Stereo Vision Systems

Comparison table with similar products

3D Vision System	Elphel MNC393-XCAM	Reviewed competing products
Released	2018	2015-2018
Depth sensing technology		Active IR and projected pattern variations or SLAM variations
Depth range, m	10-5000m	0.1-20m
Depth accuracy	Disparity accuracy is 0.05px: 1% at 100m 5% at 500m 10% at 1000m 20% at 2000m	Some do not have information The assumption is nobody can get below 0.5px disparity Passive systems (no IR), best: 1.5% at 10m 17.5% at 100m Active systems, best (ZiVID): 0.1mm @ 0.6m but the range is 0.6-1.1m HTC VIVE: 5% @ 2m (info by StereoLabs) ZED Mini by StereoLabs: 1% at 1m 9% at 10m
Depth Processing	Post-processing (real-time - future development)	Post-processing or real-time (mostly)
Number of image sensors	4 (square)	1-2
Stereo baseline, mm	250 (diagonal)	from single lens to 63-250
IR projectors	not used	some have, some don't
Framerate	up to 10fps recording in high quality	ZED Mini: varies from 15fps @4k to 100fps @1344x376 Others are somewhat similar
Output interfaces	1 GigE USB2 eSATA (external storage) internal storage (m.2 SSD) Possible to precisely sync multiple systems	some have 1 GigE most - USB3
Position and orientation	GPS Compatible with Analog Devices IMUs: up to 2460Hz	some have IMU (100-500Hz), but most don't
Image sensor shutter type	ERS (Electronic Rolling Shutter)	ERS or Global
Output format	3D model: OBJ with textures X3D with textures	RGBD Unknown Z-map
Color	Color	Color (most) Mono 2 Mono + 1 Color
Full Hardware setup	Some laptop when recording (network GUI - only browser, no extra drivers) Good PC (+NVidia GPU) when processing	Some require PC with extra drivers or other software Some require PC with extra drivers or other software + good NVidia GPU (like ZED)

Advantages over LIDARs

Product	Elphel MNC393-XCAM	Reviewed LIDARs (Velodyne, Ouster)
Range	10-5000m	<200m
Field of View	65°x51°	360°x(30-40°)
Points sparsity (horizontal x vertical):
@10m	~ 0.05x0.05 cm	~ 2x6 cm
@50m	~ 0.25x0.25 cm	~ 10x30 cm
@100m	~ 0.5x0.5 cm	~ 20x60 cm
@200m	~ 1x1 cm	~ 40x120 cm
@1000m	~ 5x5 cm
@2000m	~ 10x10 cm

Typical LIDAR angular resolutions are:

• horizontal: 0.1-0.4°
• vertical: 0.3-0.5°

Links

• StereoVisionCameras_v2.ods (internal use only) (1st tab - cameras, 2nd tab - LIDARs)
• More on cameras (xinreality.com)
• More on LIDARs (dronezon.com)
• Time-of-Flight sensors (2014 TI article)