Predicting carcass cut yields in cattle from digital images using artificial intelligence

Prediction of primal and retail cut weights, tissue composition and yields of youthful cattle carcasses using computer vision systems; whole carcass camera and/or ribeye camera

The application of two computer vision systems (CVS) was evaluated to predict primal and retail cut composition in youthful beef carcasses. Left carcass sides from a total of 634 animals were broken down into primal cuts, scanned using dual-energy x-ray absorptiometry for the prediction of tissue composition and fabricated into retail cuts. Cold carcass camera (CCC) images led to higher R² values than hot carcass camera (HCC) images. The CVS coefficients of prediction for the primal cut weights ranged from 0.61 to 0.97. For the primal cut tissue composition predictions, R² values ranged from 0.09 for Brisket HCC bone prediction to 0.82 for Chuck CCC fat prediction. Retail cut weight estimations had similar R² values, ranging from 0.10 for IMPS 112 (Ribeye roll-denuded ribeye) to 0.99 for IMPS 113C (semi-boneless chuck) both using CCC. The results suggest the feasibility of CVS technologies to predict beef primal and retail cuts weights together with tissue composition, and yield percentages.

In Vivo Ultrasound Prediction of the Fillet Volume in Senegalese Sole (Solea senegalensis)

Simple Summary

The ability to obtain in vivo information on characteristics related to fish composition is necessary for aquaculture. In addition, there is growing interest in production traits, such as growth, feed efficiency or fillet weight, but it remains difficult to precisely record in vivo individual fish traits that report to these production traits, which can increase edible fish meat production and decrease the environmental impact. In the present study, we performed an ultrasound approach for the in vivo prediction of fillet volume of the Senegalese sole (Solea senegalensis), a species considered a promising flatfish species for marine fish farming. The results show that models based on ultrasound fillet volume measurements explain above 95% of the variation observed in fillet volume. However, for fillet yield estimation, the results were modest. Therefore, further studies are necessary to better understand the potential of the ultrasound approach to this trait. Nevertheless, this work allows us to conclude that the approach with ultrasound is promising for measuring in vivo fish composition traits.

Abstract

Senegalese sole (Solea senegalensis) has been considered a promising new flatfish species for Mediterranean marine fish farming. Accurate prediction of fillet traits in live animals may allow for more efficient control of muscle deposition in fish. In this sense, this study was undertaken to develop a non-invasive method to predict in vivo fish fillet volume and yield using real-time ultrasonography (RTU). The trial was conducted with 44 market weight Senegalese sole (298.54 ± 87.30 g). Fish were scanned with an Aloka SSD 500V with a 7.5 MHz probe. Ten RTU cross-sectional images were taken from the operculum to the caudal fin at regular intervals. These images were analyzed using Fiji software. These data were then used to estimate the partial volumes of the fillet. Actual fillet volume was determined using Archimedes’ principle. Simple and stepwise multiple regression analyses were then used to develop prediction models of fillet volume and yield. The most cranial RTU sections of the fish fillet were the best single predictors of both fillet volume and fillet yield and were the ones included in the best stepwise models. The best RTU slice area explained 82% of the variation observed in fillet volume, but the other RTU slice areas used as predictors of fillet volume showed poor to moderate accuracy (0.035 ≤ R² ≤ 0.615). Single RTU partial volumes showed poor to very high accuracy (0.395 ≤ R² ≤ 0.970) as predictors of fillet volume. The best stepwise model based on the RTU slice areas included three independent variables and explained 88.3% of the observed variation. The best stepwise models based on RTU partial volumes (single volumes and/or combinations of single volumes) explained about 97% of the variation observed in fillet volume. Two RTU volume traits, V_1–5 + V_6–9, and V₁₊₍₎₊₉, showed to be practically direct predictors of the actual fillet volume, explaining, respectively, 97% and 96% of the variation observed in the actual fillet volume. The fillet yields show lower correlations with slice areas (r between 0.044 and 0.601) than with volumes (r between 0.288 and 0.637). While further studies are clearly necessary to better understand the potential of RTU for the estimation of fillet yield in fish in general and Senegalese sole in particular, the present results showed that RTU traits can be very good predictors of Senegalese sole’s fillet volume, either used in regression models or as direct predictors.

Assessing the Feasibility of Using Kinect 3D Images to Predict Light Lamb Carcasses Composition from Leg Volume

This study aimed to evaluate the accuracy of the leg volume obtained by the Microsoft Kinect sensor to predict the composition of light lamb carcasses. The trial was performed on carcasses of twenty-two male lambs (17.6 ± 1.8 kg, body weight). The carcasses were split into eight cuts, divided into three groups according to their commercial value: high-value, medium value, and low-value group. Linear, area, and volume of leg measurements were obtained to predict carcass and cuts composition. The leg volume was acquired by two different methodologies: 3D image reconstruction using a Microsoft Kinect sensor and Archimedes principle. The correlation between these two leg measurements was significant (r = 0.815, p < 0.01). The models to predict cuts and carcass traits that include leg Kinect 3D sensor volume are very good in predicting the weight of the medium value and leg cuts (R² of 0.763 and 0.829, respectively). Furthermore, the model, which includes the Kinect leg volume, explained 85% of its variation for the carcass muscle. The results of this study confirm the good ability to estimate cuts and carcass traits of light lamb carcasses with leg volume obtained with the Kinect 3D sensor.