Age Classification for work sustainability using SVM using Co-occurrence features on Fibonacci Weighted Neighborhood Pattern Matrix

: Computer vision systems are increasingly focusing on age recognition from facial images. To solve this problem, In this paper, proposed a method that computes the Fibonacci Weighted Neighborhood Pattern on an image to obtain local neighborhood information, then evaluates Co-occurrence features for work sustainability age classification with SVM classifier. These characteristics show how people's ages differ. The proposed method has been tested on the FG-Net facial images dataset as well as other scanned images. Experiments showed that the proposed approach outperformed other currently existing methods.


Introduction
Age classification plays a key role in classifying facial images.In the past two decades, age classification of people from facial photographs has gained prominence due to applications such as human-computer interaction, electronic customer relationship management, video surveillance tracking, and others.Age classification techniques can be divided into three categories.Anthropometric models, aging pattern subspaces, and age regression categories make up this trio.A wrinkling analysis of the face and the Cranio-facial development model are both included in the anthropometric model.The Active Appearance Model (AAM), a technique for age regression, is used to extract facial elements linked to shape and appearance.The estimation of age using the texture feature [1,2], the contour and texture features [3,4] Recent research has focused on LBP models, which represent spectral patterns in ECG data.Various age classification and grouping systems have recently been developed with LBP models, spectral patterns in ECG data, and by Sakthidharan et.al.[5], Reddy et al. [6,7] and other researchers [8,9,10].Research has shown that individuals of different ages hold different work attitudes [11].The current paper considers two age categories for age classification, child, and adult, in order to discuss this study in age grouping.The present paper evaluates Co-occurrence features on FWNPM of facial images and uses SVM Classifier.The rest of the manuscript is organized as follows: section 2 describes the methodology, the results are discussed in Section 3 and conclusions are made in section 4.

Methodology
The proposed method converts a colour image to a grey image, computes the Fibonacci Weighted Neighborhood Pattern (FWNPM), and then evaluates co-occurrence features for age classification as shown in figure 1. Step -1: Image conversion from colour to gray scale The colour image is converted to the grayscale image by removing the hue and saturation characteristics while preserving the brightness information.
Step-2: Fibonacci Weighted Neighborhood Pattern Matrix LBP is a two-dimensional approach to representing surface textures that includes two complementary measures: local spatial patterns and grayscale contrast.By thresholding the 3 x 3 neighborhood of each pixel with the mean and looking at the output as a binary integer, the original LBP generates labels from 0 to 255 for image pixels.
The FWNPM is used to extract local neighborhood information from pixels in an image [6].In Our approach, each pixel in the image produces FWNPM, which can be used to represent local pixel information using an integer code spanning from 0 to 54, reducing the number of permissible encoding options by 20%.An example of how to compute this pattern is shown in Figure 2. P = pc, p1,..., p8, where pc is the central pixel intensity value and pl (1<=l<=8) is intensities of surrounding pixels, makes up a 3x3 neighborhood.A set of binary values dl (1<=l<=8) can then be used to describe each of the image's 3x3 neighborhoods, as shown in equation 1. .
The equation 2 is used to calculate FWNPM for each of the 3x3 neighborhoods.
where w that not all characteristics are necessary for classification.Subsequently, numerous authors experimented with a subset of the Haralick features in image processing applications.
Our method assesses the contrast, correlation, energy, and local homogeneity of the feature set on FWNPM.These features make these methods optimal for many face detection, age group, and other image processing applications.These characteristics are calculated using Equations 3 through 6 on FWNPM in four directions for successful age classification: 0, 45, 90, and 135.
Energy = ∑ −ln (P ij ) 2  (4) Local Homogenity = ∑ P ij 1 + (i − j) 2  (5) Step 4: Support Vector Machines Support Vector Machines are supervised machine learning algorithms that can be adopted for classification and regression.They have lately gained popularity owing to their capacity to handle large numbers of continuous and categorical variables.In multidimensional space, an SVM model is a portrayal of various classes in a hyper plane.SVM builds the hyper plane iteratively to lessen the error.The objective of SVM is to partition datasets into bins to find the best boundary hyper plane (MMH).

Results and Discussions
The proposed method created a database of 1002 face images from the FG-NET database and 600 scanned photographs images, as shown in Fig. 3.The features contrast, correlation, energy and homogeneity are extracted from the Fibonacciweighted neighborhood pattern matrix of the database images under consideration and the results are stored in the feature vector.The training images are represented using the feature set.Tables 1, 2, 3, and 4 show the derived four characteristics on five facial images in four directions with 0° 45°, 90°, and 135° orientation of FWNPM.For age classification, these 16 features are used as feature vectors.These features, together with the SVM Classification algorithm, resulted in a 99.2% accuracy rate in age classification.Proposed method obtains state of the art results outperforming all previous methods.The results of this method are in agreement with our expectations.Table 5 and Figure 4 show a comparison of the proposed system and other age classification methods.

Conclusion
The proposed FWNP collects local textured details from a facial image to reduce the dimensionality of the image and the computational cost.Co-occurrence features are evaluated on FWNP.These features, along with the SVM classifier, are employed to effectively distinguish between two age groups for work sustainability.This method has been found to be simple and efficient.The proposed method demonstrated a high classification rate when compared to other current methods.Future research should be devoted to the development of methods with shape features.

Table 5 .
Analysis of the proposed method in comparison to existing methods.