SMART ATTENDANCE USING OPENCV FACIALRECOGNITION

Turkish Journal of Physiotherapy and Rehabilitation; 32(3) ISSN 2651-4451 | e-ISSN 2651-446X

www.turkjphysiotherrehabil.org 2757

SMART ATTENDANCE USING OPENCV FACIALRECOGNITION

K.Seemanthini1, V.Kamatchi Sundari2, R.Gopi3, Satyasundara Mahapatra4, Praveena Akki5 1Department of Information Science and Engineering,DayanandaSagar Academy of Technology and

Management, Bangalore, Karnataka, [email protected] 2Department of Electronics and Communications Engineering, SRM Institute of Science and

Technology,Ramapuram, Chennai. ([email protected]) 3Department of Computer Science Engineering, Dhanalakshmi Srinivasan Engineering College,

Perambalur. ([email protected]) 4Computer Science and Engineering, Pranveer Singh Institue of Technology. Kanpur, Uttar Pradesh -

209305. ([email protected]) 5School of Computer Science and Engineering, Vellore Institute of Technology, Vellore.

([email protected])

ABSTRACT

The technology is advancing day to day in every field, we believe that schools and colleges can also implement some of the new technology in their day-to-day activities. As taking attendance is important yet repetitive task, this can be tackled using some of the technologies, facial recognition being one of the technology. The aim of this paper is to examine face detection using open CV to mark the presence of recognized faces in any camera image. The image of the classroom with all students will be captured ,the image captured will be served as input for the facial recognition algorithm to recognize the faces of the students in the picture captured and mark their respective attendance. Open CV extracts countenance of the given images and retains a number of the variations within the image data. Currently, many techniques are available for the detection of faces,we will specifically focus on facial recognition using OpenCV.

Keywords: Smart Attendance, Open Computer Vision (OpenCV), Facial recognition.

I. INTRODUCTION

Facial recognition is a mechanism that calculates the locations and facial features in digital pictures. It acknowledges faces and avoids such things as structures, trees and bodies. Perception of the face image is actually a vibrant research field within the community of computer vision. In image / video surveillance, facial recognition and image management, human face localization and identification is typically the initial stage. Discovering and tracking human faces can be

important for face detection. The library is supported across many platforms. It concentrates mainly on image processing implemented in real time.TheOpenCV library was created in C, making it a lightweight resource for certain platforms, such as digital signal processors. Wrappers for languages such as C#, Python, Ruby, and Java (using Java CV) are being built in order to promote broader adoption. The C++ interface is used in the majority of recent OpenCV inventions and algorithms. The techniques and methodology used in the proposed approach to introduce and test face detection and tracking using OpenCV are detailed.

II. SURVEY STUDY

Khumbhar et cetera all , presented a method in which Haar like features are used to detect faces. The proposed technique captures faces from different angles extracted from the HD video. Simple CV and Open CV are some of the framework libraries that are used along with the Raspberry Pi BCM283 CPU processor.

Abdul Mohsen Abdul Hossen et cetera all, proposed a method with the help of OpenCV’s Viola-Jones algorithm that discards incorrectly detected faces built on coding eyes. +ve and - ve images are taken as input and

http://www.turkjphysiotherrehabil.org/�

mailto:[email protected]�



Adaboostalgorithm is applied which converts weak classifier to strong classifier. The Classifier Cascade process is faster and provides a precision of98.97%.

Mary Prasanna et cetera all, came up with an instantaneous GUI based biometric identification evolve using Open face. The three stages of Open face are: Localization, Feature extraction, and Recognition. HOG is used for altering the size of anterior part of the detected faces . In this all pixels are inspected with adjacent pixels which results in a image with 68 key features.

Kruti et cetera all, presented a method for face detection. It compares every one of the algorithms in terms of computation and determines that Haar cascades are the most effective for face detection. In order to identify the face, the image subtraction morphological process is used. Cascades of different images are captured and stored in the database. All pixels within the white region's influence are subtracted from all pixels within the black region's influence. Each image is subjected to the subtraction method one by one. When all of the images are combined, the result is called a composite image. As a result, it provides better performance and a more reliable output.

Wu-Chih et cetera all, proposed a method which is a three-stage scheme. This was conducted on a computer with an Intel Core. First, skin regions are saved. In the next stage, a box is used to remove other regions. The FERET image database was used from which the faces of different poses and size were selected and was used to evaluate the performance.

