Font Size: a A A

Application Research Of Monocular Visual Capture Technology For Supply Robots

Posted on:2019-02-18Degree:MasterType:Thesis
Country:ChinaCandidate:P YangFull Text:PDF
GTID:2358330548450456Subject:Computer technology
Abstract/Summary:PDF Full Text Request
The human brain obtains information primarily from the outside world through vision.Today,with the development of artificial intelligence,whether it is military,space or manufacturing,it is increasingly important to strengthen the visual technology of robots.With the penetration of e-commerce into three-and-fifth-tier cities,China's express delivery industry is developing at a rapid rate.In the face of a large number of express parcels,Jingdong took the lead in building the first full-process unmanned warehouse,replacing manual operations with a series of robots.In this paper,the actual situation of the courier grab robot in the unmanned warehouse is the supply robot,the main work is as follows:1,Aiming at the actual situation of the supply robot,a monocular visual courier positioning algorithm based on express single-item recognition is designed.The algorithm uses the express color of the courier on the express delivery and has a fixed shape,and extracts the four fixed express orders through image processing.Feature points,complete 3D-2D coordinate matching,this method can avoid the current 6-DOF monocular visual object capture process,the need to capture multiple objects in different positions of the picture to match the way to obtain the time-consuming three-dimensional coordinates of the object Operation;then the paper compares the accuracy of the point-based EPnP method and the parallel line method based on line features in the pose estimation,and chooses one of the better ones for the next express pose estimation.The results show that this method can achieve fast express positioning of single-image single images without the assistance of other distance sensors.2,In order to accurately extract feature points on express delivery.For the cases where the express corners are not obvious and the surface is uneven,an algorithm that combines LSD line detection and K-Means clustering is designed.After recognizing the express delivery list,the LSD algorithm detects the straight line.However,due to the unevenness of the surface of the monad,multiple straight lines are detected on each edge.When directly obtaining the straight line intersection,more than four feature points will be obtained.We will connect the connected fields.The center is perpendicular to each detected line,and the vertical points are clustered by the k-means,so that the lines are divided into four categories according to their locations.Since the long lines contain more information,we select the length of each class.The longest straight line is used to determine the intersection point,and the positioning of the feature point is finally completed.Experimental results show that the method proposed in this paper has achieved good results.For the position estimation of express delivery,the position error of XY is within 2cm,the distance error is about 3mm,and the average angle error is within 2°,which can satisfy the robot's grasping express request.
Keywords/Search Tags:Supply robot, monocular vision, location identification, pose estimation
PDF Full Text Request
Related items