The face recognition system is a computer application capable of identifying or verifying a person from a digital image or video clip. One way to do this is to compare selected facial features with the image and options from the database.
Face recognition (program) is usually used in security systems, and can be compared with other biometric systems (for example, fingerprint recognition or eye diaphragm). Recently, it has also become popular as a tool for commercial identification and advertising.
Some face recognition algorithms identify facial features by extracting landmarks or objects from the subject’s face image. For example, the algorithm may analyze the relative location, size, and / or shape of the eyes, nose, cheekbones, and jaw. This data is then used to search for other images with the appropriate parameters. Other algorithms normalize the gallery of facial images, and then compress face information, saving only the data in the image that is useful for facial recognition. Then the desired image is compared with the available data. One of the earliest successful systems is based on pattern matching methods applied to a set of facial features, providing a kind of concise view of appearance.
How does face recognition work?
It includes certain algorithms that can be divided into two main approaches:
- A geometric that considers distinctive or photometric facial features.
- Statistical, which translates the image into values and compares them with patterns to eliminate variances.
3D recognition
A new trend created to achieve higher accuracy is three-dimensional face recognition. This method uses 3D sensors to collect face shape information. This information is then used to identify features such as the contour of the eye sockets, nose and chin.
One of the advantages of three-dimensional face recognition programs is that they are not affected by changes in lighting, unlike other options. This technology can also identify from a different range of viewing angles, including the profile view. Three-dimensional data points significantly improve facial recognition accuracy. 3D research is enhanced by the development of sophisticated sensors that improve the capture of images in the form of three-dimensional images. Sensors work by projecting structured light onto the face. Up to a dozen or more of these image sensors can be placed on one CMOS chip - each of them captures a different part of the spectrum.
However, even the perfect 3D matching method can be sensitive to facial expressions. For this purpose, a team of researchers at Technion used tools from metric geometry to process expressions as isometrics. After that, Vision Access created its solution for three-dimensional face recognition. The company was later acquired by Bioscrypt Inc., which developed a version of the program for face recognition, known as 3D FastPass.
A new method is to introduce a method for capturing a three-dimensional image using three tracking cameras that point to different angles. One of them will indicate the front side of the object, the second - from the side, the third - at an angle. All of them will work together to be able to track the face of the object in real time and be able to identify it. It is believed that any program for face recognition through the camera will soon be based on this technology.
Skin texture analysis
Another new trend is using visual skin details that are captured in standard digital or scanned images. This method, called skin texture analysis, turns the unique lines, patterns, and spots visible in human skin into a mathematical space.
Tests have shown that with the addition of this technology, face recognition efficiency can increase by 20-25%.
Heat chambers
Another form of receiving input for face recognition is the use of thermal imaging cameras. Thanks to this procedure, cameras will only determine the shape of the head, while ignoring objects such as glasses, hats or makeup. The problem of using thermal images for face recognition is that databases are limited for this.
Researchers are currently studying the use of this technology in real life and production landscapes and at the same time are creating a new database of thermal imaging images. The study uses low-sensitivity, low-resolution ferroelectric electrical sensors that are capable of receiving long-wave thermal infrared images (LWIR). The results show that the fusion of LWIR and conventional visual cameras has great results in outdoor probes. With this combination, a very powerful face recognition program for the camera can work.
Mass use
While researchers are working on the latest technologies available only to specialists, developers of mass applications also do not stand still. After the announcement of the creation of Google Glass, there was a lot of noise around face recognition and programs for this on the Internet. It is believed that this will open up a lot of opportunities for users not only to interact with each other, but also with various objects.
Today there is a large list of face recognition interfaces that you can use for your applications. The most common are the following.
Face Recognition Stephen
A development from Lambda Labs, which provides facial recognition by the location of the eyes, the shape of the nose and mouth, and also conducts a gender classification. Available on the official website of the developer.
Face detection
A computer facial recognition program on the Internet that is the perfect replacement for Face.com. It is currently provided for free.
Animetrics Face Recognition
The Animetrics Face Recognition API application can be used to detect faces in photos. Information about facial features or landmarks is returned as coordinates in the image.
Skybiometry
To use it, you must have an application created in your account in SkyBiometry. To create it, a simple registration is enough.
Face ++
This application uses advanced computer vision and data mining technology to provide 3 basic services (detection, recognition and analysis). The program provides detection and analysis of Landmark (23 points), a landmark (81 points), attributes: age, gender, points, race and so on.
Facemark
This is a powerful face recognition API. He finds 68 reference points for a full face image and 35 for a profile. FaceMark detects landmarks for faces in the image specified at the URL or on the downloaded file, and outputs the result as a JSON file containing a vector of face landmarks and match points for each option found.
EmoVu by Eyeris
Intelligent emotion recognition software that allows cameras to read human micro expressions, gender and age group. This is a real-time face recognition program that works when users watch videos on their computers or mobile devices.
Rekognition.com
This is one of the best Face.com alternatives. The fast, reliable and scalable rekognition engine can perform face detection, scanning, recognition and search. It can be automatically trained using images and tags on Facebook. Since it focuses on social networks, this is the best face recognition program for Android.
Faceect
It is also a powerful and free face detection interface. It finds faces (both full face and profile) in the image specified at the URL or downloaded as a file, and can find several faces in one photo, and gives the answer in JSON format. The image is loaded with a bounding box for each face found.