Seismic surveys map 3D underground structures. The seismic waves are refracted and reflected by the structures, the waves do not travel in straight lines but by many paths. Given a single impulse, the software must solve for the 3D structure based on the return of energy via many paths to each of the several geophones on the surface.
Like seismic reflection, radar and sonar transmit a signal and produce an image of the reflectors. They do not image multiple remote sources of energy.
Interferometry has the advantage of a more isotropic propagation medium so it can reconstruct an image of the multiple sources of remote sound excitation. It may be possible to sharpen the image by deconvolving the intervening atmospheric distortions.
Acoustic direction finding was used during WW1 and WW2 for gun and aircraft location. It did not produce an image, just a direction. See the four pictures here.
https://en.wikipedia.org/wiki/Acoustic_location
https://en.wikipedia.org/wiki/Acoustic_location
There is some work on acoustic DF taking place today.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5134470/
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5134470/
Acoustic imaging should be able to highlight the sources of sounds and superimpose them on a visual camera image. It should also be possible to separate out the individual conversations in a crowd by optimising multiple synthesised acoustic spot beams.
Imagine a visual image of a highway with the noise from individual vehicles superimposed in real time, tracking the vehicles as they move. If the array is large enough there will be doppler information available.
Logs said:
a map of all the sounds at a moment
The image will not be made at “a moment” in time. It will be the accumulation of energy over a period of time. I would expect a minimum of 10 seconds recording since low frequencies are involved and full cycles of sound are needed to correlate the different channels and straddle the receive array.