Certified Robustness Bounds for Computer Vision Weed Detection in Artificial Intelligence Agriculture Under Sensor Degradation: A Computer Adversarial-Risk Guarantee

Computer vision weed detectors deployed on field robots must remain reliable when camera sensors degrade through dust, moisture and lens abrasion, yet the robustness certificates available for such systems are typically derived for idealised input perturbations and do not account for the structured, physically realisable corruptions that occur in agricultural settings. The framework formalises degradation as a composition of bounded operators on the image signal, defines a certified radius in a physically interpretable parameter space, and proves that the resulting guarantee holds uniformly over the degradation set under a Lipschitz condition on the detector. A before-after-control-impact design is used to characterise the operating characteristics of the certification procedure under a declared variance structure with an intraclass correlation of 0.64, between-cluster standard deviation of 6.0 and residual standard deviation of 4.5; power was estimated across 2000 replications per design cell for cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The analysis indicates that 52 clusters are required to detect a 5 percentage point effect at 80% power. The contribution is a formal bridge between sensor-degradation physics and adversarial-risk certification, together with a design-level assessment of how much field evidence is needed to validate such guarantees.