Validação de instrumento de medida de composição de carcaça, composição tecidual de cortes comerciais e qualidade de carne a partir de vídeo image analisys de carcaças de cordeiros.

Abstract

The mean of the paper it was investigate if the vídeo image analisys (VIA) from cold carcass, provide a consistent description of the meat quality, whereas conposition tissues from carcass and primal cuts. Information of 67 cold carcass from lambs male nesteres was submited on the regularized generalizeted canonical correlation (RGCCA) and the structural modeling equation by PLS-PM, assessing blocks of the manifest variables (MV’s) of: SHAPE from carcass (SHAPE), meat quality (QUALI_MEAT), tissue carcass composition (TISSUE_CARCASS), and the primal cuts (TISSUE_PRIMALCUTS), where it block was trated by a latent variable (LV’s). The teoric models diverged from the getting of the form characthers: by morphometric (SHAPE_MPH) and By VIA (SHAPE_VIA1 and SHAPE_VIA2). The models it was availabels to predict the of characteristic of the loss cooking and the Shearing force. All the models attended the criterion of convergente validity, composite realibity, predictivy validity and size of effect, demostrating high aacuracy on the prediction of the LV’s specialy of the QUALI_MEAT: 0.77, 0.82 and 0.78, and SHAPE_MPH, SHAPE_VIA1 and SHAPE_VIA2 respectivily. The models of VIA promoted biggers coeficientes of determination from the morphometric avaliation on the carcass in situ, except for the TISSUE_PRIMACUTS. The path coeficientes of the relation Shape relation SHAPEQUALI_MEAT, were not significantily for the models SHAPE_MPH AND SHAPE_VIA1 was considerated suitable and validated by the procedures of quality of adjustment for the structural equation modelation. The model SHAPE_VIA1 demostrated positive and high correlation between the LV’s: SHAPE_VIA1 and the TISSUE_CARCASS (r = 0.88), SHAPE_VIA1 and TISSUE_PRIMALCUTS (r= 0.81) and negative batween QUALI_MEAT and TISSUE_PRIMALCUTS (r=-0.86), QUALI_MEAT and TISSUE_CARCASS (r=-0.87), represented meat and fatness, greater weight and edible portion of comercial cuts, respctively. Lower scores for QUALI_MEAT corresponding to juicy and tender meat. A latente grouping analisys was performed from the LV’s scores, forming four cluster, where cluster 1 presented the higher scores of SHAPE_VIA1, TISSUE_CARCASS, TISSUE_PRIMALCUTS and lower scores for QUALI_MEAT. Cluster 4 groups with the lower scores of SHAPE_VIA1, TISSUE_CARCASS, TISSUE_PRIMALCUTS and higher scores for QUALI_MEAT. Cluster 2 and 3 where the groups with intermediate scores for all LV’s. The proposed and validated instrument allowed, through the geometric profiles obtained by image processing, to obtain more efficient associations with quantitative, tissue and qualitative characteristics of the carcass, 33 demonstrating interrelations between the groups of descriptors of form and composition, allowing, through the latent variable scores, establish categories for carcass classification. Thus, it is presented as a consistent methodology to be incorporated into computer vision systems capable of operationalizing such measurements in the carcass image in an instantaneous way, providing categories of interest to the production chain of the cutting sheep.