Depth of field recognition: Computer vision is being taught to see in 3D

• Author:
• Author name

  Quantumrun Foresight
• December 28, 2022

Insight summary

Machines that can accurately identify objects and environments are becoming crucial as the world adopts the use of ever more autonomous devices. However, developments in depth of field (DOF) vision technologies can also be used for controversial surveillance purposes. The long-term implications of DOF could include reduced vehicle accidents and robots having better navigation skills.

Depth of field recognition context

Depth of field refers to the distance in front and beyond the subject. Computer vision (CV) algorithms are being trained to understand the nuances in DOF and to differentiate objects from their background. This feature is what makes CV versatile as it can be developed in various ways, from image recognition to surveillance to creating safer autonomous vehicles (AVs). 

A device that captures the advancements in DOF recognition is the 3D camera. These cameras allow the viewer to see depth in images, replicating three dimensions through human binocular vision. Some 3D cameras use multiple lenses to record different points of view, while others just have one lens that shifts positions. The combination of both perspectives creates depth perception.

These devices were initially designed for content creators to produce high-quality material but have been increasingly utilized as surveillance systems. When the LiDAR (Light Detection and Ranging, uses pulsed laser light to measure distances to objects) sensor is combined with 3D cameras, it produces wireframes that enable facial recognition, even when a face is hidden by a mask or glasses. This technology also improves object recognition by using deep learning to understand how certain features differ in various environments and change over time.

Disruptive impact

According to intelligence firm Technavio, the 3D sensor market is expected to grow 22 percent annually until 2024. Global facial recognition software was forecast to make over USD $12 billion by 2025. According to 3D vision tech firm Orbbec, since the COVID-19 pandemic, 3D cameras have become the preferred method for contactless recognition and identification.

Several features allow the combination of LiDAR and 3D cameras to be good at object recognition. The LiDAR sensor is useful for detecting objects from dozens to hundreds of meters away but cannot identify close-up objects at only a few meters. For example, in AVs, LiDAR would not be able to tell if the object in question is something that should be avoided or not.

Depth vision systems can complement LiDAR to provide a more complete picture. Depth cameras are usually equipped with two RGBD (color-depth) cameras that offer stereovision. This capability allows the camera to perceive the distance, position, and speed of nearby objects. The added textures from the RGB (color) camera provide more detail.

Deep learning technology can teach 3D systems to sense and recognize objects with high fidelity. Although LiDAR enables navigation, depth cameras can ensure obstacle perception and identification. It can distinguish between objects like motorcycles and deer, or pedestrians and people on scooters or skateboards. It can even tell the difference between animals like dogs, raccoons, or opossums.

Implications of depth of field recognition

Wider implications of depth of field recognition may include: 

• Transportation safety standards eventually mandating DOF cameras be installed in all new road vehicles to improve driver safety and reduce road incidents.


• Industrial and office and home robots potentially using DOF cameras to more easily navigate their surroundings, especially if they are designed to function autonomously.


• Law enforcement agencies using 3D cameras to identify masked terrorists and other criminals. However, the tech can also be used to suppress protests and activists.


• Companies using DOF cameras to accurately identify/verify employees, particularly in countries where masks against COVID-19 are still mandated.


• DOF cameras being used to provide more contactless and remote services even post-pandemic.


• The military deploying DOF recognition to better identify enemy positions.


• Some pro-privacy companies and individuals focusing on developing equipment or devices that can counter DOF cameras.

Questions to consider

• What are other potential use cases for 3D cameras?


• What are some ways that DOF recognition can be used to improve services?