Seeing Blind No More: The Future of Safely Navigating Blind Corners

Sheila Seidel develops safer systems with various applications through her work on non-line-of-sight imaging.

Sheila Seidel (ECE)

Imagine you are a soldier walking through a tunnel in enemy territory. You come to an intersection and must choose to continue to the right or left. The wrong choice could be fatal, leading you straight into your enemy’s direct line of fire. In this situation, the ability to see around each corner could be lifesaving. Sheila Seidel, a fourth year ECE PhD student under Professor Vivek Goyal (ECE), is developing the technology that could save your life in this situation.

This technology, a type of non-line-of-sight (NLOS) imaging, forms images of objects that are obscured from view by an occluding object (objects that block the scene from direct view). In Seidel’s work, these occluding objects are walls with a vertical edge. “A wall is a structure that is naturally present in a lot of NLOS imaging applications. The wall’s vertical edge helps to discern angular information about the hidden scene”, explains Seidel. Due to this edge, light from around the wall is cast onto the floor on the visible side forming a fan-like pattern, called a penumbra. Measurements of this visible light pattern may be used to reconstruct images of the hidden scene. Seidel develops algorithms to form these NLOS images, hopefully paving the way for advances in a variety of applications, including autonomous vehicle navigation, first responder safety, and defense.

Seidel is funded through a fellowship at The Charles Stark Draper Laboratories, a not-for-profit research organization based in Cambridge, MA. Seidel previously worked for MITRE, a non-profit organization that provides technical and engineering guidance for the federal government. Seidel’s interest in signal processing and computational imaging techniques influenced her desire to work on NLOS imaging problems, and there are many compelling real-world applications. “The ability to anticipate danger around a corner could be incredibly useful for a soldier navigating a tunnel system,” says Seidel.

Seidel’s group has been developing their work in phases, building off of an original study done by Katie Bouman’s team at MIT. Her first paper, Corner Occluder Computational Periscopy: estimating a Hidden Scene from a Single Photograph, published in the June 2019 issue of IEEE Xplore, formed 1D, angular images of moving and non-moving hidden scenery. In her second paper, which recently tied for first place in the 2021 CISE Best Student Paper Awards Competition, Two Dimensional Non-Line-of-Sight Scene Estimation from a Single Edge Occluder, she added range information to form 2D images of hidden scenery. Both of these papers employ passive NLOS imaging methods, meaning only light already present in the environment is used. Seidel is currently working on an active NLOS system, where a pulsed laser is used to introduce light into the hidden scene. Building upon prior work from her group, she hopes to use time resolved measurements collected around the edge to form 3D images of NLOS objects. Although the active system makes 3D reconstruction possible, there are tradeoffs. The passive system requires less expensive equipment, less power, and may be stealthier. “The ability to image around the corner undetected may be important for certain applications, particularly in defense”, says Seidel.

Seidel’s algorithms depend on the ability to extract a clear image of the penumbra (the shadow on the visible side of the wall), however certain environmental conditions make this more challenging. For example, nearby light sources on the visible side of the wall or imperfect vertical edges may not match the team’s model. These real-world challenges could interfere with the accuracy of Seidel’s reconstructed images.

Although this work is far from being implemented into autonomous vehicles, a senior design team working with Seidel’s group is transferring select algorithms into a mobile app. To date, Seidel’s team has worked only with stationary tripod mounted cameras. The senior design team’s goal to implement these algorithms on a moving handheld phone brings this work one step closer to real world applications. Another intended purpose of the app is as an educational tool in high schools. “We hope to inspire young students to get excited about the power of science and engineering,” says Seidel.

Whether her work aids in defense situations, increases the effectiveness of first response vehicles, or improves the safety of autonomous vehicles, Seidel’s work is crucial in developing safe systems.

Read more about Seidel’s work here.