| Summary: | The Engineering in autonomous systems has many strands. The area in which this work falls, the artificial vision, has become one of great interest in multiple contexts and focuses on robotics. This work seeks to address and overcome some real difficulties encountered when developing technologies with artificial vision systems which are, the calibration process and pose computation of robots in real-time. Initially, it aims to perform real-time camera intrinsic (3.2.1) and extrinsic (3.3) stereo camera systems calibration needed to the main goal of this work, the real-time pose (position and orientation) computation of an active coloured target with stereo vision systems. Designed to be intuitive, easy-to-use and able to run under real-time applications, this work was developed for use either with low-cost and easy-to-acquire or more complex and high resolution stereo vision systems in order to compute all the parameters inherent to this same system such as the intrinsic values of each one of the cameras and the extrinsic matrices computation between both cameras. More oriented towards the underwater environments, which are very dynamic and computationally more complex due to its particularities such as light reflections. The available calibration information, whether generated by this tool or loaded configurations from other tools allows, in a simplistic way, to proceed to the calibration of an environment colorspace and the detection parameters of a specific target with active visual markers (4.1.1), useful within unstructured environments. With a calibrated system and environment, it is possible to detect and compute, in real time, the pose of a target of interest. The combination of position and orientation or attitude is referred as the pose of an object. For performance analysis and quality of the information obtained, this tools are compared with others already existent.
|
|---|