SlideShare a Scribd company logo
© 2021 Ambarella Inc.
When 2D Is Not Enough: An
Overview of Optical Depth
Sensing Technologies
Dinesh Balasubramaniam
Ambarella
© 2021 Ambarella Inc. 2
Ambarella
Founded on the premise that video is a unique type of data requiring
an optimized SoC architecture
• DNN AI processors integrated with state-of-the-art video processors
yield highly optimized computer vision SoCs
Viewing and sensing SoCs with leading performance and efficiency
• Low-power and high-performance SoCs for IoT and Automotive
camera markets
2004
Ambarella founded
2012
IPO NASDAQ: AMBA
2021
280+ million
HD/4K cameras enabled
2015
Acquisition of
Santa Clara, CA
Company HQ
Tokyo
Hsinchu
Total Headcount
Detroit Parma
VisLab
Shanghai Seoul
Munich
785
Shenzhen
World-class engineering team
• >600 software, algorithm, VLSI and
image processing engineers globally
AI Vision Processors For Edge Applications
© 2021 Ambarella Inc.
Outline
• Why Depth?
• Depth Sensing Modalities
• Stereo Camera, Time of Flight, and Structured Light
• Operating Principles
• Design considerations: Resolution, Range, Precision, Bright light and Low-light
performance
• Review and comparison
• Applications
3
© 2021 Ambarella Inc.
Why Depth?
4
© 2021 Ambarella Inc.
Why Depth?
5
Which face is real? What is in the way?
Are all objects flat?
Modalities
© 2021 Ambarella Inc.
Depth Sensing Modalities
7
Sonar
Time of Flight module
Stereo camera
Radar Lidar
Structured Light module
Stereo Camera System
© 2021 Ambarella Inc.
Operating Principle – Stereo Camera System
= disparity
Right Image
Left Image
© 2021 Ambarella Inc.
Operating Principle – Stereo Camera System
B
f
dl
dr
𝐵
𝐷
=
𝐵 − 𝑑𝑙 − 𝑑𝑟
𝐷 − 𝑓
𝐷 = 𝑓
𝐵
𝑑𝑙 − 𝑑𝑟
B – baseline (distance between optical center)
D – distance to be measured
f – focal length of lens
d – distance of projected point from optical axis
• Distance inversely proportional to disparity
• Epipolar geometry helps to reduce search
space for matching point
• Finding matching point is called
correspondence problem
D
© 2021 Ambarella Inc.
Design Considerations – Stereo Camera Systems
• Systems are flexible and scalable – can accommodate many configurations with
appropriate change in sensor resolution, field of view (FoV), and baseline.
• Depth precision decreases with distance – non-linear error model.
• Good performance in bright light conditions. Low-light performance depends on
output of image pipeline.
• Challenged on scenes with low texture – improved with active illumination.
• Manufacturing process can be complicated – alignment of optical units is critical
during and after manufacture. Improved with periodic calibration or autocalibration
software.
11
© 2021 Ambarella Inc.
Design Consideration – Stereo Camera Systems
12
Stereo Left Stereo Right
© 2021 Ambarella Inc.
Performance – Stereo Camera Systems
13
Stereo Left Stereo Right
© 2021 Ambarella Inc.
Design Considerations – Stereo Camera Systems
14
Time of Flight
© 2021 Ambarella Inc.
Operating Principle – Time of Flight
• Two types of ToF sensors: direct ToF and indirect ToF.
• Direct Tof sensors emit a pulse of light and measure the time it takes for the reflection
to arrive at the sensor. Pulse strength and duration can be varied to allow long range
measurements (up to kms).
• Indirect ToFs also called continuous-wave ToFs emit a pulsed, modulated light signal
and measure the phase difference of the reflected signal to derive depth.
© 2021 Ambarella Inc.
Operating Principle – Time of Flight
Phase
Depth
Integration time Read out time
∅ = 90
∅ = 0 ∅ = 180 ∅ = 270
© 2021 Ambarella Inc.
