Computer Vision–Enabled Evaluation of Boom Dynamics and Application Uniformity on Commercial Sprayers Under Real-World Operating Conditions.

Abstract

Efficient and uniform spraying is one of the biggest challenges in modern crop production. As sprayer booms get wider and machines travel faster, even small vibrations or terrain changes can disturb how chemicals reach the crop canopy. When a nozzle moves too far ahead or behind, or rises and dips too much, the spray pattern shifts—causing some areas to be under-applied and others over-applied. These subtle height and motion changes, though often invisible during operation, have a major impact on coverage, drift, and chemical waste. This research combined field experiments, computer vision, and advanced sensing to understand how a 120-ft sprayer boom moves and how that motion affects spraying accuracy under real field conditions. A computer vision Distance Quantifier System (DQS) Dalal et al., (2025) was developed to track a reflective fiducial target mounted on the boom tip. The system, paired with a dual-frequency GNSS receiver, recorded motion at 10 Hz with centimeter-level precision. In the first phase, this DQS–GNSS framework was used to measure horizontal (fore–aft) boom motion and its impact on chemical application rate for three commercial self-propelled sprayers: a Hagie STS12 (front boom, hybrid aluminum–steel truss), a John Deere 412R (rear boom, steel truss), and a Miller Nitro 7310 (front boom, steel truss). Field trials conducted across approximately 100 acres demonstrated that horizontal surging resulted in localized under- and over-application, even when rate controllers maintained the target flow. Among the three machines, the rear-boom John Deere remained the most stable (≤ 2 % rate-error exceedance area at the 10 % rate-error threshold), while Miller’s heavier all-steel front boom displayed the largest fore–aft excursions—reaching up to 2.7 m—and the highest rate-error exceedance area (7–10 %). The hybrid Hagie (Steel + Aluminum boom) showed intermediate behavior. Statistical analysis confirmed that boom design—specifically, position, material, and mass distribution—had a significantly stronger effect on rate uniformity than travel speed, which had only a minor scaling influence. These results provided the first field-scale evidence linking horizontal boom motion directly to measurable rate error. In the second phase, the framework was expanded to include radar canopy sensors alongside the existing DQS–GNSS setup, enabling simultaneous measurement of boom motion and canopy surface response. Using a New Holland Guardian SF310 – a 120ft front-boom sprayer with synchronized DQS, GNSS, and company-fitted Raven AutoBoom XRT radars were used to reconstruct nozzle-to-canopy clearances along the 60-ft left wing. Results revealed that vertical motion was dominated by rigid-body pitching about the boom hinge rather than terrain following. The boom often tilted more than 3°, occasionally exceeding 4.8°, with corresponding tip displacements up to 1.07 m. Strong correlations (r > 0.85) between tip and mid-radar signals confirmed coherent canopy undulations across the boom span. Fault analysis showed that both the frequency and magnitude of height deviations increased with travel speed and with lateral distance from the hinge. At the strict ±5 in (±0.127 m) tolerance, the left boom wing operated outside the target height for about 37 % of the run at 6 mph, 41 % at 12 mph, and 44 % at 18 mph. When the tolerance was relaxed to ±10 in (±0.254 m), these out-of-tolerance-time fractions were reduced by roughly half. The tip section, being farthest from the hinge, consistently exhibited the largest deviation—exceeding 59 % rate error exceedance under the tighter ±5 in band—while the center section remained the most stable. When a height-based section shut-off logic (+27in /-20in) was applied, the system predicted an average 10 % reduction in active spray time, equivalent to nearly 5 L of chemical savings per boom per pass, or more than 35 L per 100 acres of field coverage. Overall, this research demonstrated that both horizontal and vertical instabilities significantly influence spray uniformity. Together, the two studies show how complementary sensing of boom motion and canopy profile can be used to quantify field-scale spraying stability. The integrated DQS–GNSS–radar framework developed through this research provides a practical foundation for future adaptive control systems and sprayer designs aimed at improving application uniformity, chemical efficiency, and drift reduction.