OAK-D LR¶
Overview¶
The OAK-D Long Range (OAK-D LR) was designed to provide an accurate short range and long range stereo depth perception. This works by having up to 3 stereo depth pairs at different baseline distances. One can also easily change the M12 lenses for the cameras, which also affects the min/max distance of stereo depth perception (additional info).
It has 3 horizontally aligned AR0234 global shutter color cameras, which have different distances between each other, and allow for 3 different stereo baseline camera pairs (5cm, 10cm and 15cm baseline distances).
The OAK-D LR leverages our OAK-SoM-Pro to make a overall compact design. The use of the SoM reduces the design’s scale, making it easier to mount or fit in various robotic processes. The design is also open-source, allowing for any necessary modifications.
Hardware specifications¶
For communication and power, the OAK-D LR camera uses either:
USB-C cable - it supports both USB2 and USB3 (5Gbps / 10Gbps).
Power-over-Ethernet (PoE) - it offers full 802.3af and Class 3 PoE compliance with 1000BASE-T speeds (1 Gbps). A PoE injector/switch is required to power the device.
Camera Specs |
Stereo pair / Color |
|---|---|
Sensor |
AR0234 (color, PY078) |
DFOV / HFOV / VFOV |
100° / 82° / 56° |
Resolution |
2.3MP (1920x1200) |
Focus |
M12 (FF): 45cm - ∞ |
Lens size |
1/2.5 inch |
Max Framerate |
60 FPS (1200P) |
Pixel size |
3µm x 3µm |
Stereo depth perception¶
This OAK camera has a baseline of 5 cm, 10 cm, and 15 cm. Baseline distance is the distance between the left and the right stereo camera. Since we have 3 camera sensors on the OAK-D-LR, we also have 3 baseline distances, and 3 potential stereo pairs. We measured the accuracy of the 15 cm stereo pair (the left-most and right-most camera).
Depth range: 20cm - 16m
-
20cm - 6m: below 2% absolute depth error
6m - 8m: below 3% absolute depth error
8m-12m: below 4% absolute depth error
Theoretical depth accuracy¶
As M12 lenses are swappable, user can easily change lenses to achieve wider FOV, or longer depth perception (narrower FOV):
HFOV [°] |
< 2% depth error |
< 4% depth error |
< 8% depth error |
MinZ |
|---|---|---|---|---|
10 |
54.9 m |
137.2 m |
274.3 m |
1.92 m |
20 |
27.2 m |
68.1 m |
136.1 m |
95 cm |
30 |
17.9 m |
44.8 m |
89.6 m |
63 cm |
40 |
13.2 m |
33.0 m |
65.9 m |
46 cm |
50 |
10.3 m |
25.7 m |
51.5 m |
36 cm |
60 |
8.3 m |
20.8 m |
41.6 m |
29 cm |
70 |
6.9 m |
17.1 m |
34.3 m |
24 cm |
80 |
5.7 m |
14.3 m |
28.6 m |
21 cm |
82 |
5.5 m |
13.8 m |
27.6 m |
20 cm |
90 |
4.8 m |
12.0 m |
24.0 m |
17 cm |
100 |
4.0 m |
10.1 m |
20.1 m |
14 cm |
Note: we haven’t tested all of these combinations, but calculated theoretical depth error and interpolated those values with our previous real-world tests when enabling subpixel disparity:
< 2% error- 20th disparity pixel, which has 5% full-pixel error (~2% with subpixel enabled)< 4% error- 8th disparity pixel, which has 12.5% full-pixel error (~4% with subpixel enabled)< 8% error- 4th disparity pixel, which has 25% full-pixel error (~10% with subpixel enabled)
Maximum depth was calculated by using the large (15cm) baseline, while MinZ was calculated by using the small (5cm) baseline of the OAK-D-LR. You can further decrease MinZ by lowering the resolution, or using disparity shift (docs here). MinZ values are already using Extended Disparity Mode.
Lens clearance¶
There is 11.6mm of clearance between the M12 sensor housing and the cover lens. If you are using longer lenses (longer focal length which means narrower FOV), you might not be able to mount back the front cover.
RVC2 inside¶
This OAK device is built on top of the RVC2. Main features:
4 TOPS of processing power (1.4 TOPS for AI - RVC2 NN Performance)
Run any AI model, even custom-architectured/built ones - models need to be converted.
Encoding: H.264, H.265, MJPEG
Computer vision: warp/dewarp, resize, crop via ImageManip node, edge detection, feature tracking. You can also run custom CV functions
Stereo depth perception with filtering, post-processing, RGB-depth alignment, and high configurability
Object tracking: 2D and 3D tracking with ObjectTracker node
Integrated IMU¶
This OAK camera has an integrated BNO085, a 9-axis IMU (Inertial Measurement Unit). See IMU node for the API details on how to use it.
Note: due to supply chain issues, most of the OAK camera that were manufactured between Q2 2021 and Q2 2023 have integrated BMI270 - 6-axis IMU instead.
Power consumption¶
Most of the power is consumed by the RVC2, so the power consumption mostly depends on the workload of the chip:
Base consumption + PoE circuitry + camera streaming: 3W - 3.5W
AI subsystem consumption: Up to 1W
Stereo depth pipeline subsystem: Up to 0.5W
Video Encoder subsystem: Up to 0.5W
So the total power consumption can be up to ~5.5W if you are using all the features at 100% at the same time. To reduce the power consumption, you can reduce FPS of the whole pipeline - that way, subsystems won’t be utilized at 100% and will consume less power.
Operating temperature¶
The ambient operating temperature of RVC2 based devices is between -20°C and 50°C when fully utilizing the VPU.
Similarly to the Power consumption, max operating temperature depends on VPU utilization. The higher the VPU utilization, the more heat the VPU will generate. The RVC2 VPU can continuously operate at 105 °C, after which the depthai library will automatically shut down the device (to avoid chip damage).
To find out more, see our Operative temperature range documentation.
Changing lenses¶
As OAK-D-LR’s camera modules use M12 lenses, users can replace them without any special tools. By default, lenses are glued to the camera module, but users can remove the glue and replace the M12 lens with a different one. ArduCam’s M12 lens kit can be quite useful to test a variety of lenses to choose from.