0% found this document useful (0 votes)
1 views10 pages

Criminal Face Detection System

The document presents a Criminal Face Detection System that utilizes Local Binary Patterns Histogram (LBPH) and Fisherface algorithms for automated facial recognition to enhance crime identification and prevention. It discusses the methodology, including face detection using Haar Cascade Classifier, feature extraction, and evaluation metrics, demonstrating the system's potential effectiveness in real-time applications. The results indicate high accuracy in face recognition, although challenges remain in detecting faces under certain conditions.

Uploaded by

SUBHASISH BEHERA
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
1 views10 pages

Criminal Face Detection System

The document presents a Criminal Face Detection System that utilizes Local Binary Patterns Histogram (LBPH) and Fisherface algorithms for automated facial recognition to enhance crime identification and prevention. It discusses the methodology, including face detection using Haar Cascade Classifier, feature extraction, and evaluation metrics, demonstrating the system's potential effectiveness in real-time applications. The results indicate high accuracy in face recognition, although challenges remain in detecting faces under certain conditions.

Uploaded by

SUBHASISH BEHERA
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

Criminal Face Detection System

Machala Venkatesh (  manhchalavenkatesh10@gmail.com )


Vardhaman College of Engineering
Chintha Dhanalakshmi (  chinthadhanalakshmi7@gmail.com )
Vardhaman College of Engineering
Aditya Adapa (  adityaadapa2002@gmail.com )
Vardhaman College of Engineering
Mr. Manzoor MD (  manzoor@vardhaman.org )
Vardhaman College of Engineering
Anvesh K (  Anvesh@vardhaman.org )
Vardhaman College of Engineering

Research Article

Keywords:

DOI: https://doi.org/

License:   This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License

Additional Declarations: No competing interests reported.


Criminal Face Detection System
Machala Venkatesh, Chintha Dhanalakshmi† , Aditya
Adapa† , Mr.MD. Manzoor† and K Anvesh†
Department of Information Technology, Vardhaman College of
Engineering, Shamshabad, Hyderabad, 500052, Telangana, India.

*Corresponding author(s). E-mail(s):


manhchalavenkatesh10@gmail.com;
Contributing authors: chinthadhanalakshmi7@gmail.com;
adityaadapa2002@gmail.com ; manzoor@vardhaman.org ;
Anvesh@vardhaman.org;
† These authors contributed equally to this work.

Abstract
The increasing crime rate and the need for efficient criminal identifica-
tion and prevention have led to the adoption of security technologies
such as CCTV cameras. In this study, we propose an automated
facial recognition system using the Local Binary Patterns Histogram
(LBPH) classifier and Fisherface algorithm. The system utilizes a Haar
feature-based cascade classifier to detect faces in real-time, and the iden-
tified faces are then matched against a criminal database. Although
accurate face identification remains a challenge, the Viola-Jones frame-
work is utilized to pinpoint face positions and other features in a
picture. Face detection classifiers are publicly available through orga-
nizations like OpenCV. Our proposed system has the potential to
enhance criminal identification and prevention in public and private
spaces. Keywords: Criminal identification, CCTV, facial recognition,
LBPH classifier, Fisherface algorithm, real-time, Viola-Jones, OpenCV.

Keywords: Criminal identification, CCTV, facial recognition, Haar classifier,


real-time, Viola-Jones, open CV.

1
2 Criminal Face Detection System

1 Introduction
In this study, we utilize the Local Binary Patterns Histogram (LBPH) clas-
sifier and Fisherface algorithm for automated facial recognition in order to
prevent crime and ensure public safety. The use of closed-circuit television
(CCTV) cameras has become increasingly common in both public and pri-
vate settings, and a deep learning-based approach provides real-time data to
enhance the efficiency of police forces. The three processes of face identifica-
tion - face detection, feature extraction, and face recognition - are essential
for accurate results. To locate and trace facial feature points, a local search
is used. The Viola-Jones approach is utilized to detect faces and produce a
classifier using AdaBoost. The HAAR cascade classifiers are employed in this
system for face detection. After converting the image to grayscale, we create
a square shape around the face and perform face normalization to remove any
features that may cause inaccuracies. The proposed system has the potential
to improve face recognition in various applications and enhance public safety.

2 Literature Survey
Several methodologies have been proposed for real-time face recognition. Viola-
Jones developed a framework that can accurately detect faces in challenging
conditions, such as erratic head movement or poor lighting. Ni Kadek et al.
proposed an eigenface approach that uses OpenCV library for face recognition.
Shreyak Sawhney et al. created a real-time smart attendance system that
uses Eigenface values, PCA, and convolutional neural networks. Weihua Sheng
et al. established a facial recognition framework for a security system using
TensorFlow.
For my proposed methodology, I plan to use a combination of Local Binary
Pattern Histograms (LBPH) classifier and Fisherface. This method has been
shown to be effective in accurately recognizing faces in real-time applications.
The LBPH classifier is known for its robustness to lighting changes and minor
variations in facial expressions, while Fisherface is useful for reducing dimen-
sionality and improving accuracy. By combining these two methods, I hope to
create a system that can recognize faces in real-time with high accuracy, even
in challenging conditions.

