BusinessWire: Aeye reports the independent validation of performance of its camera-LiDAR combo named iDAR:
- the AEye sensor was able to detect targets with a substantial number of points at a distance exceeding 1000 meters without compromising frame rate
- AEye’s sensor produced over 175 points on an 18% target at a distance of 200m, achieving the metric with 1600 points per degree squared
- iDAR achieved over 200Hz scan rate for full frame and field of view (FOV)
- the iDAR sensor works effectively behind a first-surface made of windshield glass material at a nominal thickness at various angles of incidence.
It's not clear what part of this performance comes from camera and what's from LiDAR part of the combo - everything is fused.
Light.co publishes a webinar of its Clarity comparing multi-camera setup with LiDAR:
An IP analytics blog LexisNexis believes that Ouster has the best quality of its LiDAR patents:
My 2 cents on the Aeye demo:
ReplyDelete2. I guess the 0.25 deg point resolution only occurs at the center FOV region but I have little issues with it as it's a reasonable design optimization/compromise in practice and most people do that.
3. The trick to pull a 200 FPS is to reduce the ROI and measure near distance objects. The most simple way to think of this is that light bouncing back from a 150m target takes 1us which fundamentally limits the # of points per sec unless a massively parallel channels are used like in a flash camera design whose range however is limited by other factors like optical power and data processing. A near-range high frame rate LiDAR is certainly useful in many applications that however may not afford such a LiDAR.
4. The argument about not using co-axial design preventing windshield scattering doesn't make a lot of sense when considering a practically wide FOV. One can simply draw a ray optics diagram to see it. Direct reflection from windshield does make a difference for co-axial or non-co-axial designs however a tilted windshield greatly reduces reflection for both cases. It's interesting to see that angles that the tester put the glass: either perpendicular (which does affects the co-axial design but not used in practice) and tilting downward (which reduces reflection for both cases). It'd be fun to see if the glass is tilted to reflect light toward the receiver lens :)
The disturbing thing for me is the last scene about the LiDAR seeing the "23" number on the ground while the driver can't. It's simply because the LiDAR is installed at a higher vantage point than the driver. It's kind of an insult.