Facial Recognition Available To Police And Public Through Picasa
Originally facial recognition was used only by the government, military and police organizations. Software algorithms would take into account key alignment characteristics and relative ratios. For example, the distance between the center of the eyes, width of lips, cheekbones and jaw. These would then be compared against a database of images to find the face with the most likely match.
This type of recognition works best in well lit environments where there is a clear view of the persons face. Security systems used for authorized access to restricted areas based on facial recognition against an approved access database employ this technology. For law enforcement, images are usually grainy and low resolution.
The suspects are typically trying to conceal their identity with hats, glasses and masks which compound the issue of providing a likely match. Unlike a building security access system which has a pre-defined set of images to compare against, it’s not always the case that in a criminal investigation there is a corresponding photo for the person of interest.
Although not currently linked in the United States, the ability to take video footage from a surveillance camera and run through facial recognition algorithms against a DMV database, state prisons and local police line-up photos is one suggested method to quickly identify suspects in criminal investigations.
Another area where facial recognition falls short is comparing against a drawing or sketch.
A sketch artist will typically work with standard blocks in software to quickly build an image, and then adjust the features based on a verbal description. However, facial recognition tools generally have a difficult time to find a match when a drawing is used for the input as there are too many variations to take into account.
Computer specialists and artificial intelligence developers are working on alternative methods to scan a drawing and accurately compare this to a photograph database to deliver reliable results.
Facial recognition isn’t just for solving criminal cases, and it’s available for free to everyone. Users of Picasa , from Google, can run algorithms on both photos uploaded to the cloud and directly on images stored locally on the hard drive. It’s not quite as robust as what the professionals use, but it’s very user friendly and runs in the background. You get the option of approving and adding the faces that are identified, and can also select “Ignore” for those you don’t want to tag in the future.
Most recently, Facebook announced that they will be running facial recognition on all photos uploaded by users as well as to go back and identify all existing pictures. So whenever a new photo is added, Facebook algorithms will attempt to match the new photo with existing ones and to make a suggested name for each person it can identify in the photo. Presumably, this will make it easier for users to group and tag their photos, and to share with their friends. A point of interest, under pressure from organizations in Europe regarding privacy concerns, this feature is disabled for all European accounts.