Image recognition augments the driver’s vision, informing of perceived hazards by audible or visual methods - or both. Should the driver not react quickly enough, the ADAS could take action instead (e.g. initiating emergency braking).
For parking assist systems, things should be relatively simple. The image recognition technology needs to accurately determine the position of parking lines and stationery objects (such as walls or lamp posts), while being fully aware of moving objects (such as pedestrians) close to the vehicle. The challenges are greater in other environments, such as a busy crossroad in a densely populated city, where multiple hazards may be present. This can be exacerbated if the lighting conditions impinge on the ability of the ADAS to correctly identify the nature of these items.
Unsurprisingly, the driver will need greatest assistance from the image recognition system in areas where illumination levels are low, but to deliver this higher performance processing will be needed.
The TMPV750 and TMPV760 image recognition processors from Toshiba deliver high performance operation alongside low power consumption to provide a solution that is highly optimised for automotive implementation. These ICs perform massive multi-core processing, with hardware acceleration of complex algorithms. They can run multiple ADAS applications in parallel, so that vehicles, pedestrians, road signs and traffic signals can all be correctly identified. Through this data the ADAS is able to determine if some form of response is needed. These processor devices are able to cope with image data from up four cameras at once (and have provision for connection of up to eight different cameras).
Toshiba has developed a comprehensive white paper that discusses the various aspects of automotive imaging - including the forthcoming role that 3D technology is likely to play. To download it click here: