Perfect parking

A low-cost 3D camera adapted from military range-finding technology is the key to a spatial sensing system that could enable all cars to park themselves in the future.

The camera, which was demonstrated for the first time at the Society of Automotive Engineers World Congress in Detroit this week, could be manufactured for as little as $30 (£17) each, according to developers Qinetiq.

The company said it is working with several Tier 1 suppliers on a system that would be affordable for the mass market.

At present, ‘autopark’ and collision avoidance devices rely on bulky and expensive radar or laser sensing units and tend to be available only at the luxury end of the market.

The reduced costs of the 3D camera has also attracted interest from the retail clothing and footwear sector, which is keen to achieve better garment selection and fit.

Qinetiq said commercial applications of the technology are expected to start appearing on the high street in the next 12-months.

The ‘civilian’ applications follow work carried out for the MoD to improve the range mapping and target identification capabilities of the armed forces.

Such activities must avoid detection by the enemy, so a passive or low-power technology is best, and tends to mean the unit will also be reasonably lightweight and relatively cheap to produce.

Conventional scanning lasers are bulky, power hungry pieces of equipment. Other 3D systems such as fringe projection technologies rely on large condenser lenses. Both are also expensive to produce.

Instead, the Qinetiq device uses an array of light emitting diodes to project dots of infrared light into a space. A standard camera based on a CCD picks up the dots as they intersect with objects in the scene. The system can produce 3D co-ordinates for each dot in each frame of the camera, at a rate of five frames per second.

John Bannister, commercial development manager for optronics technologies, said Qinetiq has developed several prototypes to demonstrate the core technology to carmakers.

At present, each camera is capable of capturing data from over 1,000 individual points on an object with a range accuracy of 0.25mm across a wide field of view and with a working range of between 200mm and 2m.

For future automotive applications, the camera’s range could be increased, said Bannister, to 5m and further by reducing the number of diodes thereby boosting the intensity of the projected spots. On the other hand, for changing-room cameras in retail outlets, the number of spots might eventually be increased to 5,000.

Bannister said a 300-spot, long- range projector might eventually be developed for sophisticated automotive control applications. A range of 20 – 50m would be necessary for collision avoidance at speed.

At present the aim is for a system that will place one camera in each of the four main light clusters capable of picking up bollards and low kerbs at low speeds. ‘Existing ‘bumper-sensors’ can have limitations as they can miss low or thin objects,’ he said.

The cameras could also be used for lane detection and management when the vehicle is moving at speed. Operating in real time, with data being processed in each unit, only a warning is sent to the dashboard.

‘Ultimately they could be integrated with other onboard systems to enable a vehicle to park itself, or get out of tight parking spots.’

Bannister said Qinetiq was also investigating the use of the camera to monitor people in the passenger compartment to calculate the most effective deployment of airbags in the event of a crash.

‘Car companies are interested in enhanced intelligent safety because it is becoming more important to the consumer, perhaps even more so than performance and looks,’ said Bannister.