论文部分内容阅读
An omnidirectional vision has the ability to obtain larger field of view and richer information of environment than normal view vision,and has wide applications in mobile robots.Research on omnidirectional vision involves measurement,calibration,map building and localization,etc.,and is a hot topic in mobile robot field.There are several challenges in omnidirectional vision,such as mapping analysis between image space and physical coordinates in 3D space,because of the severe deformation with the nonlinearity of the vision model.In this thesis,an omnidirectional vision system with laser illuminations is developed,and the calibration and imaging are investigated to achieve measurement and map building for a mobile robot.Principal of omnidirectional vision with the spherical mirror is investigated,and the general mapping relationship between image space and the laser plane is given to obtain environmental depth information.OcalibToolbox is employed for calibration of the omnidirectional vision system,and the laser plane constrain is introduced to vision model to enable vision system to measure the coordinates of a viewpoint.Image unwarping is investigated based on the omnidirectional vision system.Using the extra depth information of the vision system,an image unwarping method is presented to obtain a panoramic image.The omnidirectional vision system is applied in a mobile robot.Global map building with local depth information of the mobile robot is generated.The map data is fused with the images collected at different locations to improve map quality for mobile robot navigation.