Carlos et cetera all, presented a method which is a pupil detection technique for the multiple face detection. The pupil detector used active illumination scheme. This presented a multiple detection device for the face that was cheap and real time, using recursive estimators, together with a single face tracking system. It was developed using a commercial

frame grabber on a dual Pentium unit.

Zahra Sadri Tabatabaie et cetera all, presented a paper aimed at consolidating Viola and Jones face position strategy with a technique based on shading that suggest enhanced strategy for face recognition. Exploratory outcomes show that their strategy proficiently diminished bogus positive rate and therefore expanded exactness of the face recognition framework particularly in complex foundation pictures.

Henry A.Rowley and so on allpresent a neural organization based face identification framework. A retinal linked neural organization looks at a picture's tiny windows and chooses if a face is contained in each window, they utilize a bootstrap calculation for preparing, which adds bogus findings to the preparation set as advances are being prepared. This wipes out the difficult task of individually selecting non-face preparation models that should be selected to traverse the entire space of non-face images. This framework has better execution regarding recognition and bogus positive rates.

Paul Viola and so on all, introduced a paper which a facial recognition system that is capable of handling pictures in a short amount of time while maintaining a high level of quality location appraise. There are three key commitments.

The underlying is the presentation of a new picture depiction dubbed “Indispensable Picture" which enables our highlights to be used identifier to be processed rapidly. The succeeding is a straightforward and proficient The AdaBoost learning algorithm is used to create a classifier that selects a small number of simple visual features from a large number of possible features. Ternary commitment is a method of joining classifiers in a "course" that allows back-ground areas of the image to be discarded quickly while more computation is spent on promising face-like districts.

In the paper “Student Attendance System in Classroom Using Face Recognition Technique “composed by Samuel Lukas, Aditya Rama Mitra, RirinIkanaDesanti, Dion Krisnadi” the author says the quantity of highlights of any facial understudy picture is made to be consistent, for example 16 DCT coefficient. The process is finished entirely performing grayscale standardization, histogram balance, Discrete Wavelet Transform (DWT), and Discrete Cosine Transform (DCT). Further examination of the disappointment in perceiving the rest facial




pictures demonstrates that an understudy might be perhaps perceived as other student(s)By considering the all-out degree of acknowledgment as came about because of the investigation which doesn't meet exclusive requirement.

There are a number of systems for attendance purpose, like traditional methods of data, have drawbacks and hard to use that list, a biometric presence. There is a lack of human error in the system like fingerprint scan is not accepted because of wet conditions Fingers, dirty, very dry or peeled fingers. So, the author proposes Authority to add mobile presence system and face with NFC Safety facility and possibility to store using Raspberry Pi data in the cloud. The paper reviews relatable works. Attendance management system, NFC, face authority area, Microcomputers and Cloud area. Then, it provides new Method and design system and planning. Outcome of this is a system which reduces the usage of paper, ending time and energy wasted by attendance Mobile Based Attendance System.

In the paper" Class Attendance framework the on- Face Recognition" composed by PriyankaWagh. To distinguish the understudies sitting on the last columns conveniently, the histogram leveling of picture should be finished. The picture will be passed for individual's face discovery. The productivity of Ada-Boost calculation is best of all these. In this way, this will paper utilizes this calculation for identifying countenances of understudies by utilizing the haar highlight classifiers and course ideas of Ada-Boost calculation .Every understudy's face is trimmed and the different highlights are removed from them like separation between eyes, nose, blueprint of face, and so forth utilizing these countenances as Eigen includes, the understudy is perceived and by contrasting them and the face database and their participation are stamped. A database of faces should be made with the end goal of examination.

Figure 2: Face Recognition model

Computers are intelligent to communicate with humans in various perspectives. It will be participation If it is based then it is more acceptable for both humans and computers. On the validation process. The author is concerned Integrating and developing a student recognition using survival- Ing algorithm. Then embedding is used in classification of a person's face the system offers a variety of applications such as attendance – Systems, security etc. After making a system, a resulting display, is shown in the paper.

Figure 1: Traditional neural network versus CNN.

