Utility & Mathematical Foundations

Overhead computer vision is widely used in PLF for broiler welfare scoring (gait analysis), swine weight estimation, and dairy cattle tracking. However, deep learning models (like the YOLO single-pass detector family) fail if the target animal is resolved by too few pixels, or if lenses create blind spots.

This tool evaluates camera deployment setups using optical formulas. It calculates ground coverage and details in pixels-per-centimeter (px/cm):

• Coverage Width (Wc): Calculated as \(W_c = 2 \times H \times \tan(\theta_{FoV} / 2)\) where H is mounting height and FoV is the lens angle.
• Ground Resolution (R): Calculated as \(R = \text{Horizontal Pixels} / (W_c \times 100)\). A minimum of 4.5 px/cm is required to detect small joint keypoints on broiler chickens (e.g. for lameness scoring). Swine and cattle tracking can operate down to 2 px/cm.