BELLAROMA

While at Electronic Imaging 2023 in California with the support of UL and Lero, Dept. of E&CE's Patrick Denny spoke with Robert Dingess whose consortium controls 70% of the road marking materials in the USA and Europe and Robert indicated that all road markings have been designed with the Human Vision System in mind and he asked if we could consider instead how a computer vision algorithm sees the world so that we could modify the characteristics of road markings accordingly. With this in mind, Patrick put him in touch with a friend of his, Paola Iacomussi of INRIM in Italy who looks after road markings for Europe, and the idea got traction. With that in mind, Patrick came up with a name for the associated research programme, BELLAROMA (Better Live Lab with Road Markings), which is also the Italian for Beautiful Rome as it is important that ideas have a cool name so that they can be talked about and used to build a consortium. Patrick Denny spoke about it at AutoSens in Brussels in September in a panel discussion with fellow programme members Alex Braun and Robert Dingness, while Paola Iacomussi concurrently presented a talk on it at the 30^th session of the CIE in Ljubljana in Slovenia.

The new proposed approach is necessary as the optimisation of consequential imaging systems (such as in self-driving) is predicated on optimising for characteristics of the human visual system (HVS) instead of optimising for the sensing characteristics of a computer vision system (CVS) and historically the stakeholders along the video chain from the physical environment to an actuator (e.g., the steering system of a self-driving vehicle which does or doesn't hit a pedestrian) have not considered the problem in this way, due to their immediate motivations and constrained perspectives.

Hopefully, this research programme will make the environment around automatic computer vision systems (including self-driving cars) a safer place.