תפריט נגישות
ניגודיות עדינהניגודיות גבוהה
מונוכרום
הדגשת קישורים
חסימת אנימציה
פונט קריא
סגור
איפוס הגדרות נגישות
להורדת מודול נגישות חינם תפריט נגישותניגודיות עדינהניגודיות גבוההמונוכרוםהדגשת קישוריםחסימת אנימציהפונט קריאסגוראיפוס הגדרות נגישותלהורדת מודול נגישות חינםRoii Spoliansky, Yael Edan, Yisrael Parmet - Dept. of Industrial Engineering and Management, Ben-Gurion University of the Negev
Developing an automatic body condition scoring (BCS) system is an important management tool for dairy farms, currently it is determined manually. In this research, a classification system using fuzzy logic was built to improve classification of existing computer vision algorithms. The existing computer vision system matches a body condition score to a set of Fourier descriptors, which represent the outline of the cow. Data of 151 dairy cows were collected in the ARO research farm in Bet-Dagan during 2011-2012. Model performance was evaluated by comparing BCS references given by an expert, using the average and median error of the model's evaluation and the number of correctly classified cows. The model was developed using statistical and visual analysis of the data to derive the number and type of variables, membership functions, rules and optimization methods, which will be presented in the paper. Minimizing the objective function, with suitable constraints, was performed in Matlab’s constrained nonlinear multivariable minimization algorithm. The fuzzy logic classification systems resulted in significant improvement in the training data as compared to a system with simple rules, although some cases resulted with decreased performance for the testing data. The best system resulted in 51.6% correct classification for the testing data set, with an average error of 0.389 and median error of 0.288, as compared to previous results, which resulted in 46.8% correct classification (average error – 0.343; median error – 0.315). Dividing the data into 4 groups according to BCS shows the algorithm ability to cope with extreme values. For obese cows with BCS classes of 4-5, the classifier detected 100% of the images correctly compared to 0% in previous studies. For cows with BCS classes between 3 and 4, the classifier detected 19% compared to 33.3%. For BCS classes between 2 and 3, 62.2% were correctly classified compared to 56.8% (with non-fuzzy logic classification), and for cows with low BCS (classes 1-2) both algorithms detected 100%. Classification of BCS using fuzzy logic has proved itself as an effective method and as less sensitive to extreme values.
International Conference of Agricultural Engineering, Zurich, 06-10.07.2014
Roii Spoliansky, Yael Edan, Yisrael Parmet - Dept. of Industrial Engineering and Management, Ben-Gurion University of the Negev
Developing an automatic body condition scoring (BCS) system is an important management tool for dairy farms, currently it is determined manually. In this research, a classification system using fuzzy logic was built to improve classification of existing computer vision algorithms. The existing computer vision system matches a body condition score to a set of Fourier descriptors, which represent the outline of the cow. Data of 151 dairy cows were collected in the ARO research farm in Bet-Dagan during 2011-2012. Model performance was evaluated by comparing BCS references given by an expert, using the average and median error of the model's evaluation and the number of correctly classified cows. The model was developed using statistical and visual analysis of the data to derive the number and type of variables, membership functions, rules and optimization methods, which will be presented in the paper. Minimizing the objective function, with suitable constraints, was performed in Matlab’s constrained nonlinear multivariable minimization algorithm. The fuzzy logic classification systems resulted in significant improvement in the training data as compared to a system with simple rules, although some cases resulted with decreased performance for the testing data. The best system resulted in 51.6% correct classification for the testing data set, with an average error of 0.389 and median error of 0.288, as compared to previous results, which resulted in 46.8% correct classification (average error – 0.343; median error – 0.315). Dividing the data into 4 groups according to BCS shows the algorithm ability to cope with extreme values. For obese cows with BCS classes of 4-5, the classifier detected 100% of the images correctly compared to 0% in previous studies. For cows with BCS classes between 3 and 4, the classifier detected 19% compared to 33.3%. For BCS classes between 2 and 3, 62.2% were correctly classified compared to 56.8% (with non-fuzzy logic classification), and for cows with low BCS (classes 1-2) both algorithms detected 100%. Classification of BCS using fuzzy logic has proved itself as an effective method and as less sensitive to extreme values.
International Conference of Agricultural Engineering, Zurich, 06-10.07.2014
