Advances in non-mechanical frequency-diverse apertures and reconstruction algorithms have made real-time millimeter-wave data acquisition and volumetric imaging possible. Fast frame rates allow imaging people in motion, which represents a tremendous opportunity to increase security screening throughput over existing solutions where subjects must individually strike and hold a pose. However for non-mechanical systems specularity coupled with limited viewing perspectives diminish coverage for individual images.
To mitigate these issues, a system can leverage relative motion between the aperture and subject for a diversity of perspectives across several images. Such an image set offers overlapping and complementary swaths of subject coverage. By stitching together these images a composite image of the subject can be produced with much better overall coverage.
Of course, people change shape as they move, which significantly complicates the image stitching registration and blending process. A deformable geometric model of a person suitable for real-time stitching is required. Drawing from the field of computer animation, we introduce a deformation model of a person based on Shape Key Deformation (SKD) and Skeletal Subspace Deformation (SSD). SKD blends shapes together, while SSD utilizes a simplified “skeleton” to guide deformation and modulate SKD. Assuming the pose of the skeleton is known, the deformation model is able to map any arbitrary image of a person onto a single rest image for stitching. The model is simple, fast, and robust. We go on to demonstrate image stitching of a simulated person in motion using software that models a massively multistatic MIMO metasurface computational imaging system.
Kenneth P. Trofatter, Jonah N. Gollub, and David R. Smith, "A geometric model for stitching millimeter-wave images of people in motion (Conference Presentation)," Proc. SPIE 10634, Passive and Active Millimeter-Wave Imaging XXI, 106340H (Presented at SPIE Defense + Security: April 19, 2018; Published: 14 May 2018); https://doi.org/10.1117/12.2305222.5783327730001.
Conference Presentations are recordings of oral presentations given at SPIE conferences and published as part of the proceedings. They include the speaker's narration with video of the slides and animations. Most include full-text papers. Interactive, searchable transcripts and closed captioning are now available for most presentations.
Search our growing collection of more than 22,000 conference presentations, including many plenaries and keynotes.