Operating Principle – Time of Flight
Depth Amplitude
© 2021 Ambarella Inc.
• Depth range depends on modulating frequency of the IR source - maximum
measurable distance for given frequency called unambiguous range.
• LEDs can be used up to 30MHz. VCSELs (vertical-cavity surface-emitting laser) can be
used if higher modulating frequencies are required (up to 100MHz).
Design Consideration– Time of Flight
Modulating Frequency
(MHz)
Unambiguous Range
(m)
20 7.5
50 3
75 2
100 1.5
© 2021 Ambarella Inc.
Design Considerations – Time of Flight
• Comparatively low resolution (VGA current max) but unaffected by scene texture
• Dense depth information where field of view and field of illumination overlap
• Range limited by modulating frequency of illuminator. Depth precision decreases with
distance for a given modulating frequency.
• Good performance in low-light conditions. Sunlight affects bright light performance –
improved when using 940 nm IR illuminator.
• Scattered reflections and cross talk can impact depth measurement.
• Integration time impacts number of fps but can handle motion.
• Combining multiple systems can be challenging – can be separated by frequency.
20
© 2021 Ambarella Inc.
Design Considerations – Time of Flight
21
* Data from vendor module
Structured Light
© 2021 Ambarella Inc.
Operating Principle – Structured Light
• Light source plus a pattern generator
project a random pattern.
• When image plane moves pattern
shifts depending on object distance.
• Reflected pattern correlated against a
reference stored on the system.
• Depth depends on disparity between
the reference and reflected pattern.
© 2021 Ambarella Inc.
Operating Principle – Structured Light
B
D
f
R
R1 R2
p1 p2
𝑓
𝑝2 − 𝑝1
=
𝑅
𝑅2 − 𝑅1
𝐷
𝐵
=
𝑅 − 𝐷
𝑅2 − 𝑅1
𝐷 =
𝑅𝐵𝑓
𝑅 𝑝2 − 𝑝1 + 𝐵𝑓
B – baseline (distance between optical center)
f – focal length of lens
R – distance to reference plane
D – distance to be measured
𝑅𝑛 – Point on reference plane
𝑝𝑛 – projected points on image plane
© 2021 Ambarella Inc.
System Considerations – Structured Light
• Moderately flexible – sensor resolution, baseline, and pattern density can be varied
for different applications.
• Depth precision decreases with distance.
• Good performance in low-light conditions. Lower performance when sunlight is
present – improved when using 940 nm IR illuminator.
• Spatial resolution higher than time of flight sensors but depth density depends on
projector pattern.
• Frame rate depends on sensor and depth processing speed.
25
© 2021 Ambarella Inc.
System Considerations – Structured Light
26
Comparison
© 2021 Ambarella Inc.
Comparison
Stereo Time of Flight Structured Light
Depth Range Short to long range Low to mid range Low to mid range
Depth Precision mm to cm mm to cm mm to cm
Depth Resolution
High – can detect finer features or
objects further away
Up to VGA Pattern dependent
Active emitter Optional Required Required
Bright light performance Good Fair – improved with 940 nm IR Fair – improved with 940 nm IR
Low-light performance Depends on ISP Good Good
Cost High Low Mid
Applications
© 2021 Ambarella Inc.
Applications
30
Long-range stereo for
autonomous driving
ToF for occupancy sensing Structured light for access control
Thank You

More Related Content

PDF
“Alternative Image Sensors for Intelligent In-Cabin Monitoring, Home Security...
PDF
"The Evolution of Depth Sensing: From Exotic to Ubiquitous," a Presentation f...
PDF
"The Coming Shift from Image Sensors to Image Sensing," a Presentation from LG
PDF
"Approaches for Vision-based Driver Monitoring," a Presentation from PathPart...
PDF
"Digital Gimbal: Rock-steady Video Stabilization without Extra Weight!," a Pr...
