This week, the Road Watch prepared for its first Qualification Review Board, or QRB1. At a QRB, our team will present our project to experienced faculty, who will offer feedback on the project. The main objective is to pinpoint risks in the project, and to devise mitigatory efforts. In the context of product development, a risk is an either anticipated or unanticipated factor which might hinder the development of the product. An example of a currently identified risk is that binocular vision’s increasing uncertainty with distance could render the computer vision approach to this project unviable (see the figure below for a more in-depth explanation). In the QRB, our team plans on giving a quick overview of our project, and then each member will present their current progress and discuss associated risks. Check in next week to see what feedback we got for our QRB1.
In binocular vision, two cameras are used to capture the same scene from slightly different perspectives. This allows the calculation of positions of objects in the scene, relative to the cameras. This is how the human visual system works, and offers a sense of depth perception. A drawback of this approach, however, is that the uncertainty in position calculation increases rapidly with distance. This is demonstrated in the figure above, which shows how a position in the scene gets mapped onto the X-axis of each camera’s image (note that the pixel sizes are greatly exaggerated to illustrate this principle). As the position gets further from the camera array (which is centered at (0m,0m)), a larger surrounding area gets mapped to the same pixel, which increases the uncertainty of the measurement. This makes it very challenging, or perhaps even impossible, to measure the positions of distant objects (such as cars on a highway).
