Swine live weight is an important aspect in the production of pork products and also to Stockmen,regarding market costs,feed conversion,and animal health.The determination of animal live weight has been a long-time research topic that has not yet been thoroughly exhausted due to stressful applications,time-consuming,and inconsistency in accuracy.The objective of this study was to develop a contactless,stress-free method of swine live weight estimation by machine vision technology together with machine learning systems.This novel approach was based on high definition image processing techniques for features extraction,Adaptive Neuro-Fuzzy Inference System and linear regression for modeling,testing,and results comparisons.Firstly,two models were developed:a linear regression model and Adaptive Neuro-Fuzzy Inference System model.The linear regression model established a correlation of all the inputs(extracted features)to the output(live-weight).The Adaptive Neuro-Fuzzy Inference System model was developed by determining which input variables combination(combination 2,3 and 4 of the inputs)holds the highest predictive ability,and used the feature conjunction with the best predictive power to correlate to live-weight.Secondly,for both models,their Root Mean Square Error was analyzed,it was established that the linear regression model had a higher error of 2.067 compared to 0.895 for the Adaptive Neuro-Fuzzy Inference System model.The linear regression model takes into account all the eight inputs while Adaptive Neuro-Fuzzy Inference System used only two best predictors as the inputs.Finally,the Adaptive Neuro-Fuzzy Inference System model was used to estimate the mass of 20 pigs used as the testing dataset;it was established that the average relative error of the proposed system was about 3%and a standard deviation of 0.7%.Thus,development of a practical imaging system for swine live weight estimation by the proposed method is feasible,accurate and fast. |