PDF
"Using Deep Learning for Video Event Detection on a Compute Budget," a Presen...
PDF
“Challenges in Vision-based Adaptive Traffic Control Systems,” a Presentation...
PDF
“Building an Autonomous Detect-and-Avoid System for Commercial Drones,” a Pre...
“Alternative Image Sensors for Intelligent In-Cabin Monitoring, Home Security...
"The Evolution of Depth Sensing: From Exotic to Ubiquitous," a Presentation f...
"The Coming Shift from Image Sensors to Image Sensing," a Presentation from LG
"Approaches for Vision-based Driver Monitoring," a Presentation from PathPart...
"Digital Gimbal: Rock-steady Video Stabilization without Extra Weight!," a Pr...
"Using Deep Learning for Video Event Detection on a Compute Budget," a Presen...
“Challenges in Vision-based Adaptive Traffic Control Systems,” a Presentation...
“Building an Autonomous Detect-and-Avoid System for Commercial Drones,” a Pre...

What's hot (20)

PPT
Digital imaging
PDF
HiSLED-PK
PPTX
DIGITAL IMAGING
PPTX
Mitchell Reifel (pmdtechnologies ag): pmd Time-of-Flight – the Swiss Army Kni...
PDF
Poster_ARVO_ADCIS_AGN-V6a
PPTX
8K Extremely High Resolution Camera System
PPTX
A wireless robot for networked laparoscopy
PDF
"The Path from ADAS to Autonomy," a Presentation from Strategy Analytics
PPTX
Indoor positioning system
PDF
4th Year Project - Indoor Positioning System
PPTX
Talal's Poster Presentation
PDF
The Next-Gen Technologies Driving Immersion
PDF
Innovate in new and exciting optical sensing applications in industrial marke...
PPTX
Night vision technology applications
PDF
Real Time Location Systems (RTLS)
PDF
8 k extremely high resolution camera system
PDF
Jovision jvs-n935-hy Bangladesh (01785-777722)
PPTX
Ips system
PPTX
Computer radiography and digital radiography
PDF
Mri d stream brochure_final_16 nov 12
Digital imaging
HiSLED-PK
DIGITAL IMAGING
Mitchell Reifel (pmdtechnologies ag): pmd Time-of-Flight – the Swiss Army Kni...
Poster_ARVO_ADCIS_AGN-V6a
8K Extremely High Resolution Camera System
A wireless robot for networked laparoscopy
"The Path from ADAS to Autonomy," a Presentation from Strategy Analytics
Indoor positioning system
4th Year Project - Indoor Positioning System
Talal's Poster Presentation
The Next-Gen Technologies Driving Immersion
Innovate in new and exciting optical sensing applications in industrial marke...
Night vision technology applications
Real Time Location Systems (RTLS)
8 k extremely high resolution camera system
Jovision jvs-n935-hy Bangladesh (01785-777722)
Ips system
Computer radiography and digital radiography
Mri d stream brochure_final_16 nov 12
Ad

Similar to “When 2D Is Not Enough: An Overview of Optical Depth Sensing Technologies,” a Presentation from Ambarella (20)

PDF
“Introduction to Depth Sensing,” a Presentation from Meta
PDF
ASIS Poster - Final
PPT
3D Scanning technology of industrial .ppt
PDF
Time of Flight and Structured Light Depth Cameras Technology and Applications...
PDF
Goal location prediction based on deep learning using RGB-D camera
PDF
"From 2D to 3D: How Depth Sensing Will Shape the Future of Vision," a Present...