The paper Individual Stable Space: An Approach to Face Recognition Under Uncontrolled Conditions by XinGeng, says that most face recognition systems needs the faces to be fed into it based on certain rules, like under a controlled illumination, at a particular position, under a particular view angle, and without any obstacles. Such systems are called face recognition under controlled conditions. These rules restrict the uses of face recognition in many real time applications because they cannot satisfy these rules. Real time applications need techniques which does not need any strict control over the human beings for

recognizing the face. These types of systems need face recognition under uncontrolled conditions. So, this paper proposes one such system but the system needs an image as input and one person per image which is a drawback of the system and provides a hindrance in using it in real time applications like attendance systems.




Table 1: Comparison of different techniques used in face recognition-based attendance system.

Ref No.

Title & Authors Name

Concept Used

Advantages

Disadvantages

1. Individual Stable Space: An Approach to Face Recognition Under Uncontrolled Conditions XinGeng, Zhi-Hua

Zhou, & Smith-Miles

Face recognition under uncontrolled condition.

This paper projects on face recognition under uncontrolled conditions.

Video sequences, verification, and multiple persons per image required by most of the real applications can’t be implemented.

2. Anti-Cheating Presence System Based on 3WPCA

Dual Vision Face Recognition Winarno, Wiwien Hadikurniawat i, Imam Husni Al Amin, Muji Sukur

Dual vision face recognition using 3WPCA.

It can anticipate falsification of face data

with recognition accuracy up to 98%

The relative angle of the target’s face influences the recognition score profoundly.

3. Prototype model for an Intelligent Attendance System based on facial IdentificationR

aj Malik, Praveen Kumar, AmitVerma, Seema Rawat

ADA Boost algorithm with techniques PCA and LDA Hybrid algorithm .

By using this system chances of fake attendance and proxy can be reduced.

Works only for single image of a system.

4. Convolutional Neural Network NusratMubin Ara1. Neural Network Approach for Vision Based Student

Alex NET CNNs and RFID Technology.

Uses the camera system to Monitor the scene information.

Alex NET won’t work on all the students until it is improved. RFID technology uses electronic toy which can’t be used in all the




Recognition System

cases.

5. NFC Based Mobile Attendance System with Facial Authorization on Raspberry Pi and Cloud Server Siti Ummi Masruroh

Iris Recognition.

Real time face detection and efficient

Iris condition needs to improve in different light conditions

6. Facerecognitio n-based Attendance System using MachineLearning Algorithms Radhika C. Damale

Local Binary patterns (Support vector machine), LDA based OpenCV and FLTK.

Continuous and automatic attendance system.

System has issues with system performance and accuracy.

7. ClassAttendance system based on Face Recognition" PriyankaWagh.

Local Binary patterns (Support vector machine), LDA basedOpenCV and FLTK.

continuous and automatic attendance system.

System has issues with system performance and accuracy.

III. METHODOLOGY

DESCRIPTION OF TOOLS :

In this section, we'll talk about the basic algorithm that was used to find the face. The AdaBoost algorithm is discussed first, followed by the collection of functions.

OPEN CV :

OpenCV is a package of highlights of programming that gradually relies primarily on PC vision. The library is stage cross. It is principally focused on picture preparing progressively.

FACE DETECTION :

The base calculation used to differentiate the face is discussed in this section. At that point, the AdaBoost estimate is evaluated first, and the option is addressed.




ADABOOST

It was then generally utilized in example acknowledgment.

The AdaBoostAlgorithm :

1 Input: Give sample set S = (x1, y1),... (x n,y

n)xiϵX,YiϵY={-1,+1},number of iterations T.

2 Initialize:

3 For t = 1, 2,…,T,

i). Use the Wt distribution to train a weak classifier.

ii). Determine the weight( ) training error for each hypothesis.

i). Set:

ii). Reinitialize the weights:

Output:

The stronger classifier, also the final hypothesis.

CAMSHIFT ALGORITHM

A series of steps for face detection are followed by the camshaft algorithm:

a. Acquire the desired output.

b. Create a histogram and the section that should be searched for initially.

c. Find the pixelated zone with the most populated.

d. Determine the central location of the whole are a being searched.

e. Transfer the searched region's centre to the entire region's centre.

f. Switch back to phase d, if going forward needs to be converted.

g. The new search window is now created, with the centre originating from the previous centre.

h. Find out the new processing area.

i. New window’s beginning is chosen.

