Image analysis as a tool for measuring total mixed ration uniformity on dairy farms

Achieving uniformity in total mixed rations (TMR) is crucial on dairy farms for optimizing rumen function, enhancing nutrient utilization, and maintaining overall animal health. Conventional physical sampling methods for assessing TMR uniformity are widely used, such as the Penn State Particle Separator (PSPS), nutrient analysis, and near-infrared spectroscopy (NIRS), but they can be time-consuming, labor-intensive, and prone to human error. This study demonstrated, as a proof of concept, that image analysis can be used to assess TMR uniformity. A video recording system mounted above the mixer wagon captured surface images during blending of lactating diets that differed in load size and mixing time. For each of two conditions: poorly mixed (n = 6 batches) and well mixed (n = 9 batches), ten samples were collected along the feed bunk for each batch. They were analyzed using the PSPS and nutrient analysis including manganese (Mn) and zinc (Zn). Utilizing recorded videos of the mixing process, one of the initial, middle, and final frames of each batch were obtained. Within each frame, 100 regions of interest (ROI) were selected and all pairwise comparisons were made using the Earth Mover’s Distance (EMD) to quantify the uniformity of the TMR by comparing pixel intensity distribution between ROI. Results from linear regression indicated that both the mean and 95th percentiles for EMD increased as the CV of Mn and particle size increased. The EMD mean and median for each frame and condition were plotted as boxplots, with the 5th and 95th percentiles used to visualize the distribution of EMD values. Frames showed a decrease in the EMD median and range of values from the initial to the final frame with greater reduction and reduced range in the final frame for the well-mixed diet compared with the poorly-mixed diet. The EMD mean and range from 5th to 95th percentile had strong correlation with CV of Mn, Zn, middle sieve, bottom pan and mean particle size when tested with Spearman rank order correlations. The proposed method captured the improvement on TMR uniformity over time, differentiated between poorly and well-mixed diets, and showed correlation with traditional methods, demonstrating its potential to be used to monitor TMR uniformity on dairy farms.