PDF
Distance Estimation to Image Objects Using Adapted Scale
PDF
Distance Estimation to Image Objects Using Adapted Scale
PDF
ADAPTIVE FILTER FOR DENOISING 3D DATA CAPTURED BY DEPTH SENSORS
PDF
A Fast Single-Pixel Laser Imager for VR/AR Headset Tracking
PDF
"How to Choose a 3D Vision Sensor," a Presentation from Capable Robot Components
PDF
An Accurate Scheme for Distance Measurement using an Ordinary Webcam
PDF
An Accurate Scheme for Distance Measurement using an Ordinary Webcam
PDF
An Assessment of Image Matching Algorithms in Depth Estimation
PDF
6 - Conception of an Autonomous UAV using Stereo Vision (presented in an Indo...
PPTX
MEMS Laser Scanning, the platform for next generation of 3D Depth Sensors
PDF
3d machine-vision-systems-paper-presentation
PDF
Dense Image Matching - Challenges and Potentials (Keynote 3D-ARCH 2015)
PDF
Zfx ca csm1405
PDF
"Introduction to Optics for Embedded Vision," a Presentation from Edmund Optics
“Introduction to Depth Sensing,” a Presentation from Meta
ASIS Poster - Final
3D Scanning technology of industrial .ppt
Time of Flight and Structured Light Depth Cameras Technology and Applications...
Goal location prediction based on deep learning using RGB-D camera
"From 2D to 3D: How Depth Sensing Will Shape the Future of Vision," a Present...
Distance Estimation to Image Objects Using Adapted Scale
Distance Estimation to Image Objects Using Adapted Scale
ADAPTIVE FILTER FOR DENOISING 3D DATA CAPTURED BY DEPTH SENSORS
A Fast Single-Pixel Laser Imager for VR/AR Headset Tracking
"How to Choose a 3D Vision Sensor," a Presentation from Capable Robot Components
An Accurate Scheme for Distance Measurement using an Ordinary Webcam
An Accurate Scheme for Distance Measurement using an Ordinary Webcam
An Assessment of Image Matching Algorithms in Depth Estimation
6 - Conception of an Autonomous UAV using Stereo Vision (presented in an Indo...
MEMS Laser Scanning, the platform for next generation of 3D Depth Sensors
3d machine-vision-systems-paper-presentation
Dense Image Matching - Challenges and Potentials (Keynote 3D-ARCH 2015)
Zfx ca csm1405
"Introduction to Optics for Embedded Vision," a Presentation from Edmund Optics
Ad

More from Edge AI and Vision Alliance (20)

PDF
“An Introduction to the MIPI CSI-2 Image Sensor Standard and Its Latest Advan...
PDF
“Visual Search: Fine-grained Recognition with Embedding Models for the Edge,”...
PDF
“Optimizing Real-time SLAM Performance for Autonomous Robots with GPU Acceler...
PDF
“LLMs and VLMs for Regulatory Compliance, Quality Control and Safety Applicat...
PDF
“Simplifying Portable Computer Vision with OpenVX 2.0,” a Presentation from AMD
PDF
“Quantization Techniques for Efficient Deployment of Large Language Models: A...
PDF
“Introduction to Data Types for AI: Trade-offs and Trends,” a Presentation fr...
PDF
“Introduction to Radar and Its Use for Machine Perception,” a Presentation fr...
PDF
“NPU IP Hardware Shaped Through Software and Use-case Analysis,” a Presentati...
PDF
“Voice Interfaces on a Budget: Building Real-time Speech Recognition on Low-c...
PDF
“Computer Vision at Sea: Automated Fish Tracking for Sustainable Fishing,” a ...
PDF
“Squinting Vision Pipelines: Detecting and Correcting Errors in Vision Models...
PDF
“ONNX and Python to C++: State-of-the-art Graph Compilation,” a Presentation ...
PDF
“Beyond the Demo: Turning Computer Vision Prototypes into Scalable, Cost-effe...
PDF
“Running Accelerated CNNs on Low-power Microcontrollers Using Arm Ethos-U55, ...
PDF
“Scaling i.MX Applications Processors’ Native Edge AI with Discrete AI Accele...
PDF
“A Re-imagination of Embedded Vision System Design,” a Presentation from Imag...
