Prediction of beef eating qualities from colour, marbling and wavelet surface texture features using homogenous carcass treatment

摘要: Colour, marbling and surface texture properties of beef longissimus dorsi muscle are used in some countries to grade carcasses according their expected eating quality. Handheld VIA systems being augment the grader assessments, however attempts have been made develop higher resolution image give consistent objective predictions quality based on these properties. Previous efforts unable model sufficiently variation A new approach has applied whereby were subjected homogenous post-slaughter treatment minimize related other factors such as chilling temperature hanging method. Furthermore a wider range features better characterize colour wavelet transform was texture. Objective sensory panel tests performed evaluate qualities. Classical statistical methods multilinear regression (MLR) partial least squares (PLSR) predictive models. It possible explain greater portion than before (up r^2=0.83). Carcasses classified high or low with rate correct classifications (90%). Genetic algorithms select subsets.