Over the last few years there have been some great developments in computer vision recognition. One application for these machine learning vision recognition algorithms has been to look for patterns in satellite images of the Earth.
For example, Onformative has developed an algorithm, called GoogleFaces, that scans satellite imagery looking for patterns that we might recognize as resembling human faces. Terrapattern has also used deep learning machine vision techniques. In their case to find locations which look very similar from the air (share common geographical or urban patterns). Google's Land Lines also demonstrates how machine learning can detect and recognize patterns in abstract imagery. Land Lines allows you to explore satellite imagery through finger or mouse gestures. Draw a line or pattern and Land Lines will find satellite imagery which contains the same shape.
While these applications are great fun to play with they are only demonstrations of the potential use of machine pattern recognition when applied to satellite imagery. The technology actually has some very serious potential real-world applications. For example Texty has trained a computer model to search satellite imagery for evidence of illegal amber mining.
Over recent years in the Ukraine thousands of hectares of land have been stripped in order to mine amber. In fact around 50,000 Ukrainians are now involved in illegal amber mining. Texty created and implemented a machine algorithm that was able to analyze hundreds of thousands of satellite images to identify the locations of illegal amber mines in Ukraine.
You can explore some of the results of this analysis in Земляна проказа, a story map about Ukraine's illegal amber mines. The story map includes a number of satellite images of illegal mining, all of which were found by Texty's algorithm. If you click on these images the web page cleverly turns into a map of the illegal mining location. Click on the 'x' and the map closes and returns you to the story map.
No comments:
Post a Comment