oalogo2  

AUTHOR(S): 

Ibrahim El-Henawy, Hazem. M. El bakry, Hagar  M. El Hadad, Nikos Mastorakis

 

TITLE

Bovines' Texture Feature Extraction Based on Discrete Wavelet Transform

pdf PDF

KEYWORDS

bovine classification; average filter; Median filter; Discrete Wavelet Transform (DWT) algorithm; Naive Bayes and Decition tree.

ABSTRACT

Animals especially bovines and its products growth very quickly every day, so Ministry of Animal Resources pay great attention and efforts for saving bovines' products. The main goal of this work is to identify and clearly distinguish between huge different bovines and achieving high accuracy rate. This work proposed two different techniques in the last phase which is the classification phase based on Naive Bayes and decision tree. The complete proposed two models divided into four phases namely; data base acquisition phase, muzzle image pre-processing phase, muzzle texture feature extraction and bovine classification phase. Data base acquisition phase is defined as the process starting with pressing down the camera button until saving the muzzle images in the hard memory. The muzzle images database consists of fifty-two different bovine. Pre-processing phase consists of linear filter and Non-linear filter for noise removing from muzzle images. The third phase is the important phase which based on Discrete Wavelet Transform (DWT) algorithm for muzzle texture feature extraction. The forth phase consists of Naïve Bayes in the first proposed model and used decision tree in the second proposed model. The experimental result accuracy rate denote that in case of using fifty-two different bovines group is 75.09% in case of decision tree model and 72.45% in Naive Bayes model. This accuracy rate is excellent with comparing with authors previous work where the accuracy rate in the fifty-two different group and based on box-counting for feature extraction and Artificial Neural Network (ANN) is 14%.

Cite this paper

Ibrahim El-Henawy, Hazem. M. El bakry, Hagar  M. El Hadad, Nikos Mastorakis. (2016) Bovines' Texture Feature Extraction Based on Discrete Wavelet Transform. International Journal of Veterinary Medicine, 1, 5-15

Animals especially bovines and its products growth very quickly every day, so Ministry of Animal Resources pay great attention and efforts for saving bovines' products. The main goal of this work is to identify and clearly distinguish between huge different bovines and achieving high accuracy rate.  This work proposed two different techniques in the last phase which is the classification phase based on Naive Bayes and decision tree. The complete proposed two models divided into four phases namely; data base acquisition phase, muzzle image pre-processing phase, muzzle texture feature extraction and bovine classification phase. Data base acquisition phase is defined as the process starting with pressing down the camera button until saving the muzzle images in the hard memory. The muzzle images database consists of fifty-two different bovine. Pre-processing phase consists of linear filter and Non-linear filter for noise removing from muzzle images. The third phase is the important phase which based on Discrete Wavelet Transform (DWT) algorithm for muzzle texture feature extraction. The forth phase consists of Naïve Bayes in the first proposed model and used decision tree in the second proposed model. The experimental result accuracy rate denote that in case of using fifty-two different bovines group is 75.09% in case of decision tree model and 72.45% in Naive Bayes model. This accuracy rate is excellent with comparing with authors previous work where the accuracy rate in the fifty-two different group and based on box-counting for feature extraction and Artificial Neural Network (ANN) is 14%.