21 comments

[ 2.2 ms ] story [ 56.6 ms ] thread
Help me to do hacker
If you're looking to get into Landsat imagery yourself, I highly recommend the command line tool landsat-util[1] (although you do need to install it using Python 2). It allows you to search with lat/lon pairs, cloud cover, etc, and comes with tools to automatically do the band combining for you so you can have easy real-color images.

I find the the browser-friendly tools like LandsatLook[2] or Earth Explorer[3] to be more difficult to use, but if you're interested USGS has posted some great tutorials for them on YouTube[4]. There are also some interesting ideas about how to use the data itself beyond just creating pretty pictures.

---

[1]: https://github.com/developmentseed/landsat-util

[2]: https://landsatlook.usgs.gov/

[3]: https://earthexplorer.usgs.gov/

[4]: https://www.youtube.com/user/usgs/search?query=landsat

A little off topic but are you aware of anywhere that provides high resolution night-time imagery? So far I've only been able to find a little from Himawari-8.
VIIRS DNB Nightly Mosaics can be found here:

https://ngdc.noaa.gov/eog/download.html

The Visible Infrared Imaging Radiometer Suite Day-Night Band (VIIRS DNB) images the entire Earth nightly at a resolution of about 750 meters. Suomi NPP orbits the Earth 14 times per day on a 16-day repeat cycle. The average time at which images are taken is near 01:30 in local solar time, but can vary by over an hour depending on latitude and the particular cycle.

Thanks! That's the kind of imagery I'm looking for but not the resolution. I should have made it clearer but what I'm looking for is images with resolution like in this article:

http://www.bbc.co.uk/news/resources/idt-sh/syria_from_space_...

They claim to have got these from NASA but the resolution looks a hell of a lot finer than 750m, it looks more like 1m-5m; you can see individual streets in the Raqqa timelapse.

They link to http://www.bbc.co.uk/news/resources/idt-sh/syria_from_space_... but it just mentions VIIRS again.

They should apply deep learning to this :)
If we used deep learning then we probably wouldn't have got into the journal 'Remote Sensing of Environment'

To get into top journals it is really about simple method + novel idea + good story.

The same holds for top Finance journals like 'Journal of Finance', etc.

Based on my readings, 'incomprehensible method + novel idea + good story' seems to work pretty well too.
Looks like someone has: Deep Learning-Based Classification of Hyperspectral Data http://ieeexplore.ieee.org/abstract/document/6844831/
This is exactly what I was hinting at: IEEE journals (TGRS, etc.) are filled with deep learning and machine learning on satellite imagery papers.

They are more tech focused journals and have less impact.