利用人工神经网络测量三维物体轮廓的研究-测试计量技术及仪器专业论文.docxVIP

硕士学位论文A 硕士学位论文 A Study on 3_D Profile Me船1lremci】t Usillg NeuraI Netl)l唧rk ABSTRACT Studying pro矧e measurement of 3一D objects h8s many imponam purpose．It can be used ln robocⅥslon，Indusfry automatic manuf如turc，etc．．This．paper西ve8 lhe comprehenstVe analyses about fhese methods．And it aIsO stIldies and verifIes another new measureme力t mefbod—0-D pro矗】e力oncontact measllf锄ent usjng neuraj netwo rk． This method compared to lrad{tfonal noncontact measurement methods has such advantages： 1．reduces the requircment for the卸curacy ofmeaStlrement system； 2．no need to calib豫te a|ld align the optica】system，no need to t∞t the measure瑚ent paramefers ofthe system accuratefy fn advance． This papcr presenls the detail process about accomplish profile measurement by LIsing one of artificial neumI ne咖rI(一Back-Propaga“on (BP)． netwDrk．This process includes construc“ng opt{c8l system．capturjng 8nd processing jmages， composing software，a嘶usting BP network and anatyzing the resuIts． Pro刚e mensurement using neuml nefwork is fumlled by two sceps．The fifst step is studying or lrainjng BP necwork，lhe second step is measuring objecls．First，building the system’s mapping relatjonshjp by t嘣njng BP network．Th砒is，detenn{njng t11e mapping reIationship of coordinates bdI、veen oCD camera jmnge plane and reI砒ive position in objeot‘s space，Projectjng Iaser syslem iIlumjnatcs lhe gauge block by producing stripe Iight，and CCD image 3ystem caph盯es imag∞．The c00rdinates Of line image of t}Ie ob}ect ofl CCD js ns t11e input da协of I}P netwofk aIaer some suitnble image processing． A11d the out舢t data is the con．esponding coordinates jn obiect’s space．}玷re we use№ kind of traⅫng met}Iods： the bntch processing锄d the individual processingt The咖 melhods have dif诧rcnt speed durhlg training BP netwofk，the former is鱼3ter than che Inter．But when used for’measurement，they ahnosC have the same resuIt．Du—ng fhe measuremen【s协ge，by rota“ng 360。measured object，samp“ng the Iine image丘口m the discrefe angular position。the coor