PDF
“MPU+: A Transformative Solution for Next-Gen AI at the Edge,” a Presentation...
PDF
“Evolving Inference Processor Software Stacks to Support LLMs,” a Presentatio...
PDF
“Efficiently Registering Depth and RGB Images,” a Presentation from eInfochips
“An Introduction to the MIPI CSI-2 Image Sensor Standard and Its Latest Advan...
“Visual Search: Fine-grained Recognition with Embedding Models for the Edge,”...
“Optimizing Real-time SLAM Performance for Autonomous Robots with GPU Acceler...
“LLMs and VLMs for Regulatory Compliance, Quality Control and Safety Applicat...
“Simplifying Portable Computer Vision with OpenVX 2.0,” a Presentation from AMD
“Quantization Techniques for Efficient Deployment of Large Language Models: A...
“Introduction to Data Types for AI: Trade-offs and Trends,” a Presentation fr...
“Introduction to Radar and Its Use for Machine Perception,” a Presentation fr...
“NPU IP Hardware Shaped Through Software and Use-case Analysis,” a Presentati...
“Voice Interfaces on a Budget: Building Real-time Speech Recognition on Low-c...
“Computer Vision at Sea: Automated Fish Tracking for Sustainable Fishing,” a ...
“Squinting Vision Pipelines: Detecting and Correcting Errors in Vision Models...
“ONNX and Python to C++: State-of-the-art Graph Compilation,” a Presentation ...
“Beyond the Demo: Turning Computer Vision Prototypes into Scalable, Cost-effe...
“Running Accelerated CNNs on Low-power Microcontrollers Using Arm Ethos-U55, ...
“Scaling i.MX Applications Processors’ Native Edge AI with Discrete AI Accele...
“A Re-imagination of Embedded Vision System Design,” a Presentation from Imag...
“MPU+: A Transformative Solution for Next-Gen AI at the Edge,” a Presentation...
“Evolving Inference Processor Software Stacks to Support LLMs,” a Presentatio...
“Efficiently Registering Depth and RGB Images,” a Presentation from eInfochips

Recently uploaded (20)

PDF
Reach Out and Touch Someone: Haptics and Empathic Computing
PDF
cuic standard and advanced reporting.pdf
PDF
Dropbox Q2 2025 Financial Results & Investor Presentation
PDF
Accuracy of neural networks in brain wave diagnosis of schizophrenia
PPTX
MYSQL Presentation for SQL database connectivity
PDF
Build a system with the filesystem maintained by OSTree @ COSCUP 2025
PDF
Blue Purple Modern Animated Computer Science Presentation.pdf.pdf
PDF
Assigned Numbers - 2025 - Bluetooth® Document
PDF
Building Integrated photovoltaic BIPV_UPV.pdf
PDF
Mobile App Security Testing_ A Comprehensive Guide.pdf
PDF
Network Security Unit 5.pdf for BCA BBA.
PDF
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
PPTX
Tartificialntelligence_presentation.pptx
PDF
Per capita expenditure prediction using model stacking based on satellite ima...
PDF
Approach and Philosophy of On baking technology
PDF
A comparative analysis of optical character recognition models for extracting...
PDF
Getting Started with Data Integration: FME Form 101
PDF
Encapsulation_ Review paper, used for researhc scholars
PDF
NewMind AI Weekly Chronicles - August'25-Week II
PDF
Agricultural_Statistics_at_a_Glance_2022_0.pdf
Reach Out and Touch Someone: Haptics and Empathic Computing
cuic standard and advanced reporting.pdf
Dropbox Q2 2025 Financial Results & Investor Presentation
Accuracy of neural networks in brain wave diagnosis of schizophrenia
MYSQL Presentation for SQL database connectivity
Build a system with the filesystem maintained by OSTree @ COSCUP 2025
Blue Purple Modern Animated Computer Science Presentation.pdf.pdf
Assigned Numbers - 2025 - Bluetooth® Document
Building Integrated photovoltaic BIPV_UPV.pdf
Mobile App Security Testing_ A Comprehensive Guide.pdf
Network Security Unit 5.pdf for BCA BBA.
