Guy Sata, Matthew Tancik, and Ramesh Raskar from the Camera Culture group in MIT's Media Lab have developed a camera system that can help users see through dense fog. The cameras use fast measurements to create a depth map that can improve contrast by up to 6.5 times in foggy conditions.
The problem statement used for the project was the inability of autonomous navigation systems to see through mist or fog. The group wanted to develop visible light systems, attractive for the high resolution and ability to read words and numbers. Using a laser to fire pulses of light into the fog with a camera to take images and measure how long the light takes to return, the shape of objects and their relative distance can be measured. The time-of-flight camera can count the number of photons returning every 56 picoseconds, and the algorithm uses that data to create a histogram of photon counts. Checking these histograms against the base level of fog in the field of vision creates spikes at places where physical objects exist. Optical depth was also used as a tool to decide how much light passed through the fog and the group hopes that it can soon be used to detect color. At the end of the study the system detected images at a range of 57 centimeters where the human eye could only see 36 centimeters ahead.
The group hopes that the technology can be used to aid human drivers and self-driving cars in inclement weather, but imagine other end uses. Drones might be better able to follow targets in fog, trains could have better information regarding obstacles in the tracks, and airplane traffic could be aided during takeoffs, landings and low altitude flights. Towards Photography Through Realistic Fog is the paper that was presented by the group at the 2018 International Conference on Computational Photography.
This is a great project that showcases how engineers can pull useful information from complex sets of data. Another great aspect of this is the test method devised by the group - the meter-deep chamber built for the test, the study and measurement methodology, and the new fog formulation were all created specifically for this application. The data work done on the project is explained in relatable terms throughout the paper and a technically heavy but interesting read.