Pose estimation is an essential task in many mobile robot navigation systems. Visual guidance provides a feasible means for pose estimation using the observed scene information as reference. This work presents an approach to estimate the pose of a mobile robot based on projective transformations. First, the Hough transform is used for lane detection. Next, a projective transformation is computed using the detected lines as reference. Finally, the robot's pose is estimated from the resulting projective transformation. The theoretical principles and computational implementation are analyzed. Experimental results of a visual navigation experiment are presented to validate the usefulness of the proposed approach.
Length measurements provide important information about the three-dimensional world. This is especially useful for decision making in robot vision, path planning in autonomous navigation, and people identification in security application. In this work, we present a length measurement method based on perspective transformations using an uncalibrated camera. The theoretical principles are analyzed and the computational implementation is discussed. The usefulness of our proposal is verified experimentally by measuring relative lengths from experimental monocular images.
Uncalibrated camera-projector fringe projection systems are unable to provide metric three-dimensional measurements. The main difficulty for camera-projector calibration is that independent calibration of the devices is cumbersome and susceptible to alignment errors. In this paper, an efficient and accurate method for calibration of a camera-projector pair is proposed. The operating principle and computational implementation are analyzed. The metric measurement of a three-dimensional object is carried out to demonstrate the efficiency and accuracy of the proposed method.
The capture of panoramic images requires the use of complex and specialized cameras. However, high quality panoramic images can be constructed digitally by stitching several images captured with conventional lowcost cameras. In this work, an image stitching method based on projective transformations is proposed. The theoretical principles and computational implementation are presented. Experimental panoramic images are composed to validate the usefulness of our method.