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
Tartificialntelligence_presentation.pptx
Per capita expenditure prediction using model stacking based on satellite ima...
Approach and Philosophy of On baking technology
A comparative analysis of optical character recognition models for extracting...
Getting Started with Data Integration: FME Form 101
Encapsulation_ Review paper, used for researhc scholars
NewMind AI Weekly Chronicles - August'25-Week II
Agricultural_Statistics_at_a_Glance_2022_0.pdf

“When 2D Is Not Enough: An Overview of Optical Depth Sensing Technologies,” a Presentation from Ambarella

  • 1. © 2021 Ambarella Inc. When 2D Is Not Enough: An Overview of Optical Depth Sensing Technologies Dinesh Balasubramaniam Ambarella
  • 2. © 2021 Ambarella Inc. 2 Ambarella Founded on the premise that video is a unique type of data requiring an optimized SoC architecture • DNN AI processors integrated with state-of-the-art video processors yield highly optimized computer vision SoCs Viewing and sensing SoCs with leading performance and efficiency • Low-power and high-performance SoCs for IoT and Automotive camera markets 2004 Ambarella founded 2012 IPO NASDAQ: AMBA 2021 280+ million HD/4K cameras enabled 2015 Acquisition of Santa Clara, CA Company HQ Tokyo Hsinchu Total Headcount Detroit Parma VisLab Shanghai Seoul Munich 785 Shenzhen World-class engineering team • >600 software, algorithm, VLSI and image processing engineers globally AI Vision Processors For Edge Applications
  • 3. © 2021 Ambarella Inc. Outline • Why Depth? • Depth Sensing Modalities • Stereo Camera, Time of Flight, and Structured Light • Operating Principles • Design considerations: Resolution, Range, Precision, Bright light and Low-light performance • Review and comparison • Applications 3
  • 4. © 2021 Ambarella Inc. Why Depth? 4
  • 5. © 2021 Ambarella Inc. Why Depth? 5 Which face is real? What is in the way? Are all objects flat?
  • 7. © 2021 Ambarella Inc. Depth Sensing Modalities 7 Sonar Time of Flight module Stereo camera Radar Lidar Structured Light module
  • 9. © 2021 Ambarella Inc. Operating Principle – Stereo Camera System = disparity Right Image Left Image
  • 10. © 2021 Ambarella Inc. Operating Principle – Stereo Camera System B f dl dr 𝐵 𝐷 = 𝐵 − 𝑑𝑙 − 𝑑𝑟 𝐷 − 𝑓 𝐷 = 𝑓 𝐵 𝑑𝑙 − 𝑑𝑟 B – baseline (distance between optical center) D – distance to be measured f – focal length of lens d – distance of projected point from optical axis • Distance inversely proportional to disparity • Epipolar geometry helps to reduce search space for matching point • Finding matching point is called correspondence problem D
  • 11. © 2021 Ambarella Inc. Design Considerations – Stereo Camera Systems • Systems are flexible and scalable – can accommodate many configurations with appropriate change in sensor resolution, field of view (FoV), and baseline. • Depth precision decreases with distance – non-linear error model. • Good performance in bright light conditions. Low-light performance depends on output of image pipeline. • Challenged on scenes with low texture – improved with active illumination. • Manufacturing process can be complicated – alignment of optical units is critical during and after manufacture. Improved with periodic calibration or autocalibration software. 11
  • 12. © 2021 Ambarella Inc. Design Consideration – Stereo Camera Systems 12 Stereo Left Stereo Right
  • 13. © 2021 Ambarella Inc. Performance – Stereo Camera Systems 13 Stereo Left Stereo Right
  • 14. © 2021 Ambarella Inc. Design Considerations – Stereo Camera Systems 14
  • 16. © 2021 Ambarella Inc. Operating Principle – Time of Flight • Two types of ToF sensors: direct ToF and indirect ToF. • Direct Tof sensors emit a pulse of light and measure the time it takes for the reflection to arrive at the sensor. Pulse strength and duration can be varied to allow long range measurements (up to kms). • Indirect ToFs also called continuous-wave ToFs emit a pulsed, modulated light signal and measure the phase difference of the reflected signal to derive depth.