3 Proposed Methodology
Based on the proposed methodology, the LBPH (Local Binary Patterns His-
tograms) classifier and Fisherface methods will be used for face recognition.
LBPH is a powerful feature extraction method that describes the local fea-
tures of an image, while Fisherface is a technique that focuses on the global
features of an image. The combination of both methods will enable the system
to identify faces in complex images with a high degree of accuracy. The sys-
tem will compare the extracted features of the input image with the features
stored in the dataset to recognize the face. The LBPH and Fisherface methods
Criminal Face Detection System 3

have been widely used in previous research and have shown promising results
in various applications of face recognition.

3.1 Face Detection Using, Haar Cascade Classifier


Algorithm
Our proposed approach involves utilizing two algorithms, namely LBPH clas-
sifier and Fisherface, for face recognition. However, for the purpose of face
detection, we have adopted the Haar Cascade Classifier algorithm developed
by Viola and Jones [10]. This algorithm works by searching for pre-defined
Haar features [11] instead of individual pixels on a face. Whenever one of these
features is detected, the corresponding sub-window, known as a face candi-
date, is allowed to proceed to the next round of detection. The face candidate
is a fixed-size rectangular sub-window, typically 24x24 pixels, that is a part
of the original image. To account for faces of different sizes, this sub-window
is resized before being used to scan the entire image. As the system scans the
image, each appropriate location is marked as a potential face.

Fig. 1 Architecture of FRCI

3.1.1 LBPH Classifier Algorithm


The first algorithm we use is the Local Binary Patterns Histograms (LBPH)
classifier. This method extracts features from the input image and compares
them with the features of the images stored in the training dataset. The LBPH
algorithm calculates a histogram of the local binary patterns of the image,
which captures the texture information of the image. This algorithm is used
for facial feature extraction and face recognition.

3.2 Fisherface Algorithm


The second algorithm we use is the Fisherface algorithm, which is also known
as the Linear Discriminant Analysis (LDA) classifier. This algorithm is used
for dimensionality reduction, which reduces the number of features required for
face recognition while maintaining the essential information. The Fisherface
4 Criminal Face Detection System

Fig. 2 Integral Image Generation

algorithm achieves this by projecting the input data into a lower-dimensional


space that maximizes the ratio of between-class variance to within-class vari-
ance. This algorithm is used to recognize faces in the lower-dimensional feature
space.

Fig. 3 Flow Chart of Face Detection

Fig. 4 Common Haar Features

3.3 Feature Selection:


In the proposed methodology, we select the most discriminative features to
improve the performance of the face recognition system. We use the chi-squared
test and mutual information to select the features that are most relevant to
Criminal Face Detection System 5

the task of face recognition. This helps to reduce the dimensionality of the
feature space and improves the accuracy of the system.[15]

3.4 Evaluation Metrics


To evaluate the performance of the proposed methodology, we use standard
evaluation metrics such as precision, recall, and F1 score. Precision measures
the proportion of true positives among the predicted positives, while recall
measures the proportion of true positives among the actual positives. The
F1 score is the harmonic mean of precision and recall, and provides a single
metric for evaluating the performance of the system. We use these metrics to
compare the performance of our proposed methodology with other state-of-
the-art methods. The classifier declares the final result affirmative, indicating
that the required item was found in the picture, when all stages, including the
most recent one, yield positive findings. If the labelling is unsuccessful, the
window is moved to the next location, and the region is correctly characterized
at that location. The region proceeds to the next stage of classification if the
labelling is successful.

4 Implementation
4.1 Importing Required Modules for LBPH Classifier
and Fisherface Algorithm
n order to perform facial recognition using the LBPH classifier and Fisherface
algorithm, we need to import certain modules. These include the cv2 module
for face detection and recognition, the os module for modifying image and
directory names, the image module for reading images in the gif format, and
the NumPy module for saving images as Numpy arrays.

4.2 Load the Face Detection Cascade


The first step in facial recognition is to label the faces in the images. We can do
this by loading the face detection cascade. We will be using the Haar Cascade
for face detection from the OpenCV library. The frontal face default.xml algo-
rithm will be used to identify the face. We will use the cv2.CascadeClassifier
function to load the cascade XML file. If the XML file is located in the current
working directory, relative paths are used.

4.3 Creating the Face Recognizer Object


The face recognizer object must be created in the next step. The face recog-
nizer object, like FaceRecognizer, has characteristics. Use the train() function
to train the FaceRecognizer and recognizer. Predict() [16] can be used to
recognise a face. OpenCV currently provides three face recognition algorithms:
Eigenface, Fisherface, and Local Binary Patterns Histograms (LBPH).
6 Criminal Face Detection System

Fig. 5 Training Data Creation

Because real life isn’t perfect, we used an LBPH recognizer. We simply


cannot guarantee that your images will have superb lighting or that a per-
son will appear in ten different photographs. LBPH focuses on eliminating
regional features from photographs. Instead of seeing the entire image as a
high-dimensional vector, the idea is to characterise just the local features of
an item. Each pixel is evaluated in respect to its surroundings.