HAAR CASCADES :

Haar cascades makes use of the image subtraction morphological process to detect the face. In this the cascades of different images of the same person is taken and recorded in the database. All the pixels in the influence of




white region are subtracted from all the pixels in the influence of black region. This method of subtraction is performed on each of the image in the cascade but all the images might not give us the best results. Many of the images have a lot of errors. The image with the least error is selected. The result of all the images are added together and is mentioned as weak classifier. As all the week classifiers are added together to form a strong classifier. Appling the subtraction process and determining each image error is a very time and space consuming process. Instead of applying it on each of the images the subtraction is applied on images one by one. If the last image is not a useful it is discarded.

FINDING FACES :

Finding faces is the most integral part in face detection. There are various techniques from which faces could be found. In this paper, we will be comparing the various algorithms used previously by implementing them and analyzing them. Even though face detection is the most important step in the area of image processing still the techniques used for its implementation need to be reworked upon so as to optimize its performance and bring down the hurdles it is cladding. The working of face detection algorithms is majorly based on the accuracy of the face detection, due to this face detection is the cusp point in the entire process of face detection and tracking.

Finding faces via color :

A. Images having a definite background.

One procedure is to find images in which we have a monochromatic background containing only grey scale pixels. These images have wavelength of narrow band. While using these images when we eliminate the background from the foreground we get the boundaries of the face. This is by far the easiest technique used for face detection.

B. Images having a colored background.

The system in which a face is to be detected for colored images is based on two procedural steps.

By applying a skin filter :Skin filter is applied for skin detection. The skin filter processes and defines the texture of image part that is being masked. After the initial masking the output generated contains discrete areas of human skin. Various morphological techniques such as dilation and erosion are used to develop this kind of filter.

2. By hauling out the features which are being masked. During this step the various portions of the image which are very dark or very bright are subtracted from the image. These include the area around the eyes, area under the nostrils or the area below the mouth. By subtracting these portions, we get the most relevant area which is covered by skin and effective detection can be done. The major problem during this process is the variable light sources at the time of face detection by the camera. Now it is not feasible to place all the cameras in the presence of sunlight, some might have been placed at low light source region. These cameras can be used for application only if they are placed according to need and not according to environment. To solve this problem, we make use of a skin color model proposed by outski and Healey [4]. The input should be in RGB format with caricature




intensity values ranging from 0 to 255.The RGB matrices are "zeroed" to supposedly act like a wet blanket desaturation when the image is improved from RGB enlarge space to IRgBy emphasize space. The smallest intensity outlay greater than 10 pixels from entire edge in barring.

C. Images having a complex background

Face detection for faces present in the complex background can be done using MUHULANOBIC metric. It is based on the examination of human faces in a two-dimensional natural scene image. It distinctly makes use of the process of color segmentation of the image that is being taken as the input. This segmentation of the color is being performed by Thresholding the picture in the hue color space. This takes into account the effects of distinctiveness of the color in the human skin when the lighting has been changed in the image. Then the results of the image are accumulated together for any other examination. This is followed by median filtering of the resultant image using a limited number of accumulations of pixels. At the end the difference between the faces and the remaining complex background, a multi-layer perceptron neural network is used with the invariant moments as an input factor.

FACE TRACKING:

To test the consistency of the various face indicators, we created a connection-based single face tracker that selects the most striking face to pursue. This face is followed until it is lost. During following, the framework continues to distinguish different appearances, however it doesn't respond to them regardless of whether they become more striking than the followed face.

It depends on two operation modes to keep following when the face tracker is implemented. The data from the multi-face identifier is used in one mode, and the second

is a component relationship tracker that uses the sum of total contrasts (SAD)as the capacity of the item to be restricted.

Two boundary vectors describe the state of each recursive assessor. A covariance joins each vector. The predicted state vector can be seen as the last refreshed gauge vector.

X = X

And the projected matrix of covariance is

C = C^ + (t)2W

Where, with the time interval t between the observation and the last estimate, the uncertainty in position and size grows quadratically and W notes the precise dropping of each variable and depends on the characteristics of the under lying mechanism.

The condition of the estimators is modified using new face observations (Y; C) as follows:

C^ = [ C1 + C1

X^ = C^ [ C 1X + C]

The covariance matrix C^ is an estimation of the error of the estimated state vector X^ .

It is unimaginable to expect to get perceptions from the various face indicator for each casing on account of squinting and disappointments in the gathering cycle. At the point where no face is detected, or no face closer than a certain size and location to the expected face is distinguished, the SAD relation tracker is called to determine the 2D meaning of the face last used as measurement. In order to update the place assessor, the interpreted face is the n used as the new estimate.

