Coprinus Comatus Identification And Localization Research Based On Binocular Stereo Vision

Wit h short growth cyc le and high nutr it io na l va lue,C oprinus comat us has been cult ivated in a large scale across the countr y.However,during the production work of Coprinus comat us,30% to 50% of t he labor force is wasted in t he pick ing link,where such proble ms as high time concentrat ion and big labor inte ns it y lead to low economic profits.Nowadays,the rapid develop ment of robotics and comp uter techno logy make it possib le to app ly p icking robots in agr ic ult ural into t he product ion process of Coprinus comat us.It is a prerequis ite and one o f t he key techniques to obtain the three-dimensional coordinates of Coprinus comatus during the producti on process.This paper aims to stud y the visua l s yste m of pick ing robot,composed of bino cular camera,it can perceive three-dimens io nal informat io n of objects co mpared wit h human's eyes.As for monocular,binocular vision can obtain three-dimensional infor mat ion of target object flexib ly.Besides,compared wit h mult i-View camera,it has the advanta ge of s imp le operatio n.Based on binocular stereo vis ion technolo gy,this paper stud ies the ident ific at ion and posit io ning of C oprinus comat us,inc lud ing ca mera calibrat ion,ima ge matching of C oprinus comat us,image segmentat io n and three-dimensional reconstruction.The main work and research results are as follows:(1)In this paper,the basic model o f the camera is introduced first ly,and then t he errors introduced in the product ion and assemb ly process are fully cons idered s o that the camera distort ion is ana lyzed.Then,by ana lyzing t hree kinds o f calibrat ion met hods,the two-step calibrat ion met hod between self calibrat ion and tradit iona l ca libration is c hosen as the established calibrat ion a lgorithm.F ina lly,by us ing the checkerboard to calibrate the binocular camera required in the experime nt,the inte rnal a nd externa l para meters of t he binocular camera and the distortio n coeffic ie nt of the camera are obtained thus the binocular correct ion of the ca mera is achieved,which provides basis for accurate image acquisition.(2)By us ing the ca librated camera,the author took pictures of C oprinus comat us and studied the exist ing image matching c lassica l algor ithm to ana lyze the advantages and disadvanta ges of SURF algor ithm in fi nd ing feature points and construct ing d escriptors,so as to propose an improved SURF optimizat ion algorithm.Wit h the alg orit hm ver ificatio n of C oprinus comat us image collected in the working environme nt and compare of improved algor it hm w it h SITF algor it h m and SURF algorit hm,the experimenta l result shows that the improved optimizat ion algor ithm inher its the SURF robustness and matching speed advantage,and,it a lso has a stronger descriptive abilit y.It can e liminate the mis match by us ing t he RAN SAC algor ithm and improve the accuracy of feature matching accuracy.(3)By ana lyzing the working princ ip le of two general met hods in image se gmentatio n and us ing two kinds of c lassica l segmentat io n algorit hms to segme nt the picture according to the image of C opr inus C omat us obtained in the robot working environme nt,the a uthor co mpared the ima ge qua lity a fter t he two segmentatio n a lg orit hms and t hen took the Grab cut a lgor it hm as the image se gmentat ion a lgor ithm,added the improved sur f algorit hm into the segme nt at ion algor ithm.The Grab cut algorit hm is used to segment the C oprinus comatus contour,whic h simp lifies the Grab cut algor ithm process thus improves the operation e fficie ncy.F ina lly,comparing four edge detection a lgor it hms to the edge detection of Copr inus comatus image,this system chooses canny algori thm for edge detection.(4)First ly,this paper introduces the princip le of three-dime ns iona l reconstr uction a nd then,analyzes the genera l model of three-dimens iona l reconstruct ion,precise ly ana lyzes the reconstruct ion model w ith checkerboard and sign.Finally,this paper calculates the three-dime nsio na l infor matio n of Coprinus comat us in working environme nt.The exper imenta l result shows that the reconstructed model has high reconstruct io n accuracy and the three-dimens iona l infor matio n reconstruct ion acc uracy o f C oprinus comatus in working environment ca n satis fy the e xperimental r equirement well.Based on Microsoft Visua l Stud io 2010 and open C V develop software,camera ca libration,C oprinus comat us feat ure extractio n and matc hing,Grabcut image se gmentatio n algorit hm,three-dimens io nal reconstruct ion and other functio ns are successfully imp le mented.The ident ificat ion and posit ioning of C oprinus comat us are also well completed.