Effective environment perception has become a research hotspot for mobile robots. As an important perception means, monocular vision has been widely used with broad potential applications due to its advantages of small size, fast response and flexibility. This thesis conducts the research on monocular visual measurement for mobile robots. By utilizing priori knowledge, the visual measurement models are built to improve the application convenience of the mobile robot¡¯s visual system. The main contents and contributions are as follows:
Firstly, the significance of visual system for mobile robot localization and measurement is analyzed. The research development of localization and measurement with a monocular camera is then addressed from the following three aspects: plane visual localization of fixed camera, PnP visual localization, and onboard camera localization. The contents and structure of this thesis are also introduced.
Secondly, the vision-based mobile robot following is investigated for the leader-follower robotic system, where a tracking identifier is attached at the rear of the leader robot and the follower robot captures the corresponding information by a CMOS camera. An approach based on priori knowledge to extract feature point and feature line segments is presented, and a single-point position measurement approach based on the Faugeras method is designed to estimate the position of the leader robot relative to the follower robot. On this basis, a logical controller is designed to achieve the dynamic following for the leader-follower robotic system. Regarding image-based visual servoing, a fuzzy controller is designed where the image area of the identifier and its variation are regarded as the inputs. The output of this controller is the duty cycle, which is used to control motors. A longitudinal following for the leader follower robotic system is then achieved.
Thirdly, the environmental measurement based on single image information is addressed by a mobile robot with monocular vision. The visual measurement model is established. The height of object is measured according to the vertical constraint of feature points. Take the office scene as an example. The feature points are extracted from one frame of image captured by the camera. After 2D positions of feature points on the ground are calculated, the geometric relationships of feature points are then utilized to obtain the height information of the objects.
Fourthly, a position and orientation measurement approach based on planar constraints is proposed by using a ceiling-fixed camera. The mobile robot is equipped with a tracking identifier, and its feature points are extracted from single image. With the visual measurement model, the position and orientation of the target robot are computed.
Finally, the conclusions are given and future works are proposed.