ニューラルネットを利用した肉牛の脂肪交雑値推定

概要

論文の詳細を見る
Up to the present time, estimation of Beef Marbling Standard (BMS) number based on ultrasound echo imaging of live beef cattle has been studied. However, it is difficult to establish the objective and high accurate estimation method. Therefore, this paper proposes a novel modeling technique based on a neural network to estimate the BMS number. The proposed method consists of three process steps: the extraction of texture features, principal component analysis, and the estimation of BMS number by the neural network. The neural network can be expected to model the non-linear mapping between the texture features and the BMS numbers. In the verification test with 27 live beef cows, the proposed method achieved high estimation performance. The correlation coefficient between estimated and actual BMS numbers was r=0.88 (P<0.01) by leave-one-out method. On the other hand, the correlation coefficient by conventional multple regression analysis was r=0.51(P<0.01). These results showed that the proposed method was effective in non-linear modeling between the texture features and the BMS numbers.
公益社団法人計測自動制御学会の論文