Night Vision Assistance: Notes

DESCRIPTION 

Courtesy of CHRYSLER GROUP, LLC

The major components of the Night Vision Assistance (NVA) system are:

The NVA system is a driver assistance feature that enhances driver vision in low-visibility conditions. The NVA video displays on the IPC display and it notifies the driver of a pedestrian or large animal on its line of sight. The detected objects are highlighted and boxed on the IPC video display with different colors while in different warning zones (normal is yellow, critical is red). It also provides audible chime requests and additional visual icons on the Heads-UP Display (HUD) during critical warning zone events.

The system is assembled with one NVPM performing image processing tasks and adaption to the vehicle.

OPERATION 

OBJECT DETECTION FUNCTIONS -

Vision processor performs three main functions:

1. Enhance and adapt the image for display in the vehicle
2. Detect and highlight pedestrians and present warnings to the driver
3. Detect and highlight animals and present warnings to the driver.

ANIMAL DETECTION -

Animal Detection is available when the vehicle is in a rural environment. The system is highly likely to detect an animal correctly if:

• The animal is of the deer family (roe deer, fallow deer, red deer, reindeer, caribou, moose, cow, horse etc.)
• The animal is in an upright pose.
• The animal is moving.
• Most part of the animal body is visible.
• Distance to the animal is between 7 meters to 70 meters (22 ft - 229 ft) for an animal with Withers height of 1 meter (3 ft). The Withers is the ridge between the shoulder blades of an animal, typically a quadruped. In many species, it is the tallest point of the body.
• Distance to the animal is between 13 meters (43 ft) -to 140 meters (460 ft) for an animal with Withers height of 2 meters (6.5 ft.).

The system may detect correctly while giving degraded performance if:

• The animal is of the pig family.
• The head of the animal is not visible.

The system is not likely to detect correctly if:

• Most of the animal body occluded.

Adversely the system may incorrectly detect something as an animal if:

• The heat signature in the image is similar to that of the shape of an animal. Some examples of these types of structures are sometimes found on various object such as warm stones, parts of buildings, parts of vehicles or street lights.

PEDESTRIAN DETECTION -

The system is highly likely to detect a pedestrian correctly if:

• The head of the pedestrian is visible.
• The pedestrian is in an upright pose: standing still or moving.
• Most of the body is visible.
• Distance to the pedestrian is between 8 meters (26 ft.) and 90 meters (295 ft.) for a 170 cm (5.5 ft.) adult.

The system is highly likely to detect correctly even if:

• Other objects such as a vehicle, baby carriage, dog, etc. are close to the pedestrian.
• Many pedestrians are walking close to each other in a group. The algorithm most likely detects the most visible pedestrians (which are not occluded by others) in the group or the ones who are most separated from the group. In consequence, the most 'dangerous' pedestrians are detected.
• The object is a cyclist, and the cyclist is in an upright pose similar to a pedestrian.

The system is not likely to detect correctly if:

• Head of pedestrian/cyclist is not visible. For example, head is covered by an umbrella or a helmet.
• Pedestrian is bending, sitting down, on knees, lying etc. Cyclists leaning forward on the bike will most likely not be detected correctly.
• Most part of body is occluded

The system may incorrectly detect something as a pedestrian if :

• The structure and heat signature in the image is similar to the shape of a human. Such structures are sometimes found on various objects, for example parts of vehicle tires, house facades, street lights or vehicle back lighting.

WARNING FUNCTION -

All detected objects will get a yellow enclosing box to highlight it in the image. Objects which are evaluated to be at risk will change color to a red box and a Controller Area Network (CAN) message is transmitted for the alert which can also be used to trigger other warning systems in the vehicle like sound alerts, symbols in the HUD or a combination of these. The message will contain information about what type of object (animal, pedestrian) it is and on which side the object is located (left, center, right).

IMAGE ENHANCEMENT -

The camera image is enhanced and adapted to the vehicle display to generate a grey level image suitable for night driving. Warm objects in the scene will be bright in the image and cold objects will be dark (this is called black-hot image for thermal systems). The scale of the grey levels are relative but non-linear. Detected objects are highlighted by a colored box.