A small aid and search window around the left understudy are used by the SAD link tracker, so it can monitor without any problems. To try not to track of non-face objects, if the SAD relationship tracker gets called




sequentially for in excess of a specific number of edges, without a face being identified inside the anticipated locale by the different face identifier, the cycle is re- introduced with the following generally remarkable.

FACE DETECTION AND DESIGN ANALYSIS :

Face recognition involves three basic steps which include face detection, face recognition, and face extraction. Any system needs to encapsulate the image and then manage as well as record the vital features to determine the location of the face. For the recognition of the captured image it keeps records of various features such as skin color, skin tone etc.

Face recognition based on Viola and Jones :

The fundamental rule of this estimation is to filter through a given picture of information on a sub-window fit to distinguish faces. The standard approach to image handling will be to downscale or upscale the image of data to various sizes, and then run the fixed-size indicator through these images. This procedure winds up being fairly dull because of the calculation of the unmistakable size pictures. Unlike the traditional technique, Viola-Jones re-scales the finder with all things equal to the info image and usually runs the finder each time with a different scale, through the picture.

This identifier is made up of a so-called basic picture and some simple rectangular highlights that resemble Haar wavelets. This indicator expounds the following area.

The underlying advance in figuring this face region is to get a simple image from the image of knowledge. This is accomplished by setting each pixel to the same value as the total quantity, all called things, over and on one side of the pixel concerned. This permits to do the count of the amount of all pixels inside any given square shape utilizing just four qualities. These qualities are the pixels in the essential picture that match with the corners of the square shape in the information picture.

Sum of gray rectangle = S - (Q + R) + P

Fig. Sum estimation

The V&J face locator investigations a given sub-window utilizing highlights comprising of at least two square shapes.

Fig. Different types of features

A piece factor generates a solitary value estimated by eliminating the amount of the light-coloured rectangle(s) from the amount of the dark coloured rectangle (s). They have precisely discovered that appropriate results are given by an identifier with a main objective of 24x24 pixels.

They used a simple and competent classifier that worked from computer-skilled highlights using AdaBoost to provide preference. AdaBoost is a AI boosting calculation prepared to do developing a solid classifier through a weighted blend of feeble classifiers. To coordinate this wording to the introduced hypothesis each component is viewed as an expected feeble analysis. A feeble analysis is numerically portrayed as:




Where x is a sub-window of 24*24 pixels, f is the factor that is added, p is the extremity and x is the limit that chooses whether x should be called a positive (a face) or a negative (a face) (a non-face).

The fundamental rule of the measurement of the Viola-Jones face position is to normally check the finder with a similar image, each time with another size. Regardless of whether a picture ought to contain at least one faces it is evident that an unreasonable huge measure of the assessed sub-windows would at present be negatives(non-faces).This acknowledgment prompts an alternate plan of the issue:

Rather than discovering faces, the calculation should dispose of non-faces.

The idea behind this assertion is that it is quicker to dispose of a non-face than to discover a face. With this in mind an indicator comprising of just one (in number) classifier out of nowhere appears to be wasteful since the assessment time is steady regardless of the info. Thus the requirement for a fellclassifier emerges. The strong classifier is included in each stage of the fell classifier.

Each stage's job is to figure out whether or not a certain sub-window is a face.It is easily discarded exactly when a sub-window is demanded by a given process to be a non-face Then again, the accompanying stage in the course is given a sub-window called a maybe face. It implies that the more levels a provided sub-part goes through, the greater the probability that a face is really included in the sub-window.

Fig. The Cascade Classifier

Feature-based Face Detection :

Scientists have been trying to find invariant characteristics of disclosure appearances in element-based methodologies. The basic principle that relies on the discernment that people will certainly identify appearances as well as items in different positions and lighting and that characteristics or features should consequently exist that are invariant over these abnormalities. In order to first identify facial highlights and then to stimulate the appearance of a face, different strategies have been suggested. A measurable representation is used to represent their associations and to check the appearance of a face in light of the removed highlights. Skin shading highlights will be analyzed and included in this article.

Skin Color Classification:The investigation on skin tone characterization has acquired expanding consideration in later a long time because of the dynamic examination in substance-based picture portrayal. For example, for image coding, changing, ordering or other client intelligence purposes, the ability to find picture objects as a face may be misused. Inaddition, face restriction also offers a decent venturing stone in external appearance feedback.

