Solving the Visual Perception Challenge with GatedVision Technology

Nighttime driving is the most hazardous time of the day to drive. To make matters worse, current ADAS and AV cameras do not operate satisfactorily at night, when they are needed most. 

Fortunately, cameras powered by GatedVision technology can perform very well in darkness in addition to providing obstacle detection. Slicing can eliminate the background, and shadow detection can detect obstacles with zero contrast. In this article, we present the basic concepts of GatedVision and its detection capabilities. 

The immediate automotive implementation of a GatedVision system is to have the GatedVision camera side by side with the visible (standard) camera. This allows for a full visual spectrum and operational coverage in both lighted conditions and darkness.

The Ideal Vision Sensor

Let's evaluate the properties of an ideal vision sensor for automotive applications:

1.  Seeing day and night equally

There should be no difference in the level of safety between night and day. Nights are more dangerous due to several factors related to limited visibility. At night, we lack accurate distance estimation, we are often more tired, and our vision is degraded. 

2.  Weather

One cannot expect autonomous trucks or cars to stop when it rains or snow or fog - business continuity is at the core of the industry. For ADAS, adverse weather is one of the situations where drivers require safety equipment and systems to overcome human driver vision challenges. When using active systems, especially at night, the backscatter caused by the air carries aerosols and blinds the camera, affecting the detection range.

3. Dynamic range

Dynamic range is important when using active illumination at night. Since brightness in the short-to-long range is squared law proportional and safe automotive operation requires achieving long-range imaging. A target at 10m will appear 100 times brighter than the same target at 100m. The image sensor should be able to capture both targets without saturation and with reasonable SNR (>5). In the automotive sector, there are additional constraints, such as not being saturated and compromised by oncoming headlights, similar systems, and retro-reflectors.

4.  Relevant distance

If the visible sensor captures only the relevant range and ignores irrelevant backgrounds, it will allow less computation and a lower number of false positives. It will also increase the contrast, as seen in Figure 3, where the pedestrian is more clearly detectable with the GatedVision sensor than with the visible sensor. The contrast is higher in the GatedVision image than in the visual image. The range discrimination in GatedVision is defined as a range slice. The location and width of the slice vary in real-time and can correlate with the car’s speed.

‍

5.  Low cost

Up to L3 (SAE level), the price of adding sensors to the automotive sensor suite might be a limiting factor when it comes to mass-market quantities and deployment. At L3 and above, the cost of the sensor suite and the overall autonomous enabling system requires a higher budget. Having a low-cost solution for mass-market applications is critical. 

Truck OEMs allow for higher sensor costs - this can be accommodated especially in class 8 trucks on highway routes (hub to hub) where no driver is required. In addition, L4 trucks will prefer nighttime driving, which flows more freely, making them more fuel-efficient and environmentally friendly. This concept completely revolutionizes the economics behind long-distance hauling/trucking. Once again, a lower-cost solution makes it easier, faster, and more worthwhile for OEMs to upgrade their sensor suites to allow for nighttime driving.

The GatedVision Camera 

The requirements of the automotive image sensor are often versatile and demanding. GatedVision is the only technology that can address all these requirements and more. 

GatedVision technology provides the following capabilities:

1. Depth
   The GatedVision camera system uses overlapping range slices, allowing for the calculation of range to any point in the scene. The slice itself can provide a range at a lower resolution but is sufficient at the longer ranges.
2. Shadow
   All vision systems rely on contrast; if there is no contrast, no AI or human can detect a target lying on the road. Any active system generates a shadow in the scene. The issue is to differentiate the dark spot as either a shadow or a non-reflective material. GatedVision is capable of spatially modulating the shadow in the captured image within a single frame time (< 5 msec). A dark spot or a non-reflective material will not be modulated. 
3. Retro
   GatedVision systems can distinguish between defuse reflections and retroreflectors, which are not neutral and belong either to road markings, traffic signs, or the rear end of a car. This is helpful in the rain for LIKE/LDW and for AEB cars with or without lights.
4. Non-trained obstacle detection
   Slicing combined with shadow and passive NIR imaging can enable a new kind of detection. Even for non-trained obstacles. This means that the system is capable of detecting new obstacles it has never encountered before. 

OEMs are destined to overcome the limitations of hazardous driving conditions, including nighttime and adverse weather. They need to broaden the scope of their ODD (operational design domain) to ensure better safety, which can be accomplished by using GatedVision technology.

Moving Forward with GatedVision

While roads are less populated at night, nighttime driving still results in the highest number of casualties and fatal accidents, particularly when combined with adverse weather. A solution that allows both drivers and AV software to get a clear image in all weather and lighting conditions is now essential and on the shopping list of the most advanced manufacturers. 

GatedVision’s NIR wavelength imaging provides clear images with high contrast while supporting range discrimination as a detection enhancer and suppressing the interferences that come with adverse weather, including backscatter. 

‍

Additionally, the system’s inherent capability of overlapping slices means that a depth image can be extracted to provide an accurate enough depth map, as shown in recent studies.

GatedVision technology has reached maturity and is ready to provide the automotive industry with a high-performing solution at an affordable cost for any low lighting and adverse weather conditions to enable non-stop business continuity for both private cars and trucks. 

‍