Every time you unlock your mobile phone with your face or pass through your high-tech office surveillance system, an elaborate facial recognition technology is operating in the background. So what is face recognition, and how can you carry out face recognition using MATLAB?
Facial recognition is the process of identifying human faces through technology. The facial recognition system uses computer vision and Machine Learning techniques to model and classifies facial features extracted from images and videos. Algorithms for face identification extract and map facial features and compare them to a database of known faces to find the best match.
MATLAB in Face Recognition
It is possible to achieve face recognition using MATLAB code. The built-in class and function in MATLAB can be used to detect the face, eyes, nose, and mouth. The object vision.CascadeObjectDetector System of the computer vision system toolbox recognizes objects based on the Viola-Jones face detection algorithm.
Description of the MATLAB Object Detector
The vision.CascadeObjectDetector makes use of the Viola-Jones algorithm for the identification of faces, eyes, mouth, nose, or the upper body. A custom classifier can be trained by using MATLAB’s Image Labeler and used along with the System object. So how are facial features or the upper body detected in an image? Here are the steps:
- The first step involves the creation of the vision.CascadeObjectDetector object and setting its properties.
- In this step, the object is invoked with arguments (as if it were behaving like a function).
Syntax for the Creation of the Object Detector
The syntax used for the creation of the Object Detector is as follows:
- detector = vision.CascadeObjectDetector
- detector = vision.CascadeObjectDetector(mode1)
- Detector = vision.CascadeObjectDetector(Name,Value)
- detector = vision.CascadeObjectDetector(XMLFILE)
Description of the Syntax
- detector = vision.CascadeObjectDetector: This syntax is used for the creation of a detector that detects objects using the Viola-Jones algorithm.
- detector = vision.CascadeObjectDetector(mode1): This syntax is used for the creation of a detector that is configured for detecting objects defined by the input vector – mode1.
- detector = vision.CascadeObjectDetector(Name,Value): This syntax is used for setting properties by using one or more than one name-value pairs, where each property name is enclosed within quotes. For example: detector = vision.CascadeObjectDetector(‘ClassificationModel’,’UpperBody’)
Check out: Machine Learning Project Ideas
Unless otherwise specified, the features of the cascade Object Detector are non-tunable, meaning that after calling the object, their values cannot be altered. Objects get locked when they are invoked, and their unlocking is done by the ‘release’ function.
On the other hand, a tunable property would mean that its value can be changed at any time. So, let us have a look at some of the features before we understand face recognition using MATLAB codes.
- ClassificationModel: Defined as a character vector, this property is responsible for controlling the type of object to detect. The default configuration of the detector detects faces.
- MinSize: The size of the smallest recognizable object is denoted as a two-element vector [height width]. Unless a property value is specified, the detector sets it to the image size used for training the classification model.
- MaxSize: The size of the smallest recognizable object is denoted as a two-element vector [height width]. Unless a property value is specified, the detector sets it to size (I).
- ScaleFactor: It has a specified value higher than 1.0001. This property is for incremental scaling of the detection resolution between MinSize and MaxSize.
- MergeThreshold: It has a specified integer value equal to 4. In case there are multiple detections around a target object, the threshold defines the final detection criteria.
- UseROI: Specified as false, this property can be set to true for the detection of objects within a rectangular region of interest in the input image.
Syntax for Using the Object Detector
- bbox = detector(I)
- bbox = detector(I,roi)
Description of the Syntax
- bbox = detector(I) returns bbox, an M-by-4 matrix, that defines ‘M’ bounding boxes that contain the detected objects.
- bbox = detector(I,roi) is used for detecting objects within the rectangular region of interest, specified by roi.
Natural Language Processing
- I — Input image: It is specified as true color or grayscale (RGB).
- model — Classification model: It is specified as a character vector and describes the object type to be detected.
- XMLFILE — Custom classification model: Specified as an XML file, it can be created using OpenCV training functionality or the trainCascadeObjectDetector function.
- roi — Rectangular region of interest: A four-element vector [x y width height] is used to specify this input argument.
bbox — Detections: Detections are returned as an M-by-4 element matrix, each row of which contains the four-element vector [x y width height].
Object Functions Common to All System Objects
- step: For running System Object algorithm
- release: For releasing system resources
- reset: For resetting the internal states of System Object.
MATLAB code for face recognition
In this section, we will see an example of face recognition using MATLAB code.
The step(Detector,I) will return Bounding Box value containing [x,y,Height,Width] of the objects under detection:
- The passing of the argument ‘Nose’ denotes that the object of interest is the nose.
- The default nose detection syntax is vision.CascadeObjectDetector(‘Nose’)
- The default parameter values passed to vision.CascadeObjectDetector can be modified based on the input image.
- The ‘MergeThreshold’ value can be overridden to avoid multiple detections around the target object (as in the image above).
While face recognition technology is beneficial for verification of personal identity, it does raise privacy issues. Since the technology uses an individual’s faceprint, it is often regarded as a breach of one’s privacy, safety, and security. Face recognition using MATLAB can be employed in several cases where security is of utmost concern. From airports and offices to smartphones, facial recognition has become an integral component of many systems and organizations.
If you’re interested to learn more about machine learning, check out IIIT-B & upGrad’s PG Diploma in Machine Learning & AI which is designed for working professionals and offers 450+ hours of rigorous training, 30+ case studies & assignments, IIIT-B Alumni status, 5+ practical hands-on capstone projects & job assistance with top firms.