CHANNELS

Aerospace

Automotive

Automated & Autonomous Vehicles

Battery & Electrification Technology

Defense

Drone TV

Electronics

Manufacturing & Prototyping

Materials

Medical

Motion Control

RF & Microwave Technology

Robotics & Automation

Sensors & IoT

Test & Measurement

Aerospace
Automotive
Automated & Autonomous Vehicles
Battery & Electrification Technology
Defense
Drone TV
Electronics
Manufacturing & Prototyping
Materials
Medical
Motion Control
RF & Microwave Technology
Robotics & Automation
Sensors & IoT
Test & Measurement
Aerospace & Defense
Search Results

Camera Can 'See' Moving Objects Around Corners

A camera system developed at Stanford University  can reconstruct room-sized scenes and moving objects that are hidden around a corner. The technology could one day help autonomous cars and robots operate even more safely than they would with human guidance. Previous systems for viewing scenes outside a camera’s line of sight relied on objects that either reflect light evenly or strongly. But real-world objects, including shiny cars, fall outside these categories. Central to Stanford's advance is a very powerful laser; it scans a wall opposite the scene of interest and that light bounces off the wall, hits the objects in the scene, bounces back to the wall, and to the camera sensors. The system can scan at four frames per second and can reconstruct a scene at speeds of 60 frames per second on a computer with a graphics processing unit, which enhances graphics processing capabilities.

Topics:
Automation Automotive Data Acquisition Imaging Photonics Robotics Sensors

More From SAE Media Group

    Transcript

    00:00:06 omitting objects outside a camera's direct line-of-sight has important applications in robotic vision medical imaging remote sensing and many other domains time-of-flight based on line-of-sight imaging systems use pulsed lasers and time-resolved sensors to illuminate a position on a surface probe the indirect light which returns from a hidden object and recovered 3d geometry

    00:00:28 recent approaches to solving this challenging inverse problem include filtered back projection and the light cone transform however these approaches are either slow assume a hidden object with lambertian reflectance or may not account for non planar scanning surfaces or non-conformance we present a fast wave based image formation model for non line-of-sight

    00:00:49 imaging called FK migration inspired by inverse methods used in seismology this frequency domain approach generalizes to reflective or specular surfaces and improves upon other approaches to demonstrate our method we build a hardware prototype capable of scanning room sized scenes and capturing outdoor results here is a set of control commands captured for an indoor scene if

    00:01:15 we play the measurements over time we can directly visualize light from the hidden volume splashing against the wall like ripples on a pond also note the bright specular flashes from the disco ball here's a comparison of the reconstructed 3d volume with each of these methods note that FK migration produces a clean high-quality result

    00:01:51 [Music] [Music] in seismology FK migration is used to map out the structure of underground surfaces an explosion generates shock waves which scatter off of underground structures and are then recorded by geo phones at the surface here the shockwave is analogous to the wavefront of light an optical non line of sight which

    00:02:30 propagates out words from each illuminated point on the wall rather than modeling the two-way propagation of light from the wall to the object and back we can think of the light as being emitted by the hidden object at time 0 and traveling at half the speed of light to the wall the captured measurements thus contain the intensity of a wave field arriving at the wall over time and

    00:02:51 reconstructing the geometry of the hidden volume corresponds to finding the value of this wavefront at time 0 this is a boundary value problem which requires us to migrate the field from one boundary condition at the wall to another in time FK migration solves this problem in three simple steps starting with the measurements at the wall we take the

    00:03:12 Fourier transform to calculate the spectrum we interpolate the spectrum based on constraints from the wave equation as detailed in the paper applying the inverse Fourier transform returns the hidden volume our hardware prototype consists of a laser detector electronics and various optics and optima Kanaka parts used for focusing and scanning this system

    00:03:36 enables fast high-resolution scanning with an average laser power of roughly ten thousand times greater than previous non-line of sight systems this scene consists of simulated measurements of a specular Bunny using FK migration allows us to more accurately recover the 3d geometry compared to the light-cone transform moreover as shown in this captured scene

    00:04:03 of a bicycle FK migration is more resilient to noise in the measurements for short exposure times we demonstrate capture at interactive rates of two frames per second with a higher resolution scan and four frames per second at lower resolution here the actor is dressed in a retro reflective suit to increase the amount of available signal

    00:04:26 note that the position and pose of the actor can be readily recognized in the reconstruction an outdoor scene consists of a statue and a plant on top of a table with a white tablecloth we capture the scene for 50 minutes under in direct sunlight during Twilight and reconstruct the 3d geometry we capture a number of other results and demonstrate extensions of

    00:05:11 our method to handle non-conformance and non planar scanning surfaces as detailed in the paper we will also make these datasets publicly available in conclusion we make a step forward for an on line of site imaging with a new wave migration procedure which handles varied surface reflectance properties is fast robust to noise and which we demonstrate on large-scale scenes