3D Imaging—the Heart of Machine Vision—Forges Ahead
Behind the explosion in computer vision is tremendous growth in the world of 3D imaging. What’s the state of 3D imaging today and where are we headed?
One of the most useful and well-studied subsects of AI/ML is machine vision. Machine vision, or the ability for a computer to analyze and predict the contents of an image, is often only as good as the images it’s fed.
Example of a machine vision application. Image from Robotics Tomorrow
This is why engineers have been working tirelessly to create new, better means of 3D imaging technology from both a hardware and software perspective.
What Is 3D Imaging?
When studying imaging, engineers have, understandably, found inspiration from the way the human eye captures and interprets images.
Working principle of a camera mimicking parallax. Image from SPIE Digital Library
The way the human visual system works is that each individual eye sees the world from a different angle and merges these images as one through a process called parallax. 3D imaging follows the same approach, using two lenses for every shot—each capturing an image that’s different from the other.
Active vs. Passive 3D Imaging
In general, 3D imaging techniques can be classified as either active or passive.
Active 3D imaging systems are ones that use artificial illumination in order to capture and record digital representations of objects. This artificial illumination provides dense and accurate images, even of textureless objects, which would be otherwise difficult to obtain.
An active 3D imaging system uses different methodologies, including time-of-flight, triangulation, and interferometry. Time-of-flight, for example, requires the encoding of 3D data into each pixel by measuring the time that elapses as light travels to the target object and then returns to a sensor. LiDAR is a popular example of active 3D imaging.
A LiDAR-represented spatial plane. Image used courtesy of Geospatial World
Passive methods, on the other hand, are systems that recover 3D information from scenes that are illuminated only with ambient lighting. They tend to utilize depth from focus and light field. In snapshot-based methods, the difference between two snapshots captured at the same time is used to calculate the distance to objects—a process called passive stereo imaging.
Expanding 3D Imaging Use Cases
Many fields may greatly benefit from 3D imaging.
An obvious application is autonomous vehicles. With improvements in 3D imaging technology will come better visual systems for self-driving cars, allowing them to make more accurate and safe decisions in real-time. Augmented reality is also a use case that will benefit from 3D imaging.
Beyond machine vision, the medical field in particular can see many notable applications. The implementation of 3D imaging in procedures like ultrasound, cardiological exploration, and surgical vision will greatly benefit doctors and patients alike.
Image-guided spinal surgery. Image used courtesy of ODT Magazine
It’s obvious to see why 3D imaging is such a hot field right now, and as the technology continues to improve at the circuit level, many fields may establish 3D imaging as a commonplace technology.