Proposal of interface to recognize depth error to arbitrary points while constantly observing 2-D image consisting only of information from robotic perspective and practical issues

Abstract

In this paper, "the stereo matching interface" as a new visual recognition interface for remote operations is proposed. It satisfies the following functional requirements, which have been developed through a survey of prior arts. The stereo matching interface allows operators to recognize depth information about position errors to any remote target objects, 1) only uses information interpreted from the robotic perspective, 2) superimposes depth information on the image without separating the display area, 3) is a planar (2-D) image, 4) measures and presents depth information in the parallel to elevation angle direction and 5) allows users to selectively use arbitrary point depth information on the image. As a design that satisfies these functional requirements, we use that the parallax of the stereo camera images for the same spatial feature points changes with depth distance. The feature of this interface is that one of the images is edged by image processing to improve visibility, and it uses the human mechanism of selective attention. Furthermore, it was confirmed that the stereo matching interface satisfies the above functional requirements and fully performs the basic function. It has advantages of being able to recognize the presence or absence of depth distance errors over the entire display area (functional requirement 4 and 5) and to complete main operations using only intuitive images from the robotic perspective (functional requirement 1 and 2), without additional operations to identify any target objects. Furthermore, it is suggested that the following design values can be obtained: the reduction of falling and collapse of target objects, the synergistic effects in large scale-spaces, and the ease of integration into conventional equipment. Therefore, an alternative interface worthy of comparison with prior arts is considered to have been provided. On the other hand, the depth position error during turning and the convergence angle shift are practical issues. Besides, usability issues are the generation of depth indexes in the absence of spatial feature points on the surface of the target object and the visibility of the depth indexes.