Research On Vision-guided Technologies Of AGV Complex Path Based On Illumination Constraints

Automated guided vehicle (AGV) has broad application prospects as an automatic materialtransmission equipment in modern manufacturing system. Due to real time ability and robustness,vision-guided AGV has not been widely used. On the basis of reviewing the current researches onAGV and vision-guided technologies, the paper deeply researches the key technologies ofvision-guided AGV.Firstly, the working principles and image processing algorithms of vision guidance system arediscussed in this paper. To solve the problems of non-uniform illumination and high light phenomenacaused by the coaxial annular LED array light used for a vision-guided AGV, an image processingtechnology for non-uniform illumination based on illumination constraints is proposed. It improvesthe reliability of feature extraction for guide path. Multi-branch path recognition is a key problem inthe practical application of AGV. A method based on region descriptor and minimum spanning tree ispresented for models of cross turns and circular turns. Firstly, a method based on region descriptor isused to extract path features. Then, the support vector machine based on minimum spanning tree isapplied to obtain optimal classifier surface. Finally, simulations and experiments at variousenvironments verify the validity and reliability of the method.Secondly, the multi-camera vision collaborative guidance technology of dual drive AGV isdeeply researched in this paper, including collaborative camera calibration and multi-camera visioncollaborative guidance. A method of multi-camera vision collaborative guidance in the motion case ispresented. The method can accurately calibrate the position relationship between not only multiplecameras and camera vision center but also camera vision centers and drive units. With this method, aunified world coordinate system can be established. On this basis, the circular turning process of thedual drive vision-guided AGV is analyzed in the paper. To optimize the circular path, a method basedon multi-camera vision collaborative guidance is proposed. The minimum turning radius of the dualdrive AGV is effectively reduced by using this method.Finally, the main research results of the thesis are summarized. The difficult problems andvaluable directions for further researches on vision-based AGV are pointed out, which are expected tobe solved and improved gradually in future studies.