Читать книгу Deep Learning Approaches to Cloud Security - Группа авторов - Страница 25

Facial acknowledgment is a classification of biometric programming that maps a person’s facial features mathematically and stores the information as a face print.

Here, we propose a model for application in criminal justice systems. The model suggests the capacity to perform face identification in a group continuously or post-occasion, for open security, in urban communities, air terminals, at fringes, or other sensitive spots like religious congregations. It can help law enforcement agencies in better identification of possible suspects.

First of all, facial detection technique is used to confirm whether the image given as input is a face or not. On being detected as a face, facial marking is done using fiducially point analysis.

We will access the database from the cloud and the personal recognition will use the feature vectors from the training images to train or learn the classification algorithm. Data mining algorithms will be used for further processing and evaluation of data. Feature selection and feature extraction techniques can be used for improved accuracy shown in Figure 1.3.

Figure 1.3 System architecture.

In the end, similarity detection is achieved by using the Nearest Neighbor Approach where the classifiers can be trained using a suitable algorithm like K-Nearest Neighbor.

STEPS:

1 1. An image captured is used as input.

2 2. Detect the image using fiducially points.

3 3. Transform the image by cropping the subsequent markers.

4 4. Apply Feature Selection and Feature Extraction for improved accuracy.

5 5. Use the Feature Vector Extraction method requiring features offloaded from the criminal database on the cloud.

6 6. Interpret and evaluate by comparing the two data sets.

7 7. Use the Nearest Neighbor Approach to identify the similarities between two or more images.

Facial Recognition Technology is intended to operate at a distance, without the knowledge of the target, so that it becomes hard to prevent the face from being captured. Along these lines, it allows for targeting of multiple persons one after another. Moreover, it is a non-consensual and clandestine reconnaissance innovation. The proposed model, when brought to realization, could act as an effective tool for criminal identification by comparing a live capture or digital image to the stored face print in order to confirm an individual’s identity.