It is reasonable for express that the most main stream calculation to face restriction is the utilization of shading data, whereby assessing zones with skin tone is frequently the principal imperative advance of such methodology. Thus, skin shading characterization has become a significant undertaking. Most of the study of face limitation and exploration based on skin shading relies on RGB, YCbCr and HSI shading spaces. This study utilizes RGB shading area for skin shading order.

The speed of the RGB colors is a 3-D shading area whose segments are the powers of R,G& B that form a particular sound. The tones red, green, and blue were picked in light of the fact that every one compares generally with one of these three kinds of shading delicate cones in the human eye. It is quite possibly the most generally utilized color spaces for handling and putting away of advanced picture information.

Hybrid Face Detection :

Mixing an appearance-based and highlight- based face position structure is our suggested approach to enhancing face recognition frameworks. Because of fast and exactness in identifying faces in a picture, we picked Viola




furthermore, Jones proposed face location framework. In this outlook, we consolidated it with a (skin) based upon shading strategy to enhance the face position speed rate and minimize bogus positive rate. This methodology is extremely basic however it functions admirably.

Fig. – Overall construction of our technique

One strategy to fabricate a skin classifier is to characterize expressly (through various guidelines) the limits skin group in some shading space.

In the given methodology, initial skin districts within info picture are recognized utilizing previously provided strategy also afterward their calculation is used for distinguishing faces. All non-skin areas are replaced by dark in the wake of the application of the skin shading classifier, while skin districts remain set. This encourages face identification calculation to rapidly distinguish non-faces which incorporate lion's share pixels of each picture. Like wise this technique effectively decreases bogus positive rate. Before and after applying our hybrid algorithm, the figure illustrates the effects of face detection.

(a) Originalimage

Faces recognized by V&J face detector




Faces recognized by the hybrid approach

Fig.- Results of face recognition with V&J and hybrid approach

In this methodology, first skin districts in the information picture are distinguished utilizing previously mentioned strategy and afterward V&J calculation is used for identifying faces. Subsequent to applying skin shading classifier, all non-skin locales supplant with dark, while skin districts stay fixed. This causes face recognition calculation to rapidly distinguish non-faces which incorporate greater part pixels of each picture. Moreover this strategy productively lessens bogus positive rate. The after-effects of face recognition before and then after applying our crossover estimate are shown in Figure.

Face detection: With the end goal of investigation and ID of the face, this gadget can recognize the appearances from the caught picture from HD Video. The Prototype is planned with OpenCV'sHaar-Like Feature work. To assemble an inquiry window that slides through a picture, Haar classifier face recognition is used to check whether a specific area of a picture resembles a face or not. To group a specific picture as face or non-face, haar-like highlights and a wide assortment of frail classifiers utilize a solitary capacity.

Figure: Haar-like features-based decision tree (Cascade of classifier)

The pursuit window effectively checks the main classifier on the course, on the off chance that the classifier returns bogus, the figuring on that window also ends at that point, leaving no identifiable face (bogus). Furthermore, in the event that the classifier returns valid, at that point to do exactly the same thing, the window is passed down to the following classifier in the course. On the off chance that all classifiers return valid for that window, the outcome returns valid too. This detects a specific window face.

Software Required: For Windows, there is a super pack called OpenCV 2.3.1.

OPEN CV V/S MATLAB:

A. Speed




Since, MATLAB is fabricated from java which in turn is fabricated from C. Therefore, when a code is scripted and run-on MATLAB, the computer initializes by interpreting the code and converting it into java and then finally executes the script. Whereas, open CV uses c/c++ library functions. Which directly provides the computer with the machine language code and hence helps in faster execution.Using OpenCV results in more utilization of time and resources in image processing and less in interpreting.

B. Portability

As OpenCV runs on C, therefore any device which runs on C can run OpenCV. It can toil well with Windows, mackintosh or Linux.

C. Cost

MATLAB is much more expensive than OpenCV. MATLAB costs around USD2150 whereas, OpenCV is Free of cost. Even the base MATALAB is expensive as it has commercial, single user License. And OpenCV is a BSD license so it is free of cost.

Equipment Required:

Webcam: The USB port on the webcam connects to the device. A number addressing its related USB port will be used to describe it in the code.

