PyCon 2016 in Portland, Or
hills next to breadcrumb illustration

Single pixel thermal imaging with the Raspberry Pi and Python

Scott Sievert

Audience level:


Thermal cameras are expensive ($4,000 to $40,000). However, exploiting sparsity and implementing the algorithm on a Raspberry Pi (with Python!) allows for a roughly $400 single pixel thermal camera.


For thermal imaging, we exploit an agile architecture with a single low-cost thermal sensor that can be independently actuated in two dimensions to collect narrow field-of-view spatial “samples” of the scene of interest, equipping it with a simple adaptive sensing methodology to guide the sampling process toward spatial regions of interest, to effectively produce high-resolution thermal images from relatively few samples. With an accompanying hardware system, we provide experimental evaluations to demonstrate the viability of our approach. This hardware system utilizes two precision stepper motors, a Raspberry Pi and a thermal sensor.