Authors:

Han Li

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Achieving uniformity in total mixed rations (TMR) on dairy farms is crucial for optimizing digestive function, enhancing nutrient utilization, and maintaining overall animal health. However, there remains a lack of automated, real-time systems for monitoring the preparation of TMR. This study explores the use of textural feature extraction from RGB images as a tool for assessing TMR uniformity. Cameras were mounted above a mixer wagon to record the mixing process of poorly-mixed and well-mixed lactating cow diets. Both diets were composed of the same ingredients, with the poorly mixed diet having a larger load size and shorter mixing time compared to the well-mixed diet. From each batch, the initial, middle, and final frames of the mixing process were extracted. In each frame, 100 regions of interest (ROI) were randomly selected, and various handcrafted textural features were computed, such as those derived from the Gray Level Co-occurance Matrix, Gabor filters, and Local Binary Patterns. Each feature was tested for robustness against variations in rotation, lighting, and camera distance to identify those that are most generalizable across different feed mixing conditions. Features were also tested for sensitivity to mix uniformity by calculating the coefficient of variation for each feature across all ROIs within a frame, and comparing the distribution of poorly-mixed and well-mixed groups. Features that effectively differentiated between groups showed statistically significant differences in their distributions, demonstrating the potential that textural analysis has in assessing diet uniformity as part of an automated system for real-time monitoring. Keywords: Video Analytics; Computer Vision; Texture Analysis; Feed Mix Uniformity; Dairy Feed Management

Source:

Purdue University / 2025

Topics:

No topics listed

Co-authors:

Han Li