Implementation:

It appears to be added to a locale of interest (of a comparable size as used during the preparation) in an information picture after a classifier is prepared. The classifier yield is "0" if the area is likely to show the face and "1" in any case.

Figure: Extended Haar-like Features

You can get the inquiry window around the image to look for the article in the whole picture and search each region using the classifier. For face recognition, we use two different codes and have a separate follow-up procedure. The formula used for the codes for both(Processing and Raspberry Pi).

IV. IV.RESULT AND ANALYSIS

The figure depicts the after-effects of facial exploration. Those are the edges of the HD video isolated from the actual time. Face area figuring will every so often get more than one answer, even though there is just one face on the edge. A post image preparation was used for this scenario with OpenCV and SimpleCVHaar Classifier libraries to eliminate the unique face facilitates. On the off chance that the framework yield gives more than one square shape, which demonstrates the situation of the face, the distance of focus purposes of these square shapes has been determined. The normal of these square forms will be figured and set as the last situation of the distinguished face, on the off chance that this gap is more modest than a pre-set edge. We accomplish the face following application in Python in this paper by using face recognizable evidence in the same way. This




procedure is proven, so by analyzing and troubleshooting our codes, the flaws of the plan are seen. Furthermore, we switch to OpenCV to evaluate the speed of this face after arrangement after a while, confined by Python execution. We discovered that the recognition of the Viola and Jones faces is more appropriate for continuous face discovery because they require less CPU resources and take less time.

Fig. Algorithm performance showing face detection

V. CONCLUSION

Facial recognition is the most popular field in the area of computer vision study, it is this technology that is used in most applications for detection of faces in soft images. From all the above papers, it is evident that the OpenCValgorithm is most efficient and suitable for our project. Open CV has many advantages like c/c++ library functions is been used, so machine learning code is directly provided to the computer and it will be executed faster. OpenCV provides additional edge over others and as it is a Berkeley Software Distribution's license it is cost free.

REFERENCES 1 Abdul Mohsen Abdul Hossen, RaheemAbdAlsahebOgla, MaithamMahmood Ali (2017). Face Detection by Using OpenCV’s Viola-Jones Algorithm

based on coding eyes. 2 D. Mary Prasanna 1, Ch. Ganapathy Reddy 2 (2017). Development of Real Time Face Recognition System UsingOpenCV. 3 KrutiGoyal, KartikeyAgarwal, Rishi Kumar (2017). Face Detection and Tracking Using OpenCV. 4 Wu-ChihHu1 ,Ching-Yu Yang1 , Deng-Yuan Huang2 , and Chun-Hsiang Huang1 (2011). Feature-based Face Detection Against Skin-color Like

Backgrounds with Varying Illumination. 5 Prof. P Y Kumbhar , Mohammad Attaullah , ShubhamDhere , ShivkumarHipparagi (2017) . Real Time Face Detection and Tracking Using OpenCV. 6 Paul Viola, Michael J. Jones (2004). Robust Real- Time Face Detection. 7 Henry A.Rowley,ShumeetBaluja,TakeoKanade. Human Face Detection in Visual Scenes. 8 Carlos H. Morimoto, Myron Flickner. Real-Time Multiple Face Detection Using Active Illumination. 9 Mohammed Abdel-Mottaleb (2002). Face Detection in Color Images. 10 Zahra Sadri Tabatabaie, RahmitaWirzaRahmat, NurIzuraBintiUdzir and EsmaeilKheirkhah (2009). A Hybrid Face Detection System using

combination of Appearance-based and Feature- based method. 11 Face Recognition Based Attendance System Nandhini R, Duraimurugan. 12 Protype model for an Intelligent Attendance System based on Facial Recognition by Raj Malik, Praveen Kumar, AmitVerma, SeemaRawat, Amity

University Uttar Pradesh. 13 Class Attendance System based on Face Recognition, PriyankaWagh. 14 Convolutional Neural Network Approach for Vision Based Student Recognition System, NusratMubin Aral, Dept. of CSE, SUST, Sylhet, Bangladesh. 15 XinGeng, Zhi-Hua Zhou, Smith-Miles, K.(2008).Individual Stable Space: An Approach to Face Recognition Under Uncontrolled Conditions.


SMART ATTENDANCE USING OPENCV FACIALRECOGNITION

Documents

Transcript of SMART ATTENDANCE USING OPENCV FACIALRECOGNITION