Depth-sensing imaging system can peer through fog

Computational photography could solve a problem that bedevils self-driving cars.


A failure to deal with hazy driving conditions has been one of the central impediments to the improvement of self-governing vehicular route frameworks that utilization obvious light, which are desirable over radar-based frameworks for their high determination and capacity to pursue street signs and track path markers. Thus, MIT scientists have take a step forward to develop a depth-sensing imaging system that can deliver pictures of articles covered by mist so thick that human vision can’t infiltrate it. It can likewise check the items’ separation.

Amid testing, scientists used a small tank of water with the vibrating motor from a humidifier immersed in it. They found that the system could enter just 36 centimeters, the framework could resolve pictures of articles and measure their profundity at a scope of 57 centimeters.

According to scientists, 57 centimeters is not a great distance, but the fog produced for the study is far denser than any that a human driver would have to contend with. The imperative point is that the system performed superior to human vision, though most imaging frameworks perform far more terrible. A route framework that was even tantamount to a human driver at driving in haze would be a tremendous leap forward.

Guy Satat, a graduate student in the MIT Media Lab said, “I decided to take on the challenge of developing a system that can see through the actual fog. We’re dealing with realistic fog, which is dense, dynamic, and heterogeneous. It is constantly moving and changing, with patches of denser or less-dense fog. Other methods are not designed to cope with such realistic scenarios.”

Guy Satat, a graduate student in the MIT Media Lab, who led the new study.  Image: Melanie Gonick/MIT
Guy Satat, a graduate student in the MIT Media Lab, who led the new study.
Image: Melanie Gonick/MIT

The system involves a time-of-flight camera, which fires ultrashort bursts of laser light into a scene and measures the time it takes their reflections to return. The camera also counts the number of light particles, or photons, that reach it every 56 picoseconds, or trillionths of a second. This system uses those counts to generate a histogram- essentially a bar graph, with the heights of the bars indicating the photon counts for each interval.

After that, the system search for gamma distribution that best fits the state of the visual diagram and basically subtracts the related photon checks from the deliberate sums. What remains are slight spikes at the separations that relate with physical impediments.

When scientists tested the system using a fog chamber a meter long, they mainly mounted regularly spaced distance markers, which provided a rough measure of visibility. They also involved other objects such as a wooden figurine, wooden blocks, silhouettes of letters, so that the system can image even when they were indiscernible to the naked eye.

There are different ways to measure visibility, however: Objects with different colors and textures are visible through fog at different distances. So, to assess the system’s performance, scientists used a more rigorous metric called optical depth, which describes the amount of light that penetrates the fog.

Srinivasa Narasimhan, a professor of computer science at Carnegie Mellon University said, “Bad weather is one of the big remaining hurdles to address for autonomous driving technology. Guy and Ramesh’s innovative work produces the best visibility enhancement I have seen at visible or near-infrared wavelengths and has the potential to be implemented on cars very soon.”

Optical profundity is free of separation, so the execution of the framework on mist that has a specific optical profundity at a scope of 1 meter ought to be a decent indicator of its execution on mist that has the same optical profundity at a scope of 30 meters. Actually, the framework may even charge better at longer separations, as the contrasts between photons’ entry times will be more prominent, which could make for more exact histograms.

Satat and his colleagues describe their system in a paper they’ll present at the International Conference on Computational Photography in May. Satat is first author on the paper, and he’s joined by his thesis advisor, associate professor of media arts and sciences Ramesh Raskar, and by Matthew Tancik, who was a graduate student in electrical engineering and computer science when the work was done.

- Advertisement -

Latest Updates