Vision System Design Of Zanthoxylum Bungeanum Maxim Picking Robot

Chinese prickly ash is one of the important condiments that must be lacked on the dining table,but the growing environment of Chinese prickly ash is mostly mountainous,the terrain is complex and irregular,especially in Gansu,Sichuan,Qinghai and other places,the terrain is more rugged,due to the height of the Chinese prickly ash tree is 2-5 meters,the epidermis of the Chinese prickly ash branch is mostly prickly,the growth of Chinese prickly ash is mainly based on the cluster,and the fruit particles of each cluster are relatively scattered,etc.,Picking the fruit is very difficult.At present,it is still manual picking,which is very difficult.In view of the above many complex situations,it is urgent to develop a pepper picking robot that can independently move in the field and complete the whole process of pepper picking collection,so as to solve the automation problem of pepper picking.Pepper growing in an unstructured environment,pepper fruit growth characteristics are uncertain,pepper picking robot vision system is the key technology of pepper automatic picking.Based on the scheme design of pepper picking robot system,this paper designs the system for pepper picking vision system,and finally obtains the three-dimensional world coordinates of the robot picking point for pepper picker The realization of artifact and human lays the foundation of theory and practice.(1)The establishment of local map of pepper farmland.The pepper picking robot uses the orb feature extraction visual Slam(simultaneous localization and mapping)system,which is the real-time positioning and local map building technology of the robot.In this paper,the local sparse map of farmland is established,which lays the foundation for the harvesting robot to move in the field autonomously,to avoid the obstacles in the field intelligently,and to complete the fruit transportation autonomously after harvesting.This requires the robot itself to have strong ability of autonomous movement and obstacle avoidance,and the vision slam technology is the foundation for the realization of this ability of the robot.(2)The algorithm of image recognition.Image segmentation is the key to locate the target,and accurate segmentation of the target is the key basis for the subsequent centroid location,pepper picking point location,and parent branch extraction.Through analyzing and studying the performance of various image segmentation algorithms,such as threshold segmentation algorithm,fuzzy C-means algorithm,maximum entropy segmentation algorithm,K-means clustering algorithm and so on,the image segmentation algorithm suitable for prickly ash target extraction is determined.Through the experiment,the K-means clustering algorithm can extract prickly ash target stably and efficiently.(3)Identification of the mother branch of Zanthoxylum bungeanum L.Because the growth characteristics of Zanthoxylum bungeanum are uncertain,the traditional method of locating the picking point by gravity is not practical and the location of the picking point of Zanthoxylum bungeanum bungeanum bungeanum bungeanum bungeanum bungeanum bungeanum bungeanum bungeanum bungeanum bungeanum bungeanum.In the lab color space,by analyzing the color distribution characteristics of pepper image,this paper uses the maximum entropy threshold segmentation algorithm and Otsu algorithm to extract the pepper string,and then the pepper string and the extracted pepper fruit are made image subtraction,finally the pepper fruit is the result of the mother branch.(4)The location of pepper picking point,through the study of the growth characteristics of pepper,we know that the extension line of the straight line where the parent branch is usually through the center of mass position of pepper string,using Hough algorithm to detect the straight line on the parent branch,using the shortest distance from the point to the straight line as the limiting condition to filter out the redundant straight line,and get the midpoint of the straight line as the final picking point of pepper.(5)To obtain the depth information of the picking point,the pepper picking point is the point of the three-dimensional world,and the image is the plane 2D data.To obtain the three-dimensional world coordinates of the picking point,we must obtain the depth distance between the picking point and the camera,and then obtain the real three-dimensional coordinates of the picking point to complete the picking task.In this paper,we use orb feature point extraction algorithm to extract the feature points of the parent branch of prickly ash,then form a convex hull of the extracted feature points,use the convex hull algorithm to calculate the irregular shape of the convex hull area,use the basic principle of pinhole imaging to bring the camera’s front and back position into the formula,finally find the depth of prickly ash picking point,and then turn the picking point in the image from the image coordinate to the phase Finally,the machine coordinates are converted to the three-dimensional world coordinates,and the threedimensional world coordinates of the picking point are finally obtained.(6)Field experiments.In Gansu pepper picking field visual image collection,verify the feasibility of this paper pepper picking visual system.