  • 17. © 2021 Ambarella Inc. Operating Principle – Time of Flight Phase Depth Integration time Read out time ∅ = 90 ∅ = 0 ∅ = 180 ∅ = 270
  • 18. © 2021 Ambarella Inc. Operating Principle – Time of Flight Depth Amplitude
  • 19. © 2021 Ambarella Inc. • Depth range depends on modulating frequency of the IR source - maximum measurable distance for given frequency called unambiguous range. • LEDs can be used up to 30MHz. VCSELs (vertical-cavity surface-emitting laser) can be used if higher modulating frequencies are required (up to 100MHz). Design Consideration– Time of Flight Modulating Frequency (MHz) Unambiguous Range (m) 20 7.5 50 3 75 2 100 1.5
  • 20. © 2021 Ambarella Inc. Design Considerations – Time of Flight • Comparatively low resolution (VGA current max) but unaffected by scene texture • Dense depth information where field of view and field of illumination overlap • Range limited by modulating frequency of illuminator. Depth precision decreases with distance for a given modulating frequency. • Good performance in low-light conditions. Sunlight affects bright light performance – improved when using 940 nm IR illuminator. • Scattered reflections and cross talk can impact depth measurement. • Integration time impacts number of fps but can handle motion. • Combining multiple systems can be challenging – can be separated by frequency. 20
  • 21. © 2021 Ambarella Inc. Design Considerations – Time of Flight 21 * Data from vendor module
  • 23. © 2021 Ambarella Inc. Operating Principle – Structured Light • Light source plus a pattern generator project a random pattern. • When image plane moves pattern shifts depending on object distance. • Reflected pattern correlated against a reference stored on the system. • Depth depends on disparity between the reference and reflected pattern.
  • 24. © 2021 Ambarella Inc. Operating Principle – Structured Light B D f R R1 R2 p1 p2 𝑓 𝑝2 − 𝑝1 = 𝑅 𝑅2 − 𝑅1 𝐷 𝐵 = 𝑅 − 𝐷 𝑅2 − 𝑅1 𝐷 = 𝑅𝐵𝑓 𝑅 𝑝2 − 𝑝1 + 𝐵𝑓 B – baseline (distance between optical center) f – focal length of lens R – distance to reference plane D – distance to be measured 𝑅𝑛 – Point on reference plane 𝑝𝑛 – projected points on image plane
  • 25. © 2021 Ambarella Inc. System Considerations – Structured Light • Moderately flexible – sensor resolution, baseline, and pattern density can be varied for different applications. • Depth precision decreases with distance. • Good performance in low-light conditions. Lower performance when sunlight is present – improved when using 940 nm IR illuminator. • Spatial resolution higher than time of flight sensors but depth density depends on projector pattern. • Frame rate depends on sensor and depth processing speed. 25
  • 26. © 2021 Ambarella Inc. System Considerations – Structured Light 26
  • 28. © 2021 Ambarella Inc. Comparison Stereo Time of Flight Structured Light Depth Range Short to long range Low to mid range Low to mid range Depth Precision mm to cm mm to cm mm to cm Depth Resolution High – can detect finer features or objects further away Up to VGA Pattern dependent Active emitter Optional Required Required Bright light performance Good Fair – improved with 940 nm IR Fair – improved with 940 nm IR Low-light performance Depends on ISP Good Good Cost High Low Mid
  • 30. © 2021 Ambarella Inc. Applications 30 Long-range stereo for autonomous driving ToF for occupancy sensing Structured light for access control