Humans are very good at recognizing faces and complex patterns. Even a passage of time doesn’t effect this capability and it would be very much helpful if computers become as accurate as humans in face recognition…
NEC, Facebook, Google, Microsoft and many other large companies are now developing face recognition technology which used in almost all of their products.
Humans are very good at recognizing faces and complex patterns. Even a passage of time doesn’t effect this capability and it would be very much helpful if computers become as accurate as humans in face recognition.
Over the past decade, there are many studies on the problem of determining the human faces from black and white, grayscale to color images. In order to meed our needs, multiple studies ranges from simple problem with an image contains only a human face looking straight into the capture device, to the color image has multiple human faces in the photo, the face turned a small angle, or partially obscured, and the complexity of the image background (e.g. snapshots in outdoor).
The face recognition is an algorithm to determine the location and size of many faces in the photos (digital photo). This technique identifies features of faces and ignore other things such as building, trees, body,…
Face recognition system can be applied in many ways:
Checking for criminal records
Enhancing of security by using surveillance in conjunction with face recognition system.
Finding lost children’s by using the images received from the cameras fitted at some public places.
Knowing in advance if some VIP is entering the hotel.
Detecting of a criminal at public places.
Can be used in different areas of science for comparing an entity with a set of entities.
How does face recognition work?
Identifying or verifying a person from a digital image or a video frame extracted from a video source.
Face recognition is a process automatically identifying (detection) faces and recognizing a particular person from a photo or video. One way is to compare the characteristics of the face identified with the images in the database.
Until the year 2000, there are many different techniques to detect the face, but all were either slow or unreliable, or both.
A major change in happened in 2001, when Viola and Jones invented Haar-based cascade classifier, a technique used to identify object and it was improved by Lienhart and Maydt in 2002.
The result of identifying objects was fast enough (identifying in real-time on normal PC) and was reliable (more than 95% accuracy).
There are two approaches to the facial recognition: Feature-Based Face Recognition and Appearance-Based Face Recognition.
Haar-like features are the rectangle which is divided into different rectangles.
First, the image is grayscaled, then the haar-like features (rectangle) are shifted through the image, comparing similar image rectangles with Haar-like features, similar ones are marked.
The following are three main steps in the face recognition using haar-like method:
Step 1. Detects the “position of the face”
Step 2. Finds the “feature of the face”
Step 3. Search and identify of the “detected face” in the database.
Why is Face Recognition difficult?
There are many factors that affect the recognition results:
Light: digital image represents the brightness of the object, so that when the light changes, information about the object will be affected.
Distance of the object from the camera: the distance of the object will determine the number of pixels of the face.
Emotional on the face: the emotional expressions on faces caused noise, there is no good method to filter this noise has been developed yet.
Standing position of the object: position of the object affects the recognition result. Big varriant will will lead to fail detection.
Costumes of object: recognition can be affected if the objects are having different outfit than the sample images, such as glasses, hat, …
Real World Application
At Research and Development center (R&D) of FPT Telecom, we are using this technology for Workforce Management project. This project is a part of Workforce Management program.
Workforce management (WFM) is an integrated set of processes that an institution uses to optimize the productivity of its employees on the individual, departmental, and company-wide levels.
Workforce Management (WFM) include the following sub-projects:
Workforce Absence manages the leaves or attendance of employees instead of using traditional paper.
Workforce Scheduling helps to create and manage staff’s schedules.
Workforce Mobility helps users to experience all the features above on mobile devices.
Workforce Performance provides features to manager to evaluate work performance of employees.
Workforce Payroll creates payrolls based on many different criteria.
Workforce Attendance project uses face recognition technology for employee attendance at the business units and departments of the company.
Face recognition progress is performed on the company’s server, and the software still supports offline running when the network is down. The data is synchronized from client to server for storage and backup purpose.
WFM is being deployed for more than 13 business units and departments with more than 2,000 employees. In 2015, WFM system also received about 3,000 requests for leave and late for work.
What is next
We are planning to develop this technology to support access control based on face recognition. The access control system has the ability to proactively and real-time monitor all the employees in a large area.
This technology can also be applied to solve the problem of traffic congestion by monitoring and controlling vehicle traffic flows, vehicle density to be able to give the warning, urban planning guidelines.
Nguyen Ngoc Dinh