During the presentation Mark explained that MIRA had chosen to utilise a vision system for edge detection rather than using LIDAR or RADAR, as the engineering consultancy’s UGV group had already used LIDARs and RADARS extensively over a number of years. It was felt that it would be beneficial to expand the knowledge of visions systems; what they could offer and their limitations. Mark also pointed out that the cost of vision cameras needed for this type of research can be an order of magnitude less than a LIDAR device and considerably cheaper than a commercial RADAR unit, which in turn would pass on cost benefits to customers once the technology had matured. Engineers at MIRA also aimed to feed edge detection and tracking algorithms developed from vision system work back into the company’s other LIDAR and RADAR systems.
Delegates were impressed to hear that unlike most other road edge detection programs which initially remove the sky from the image in order to reduce the amount of computing power required, MIRA’s program was capable of removing everything from the image except a narrow band that they termed a ‘letterbox’. Mark explained this approach, saying that the team had theorized that there was enough information still in that band to adequately find and track road edges. By applying various filters to the image and other computational transformations, MIRA found that the detection stages could be run more frequently – allowing the vehicle to be driven faster whilst maintaining a suitable level of accuracy. In order to further develop this cutting edge system, Mark highlighted that MIRA was planning a number of improvements to the system including detection in low or artificial lighting conditions.