5 Results
5.1 Homepage

Fig. 6 Homepage

5.2 Criminal Registration

Fig. 7 Criminal Registration


Criminal Face Detection System 7

5.3 Detect Criminal Face

Fig. 8 Detecting the Criminal Face

The output of a facial recognition system that uses the LBPH classifier
and Fisherface algorithm to detect criminal faces would be a prediction or
identification of the criminal in question. The result of this prediction would
depend on the accuracy and reliability of the facial recognition system, as well
as the quality and quantity of the images available for training the system. It
is important to note that facial recognition systems, including those that use
the LBPH classifier and Fisherface algorithm, are not foolproof and may have
biases or inaccuracies. Therefore, they should not be solely relied upon for the
detection of criminal faces

5.4 Video Surveillance

Fig. 9 Detecting the Criminal in Video Surveillance

The result of the output of a facial recognition system using the LBPH
classifier and Fisherface algorithm would be the identification or prediction
of a person’s face in an image or video frame. The system can detect facial
features in the input image or frame, and use those features to match the face
with the closest match in its database.
In the case of video frames using the webcam on a computer, the facial
detection module would be applied to each frame in real-time to identify any
offenders that appear in the video. This can be useful in scenarios such as
security monitoring, where real-time identification of individuals is required.
8 Criminal Face Detection System

5.5 Local Binary Pattern face recognition results


Our methodology involved detecting faces in all of the input photographs in
our dataset before running the face recognition algorithm to extract Local
Binary Patterns (LBPs). While this approach yielded a high accuracy of 98
percent, it also took considerable time since LBPs had to be computed for
each cell, resulting in over 2 minutes being taken to identify every face in our
sample. Furthermore, the inference was slow as a nearest-neighbor search had
to be run throughout the entire training set.
For each face, we used the recognizer’s predict method to return a 2-tuple
of the subject’s integer label and the confidence, which was based on the chi-
squared distance between the current testing vector and the nearest data point
in the training data. As the distance decreased, the likelihood of two faces
belonging to the same individual grew.
1.We began by loading the face detector model to detect faces in the input
photographs.
2.We then loaded the dataset containing 397 images that were to be used
for training and testing the face recognition model.
3.The face recognition model was trained using LBPH and Fisherface
algorithms, and the process took 3.0534 seconds.
4.Next, we gathered predictions by running the trained model on the test
dataset.
5.The inference, which involved comparing the test data to the trained
model, took 127.8619 seconds to complete.

Table 1 Comparison Table

Names Precision Recall f1 Score Support

Abraham 0.95 0.94 0.94 5


Allen 096 0.96 0.96 8
David 0.97 0.80 0.89 5
Jennifer 0.86 1.00 0.92 6
Accuracy 0.98 98
Macro avg 0.98 0.93 0.95 97
Weighted avg 0.97 0.96 0.98 97

Results Obtained by Using LBPH Algorithm

6 Conclusion
In this proposed methodology, we aim to enhance facial recognition capabilities
by using a combination of LBPH classifier and Fisherface algorithm. LBPH
classifier is able to accurately detect faces in a range of lighting conditions,
and can identify individuals even with only one training image. Meanwhile,
Fisherface algorithm can effectively extract relevant facial features and clas-
sify them. By utilizing these two approaches, we can improve the accuracy,
Criminal Face Detection System 9

precision, recall, and F1 score of our facial recognition system. However, it is


important to note that the system may have difficulty in detecting faces that
are rotated by 45 degrees along the vertical and horizontal axes, as the detector
is optimized for frontal photographs of faces.

References
[1] Belhumeur, P. N., Hespanha, J.P., Kriegman, D. J. (1997). Eigenfaces
vs. Fisherfaces: Recognition Using Class Specific Linear Projection.IEEE
Transactions on Pattern Analysis and Machine Intelligence. 19, pp. 711-
720. IEEE Computer Society.

[2] Bornet, O. (2005, May 19). Learning-Based Computer Vision with Intel’s
Open Source Computer Vision Library. Retrieved April 2007, 2007.

[3] Brunelli, R., Poggio, T. (1993). Face Recognition: Features versus tem-
plates. IEEE Transaction on Pattern Analysis and Machine Intelligence,
15 (10), 1042-1052.

[4] Viola, P. and Jones, M. Rapid object detection using a boosted cascade
of simple features. IEEE Conference on Computer Vision and Pattern
Recognition, 2001.

[5] P. Viola and M. Jones. Robust Real-time Object Detection. International


Journal of Computer Vision, 57(2):137–154,2002.

[6] Lorencik, D., Ondo, J., Sincak, P., Wagatsuma, H. (2015). Cloud-Based
Image Recognition for Robots. In Robot Intelligence Technology and
Applications 3 (pp. 785-796). Springer, Cham.

[7] https://en.wikipedia.org/wiki/Cascadingclassifiers

[8] Abin, A. A., Fotouhi, M., Kasaei, S. (2009, October). Realtime multi-
ple face detection and tracking. In 2009 14th International CSI Computer
Conference (pp. 379-384). IEEE.

You might also like