文档介绍：Abstract
Vis ion s ystem has lo ng been a key research fie ld of robot techno logy. The funct ion of vis ion syste m on robots equa ls to that of the eyes for human being, which provides large a mount o f optica l infor mat ion for body control. Vis ion system is a dominant sens ing s ystem for mobile robots for it is t he source of surround ing informat ion. The system work in two steps, putes out the depth value for each pixe l of the reference ima ge of the ima ge pair collected by two hor izontal paralle l fixed cameras, second realized 3D terrain reconstruct ion by us ing the depth va lue and the p lane coordinates of each pixe l. Robots do path planning according to the reconstructed 3D terrain obtained by the vis ion processing syste m, whic h means the system indirect ly instr ucts the mot ion and control of the robots.
Stereo matching is the core algorit hm of 3D reconstructio n of the binocular system, it is characterized by put ing amount, restrict ion between accuracy and speed, poor robustness for vario us image collect ing cond it ion. The put ing amo unt attributes to large ima ge infor matio n memor y costs, since ima ges is stored by each pixe l which contains mult i- dimens ion vector recording color value of 3 channe l, pixel coordinate and intens ity va lue. Though it cost high storage space, it has single store pattern for each pixe l, therefore, image object is quite proper for paralle p ut ing. According to the above reasons, a paralle l based stereo matching system us ing CUDA was proposed to accelerate the image processing speed. The syste m was devised based on paralle l processing functio n of GPU hardware to realize real- time stereo matching. The system contains paralle l a lgor ithms for image read and write, pre- processing, disparity ca lculat ion and d isparity re fineme nt. In order to obtain better matching result, the system set several adjustable parameters to control the final e.
Through the study of tradit iona l stereo matc