To answer the question, we should understand what is the exclusive capability for cameras, compared to other established sensing technology radar and lidar.
Exclusive capability benchmark
Camera is the only sensor that can be relied on to locate and read the road lane and traffic sign. But for many weather and environmental (W/E) conditions the camera would lose its reliability, such as in fog, rain and snow. This is also the reason nowadays we see the coexisting of different types of sensors: Camera, Lidar, Radar and so on. They are complimentary in different use cases and W/E conditions.
We can’t say which one of them is replaceable for the current AV development stage, because the redundancy is very much necessary for security consideration. What we know now is that cameras and radar have already become a mandatory sensor for ADAS, AV, due to their function, performance, technology readiness and cost.
Sensing Technology
What needs to be mentioned is the very much discussed Lidar technology in the past 5-10 years. Already the lidar market is crowded with hundreds of providers, and many well known car makers have already launched their mass production and have it installed on their cars. Among those models, time of flight (TOF) is still the dominant technology. In the meantime, Lidar and camera data fusion already shows the ability to improve detection accuracy.
As could be seen from the above table, lidar sensors in cars are more or less a bridging technology which is filling the specification gap between camera and radar, from the technical parameters like resolution, detectability, data consuming etc. Would it be able to replace one of the two sensing devices, especially the camera?
Exclusive Character of Camera
Camera detects and recognizes objects relying on the good quality of the picture, and the algorithm is relying on the information like color, contrast, texture, which current lidar technology point cloud does not contain. From this point, lidar couldn’t do the same job as cameras for road mark and traffic sign recognition. But the drawback of camera systems besides the incompatibility under bad W/E conditions, is the computation cost, cameras generate dozens times more data every second compared with lidar point cloud. And then it needs sophisticated algorithms for object detection and classification.
Some Issues with Lidar...
Outlook - Is Lidar able to substitute Camera?
Lidar with FMCW technology which has higher resolution, longer distance detection range, better worse weather compatibility, still would need another 5 to 10 years to develop and evolve. The interesting information FMCW lidar provides is the speed besides distance and spatial coordinate. This information is very useful for highlighting and screening moving objects, which may be very helpful to accelerate time and computation cost of object recognition for cameras.
Conclusion
As the current stage of AV development, camera and radar are both mandatory sensors for analysis and detection. Lidar is very much necessary to provide redundancy and environmental 3D representation. For the exclusive capability that camera has, for the coming decade, it can’t be replaced by lidar or other sensors.
Comments