Seoul will install 3,000 cameras with artificial intelligence to analyze movement patterns and detect possible crimes
In ‘ Minority Report ‘, a movie starring Tom Cruise in which he plays the head of the PreCrimen police force in Washington DC, they use some young mutants who can see the future to predict the crimes that will happen soon and stop the suspicion before I do nothing. It is science fiction, of course, not like the 3,000 cameras with artificial intelligence that are going to be installed in Seoul, capital of South Korea.
This initiative is the responsibility of the Electronic And Telecommunications Research Institute (ETRI) and the District of Seocho, which according to ZDNet , will install 3,000 cameras throughout the district to process data from pedestrians and “measure the probability that a crime takes place .
Movements and suspicious objects
The cameras will use artificial intelligence to process the location, time, and behavior patterns of people driving through the streets to predict possible crimes, see a robbery. In that sense, the cameras will analyze when someone is at a normal pace or is chasing someone , in addition to detecting if he is wearing glasses, a mask or a hat, and even if they carry a bag or suspicious object. They will also know if it is day or night.
The software is still in development, but ETRI intends to have the final version ready by 2022 . Subsequently, the cameras will expand to other districts in Seoul, as well as to other provinces. It is a titanic task, since in Seoul there are 25 districts where 9.7 million people live (in Madrid there are 6.55 million inhabitants, to put it in context). Throughout Seoul there are more than 55 million inhabitants.
To train the software, the Seocho District and ETRI will use 20,000 court sentences and video files associated with crimes for AI to learn to detect patterns. Once the system is capable of doing so, it can “compare whether what is being recorded in the present matches past criminal patterns,” they explain on ZDNet.
The software will be trained with sentences and recordings of previous crimes
It is easy to see it with an example: if there are many videos that show how someone with a mask follows a person and steals a backpack, the AI will understand that if that is happening again, a similar crime is likely to be committed. In that case, the camera will analyze the odds and, if higher than a certain percentage, will give the notice to the nearest police station and will send a patrol to take a look.
Similar systems have already been tested on occasion and, in fact, some studies show positive results. In a NIST study , a database of 1.6 million people registered by the police was analyzed and the result was 95% successful in identifying criminals . In another that took place in the United Kingdom , however, only 45% of 200 tests were successful. The Chinese police, meanwhile, have glasses with facial recognition that have apparently worked well by detecting certain individuals