstly,then uses the region growing algorithm to segment the image,gets a good segmentation effect finally.A rotation correction algorithm is used to solve the problem that the image may be distorted and the bilinear interpolation is used to process the corrected image for maintaining the integrity of the edges of the image after comparing different interpolation algorithms.The paper studies a detailed process of the image processing in order to position the Marks.For location position,the paper uses Canny operator to extract circular edges firstly after a series of image processing including image preprocessing and image segmentation and etc.and then applies the RANSAC circle fitting algorithm to fit the circle edges.And the center coordinates are acquired in the end.For angle position,the paper mainly applies the gradient algorithm based on interpolation to extract the board edges after comparing different sub pixel linear edge extraction algorithms.And then uses an adapted RANSAC straight-line edge fitting algorithm to obtain the angle of the board edges.To improve the efficiency of algorithm,the paper proposes an improved algorithm of template matching.The algorithm uses the pyramid-search strategy combined with the optimization of the processing domain and the use of integral graph to achieve the goal.In the end,the paper analyzes the feasibility,stability and time-consuming of the system by a lot of experiments which showed that the system has the ability to achieve the goal of automatic loading and unloading and meets the actual production requirements towards the accuracy.And then the paper makes a detailed analysis of the design ideas,main functions and the components of the software which uses a modular thought to connect all parts together tightly and orderly.