Proceedings
Authors
| Filter results7 paper(s) found. |
|---|
1. Knowledge-based Approach for Weed Detection Using RGB ImageryA workflow was developed to explore the potential use of Phase One RGB for weed mapping in a herbicide efficacy trial in wheat. Images with spatial resolution of 0.8 mm were collected in July 2020 over an area of nearly 2000 square meters (66 plots). The study site was on a research farm at the University of Saskatchewan, Canada. Wheat was seeded on June 29, 2020, at a rate of 75 seeds per square meter with a row spacing of 30.5 cm. The weed species seeded in the trial were kochia, wild oat, wild... T. Ha, K. Aldridge, E. Johnson, S.J. Shirtliffe, S. Ryu |
2. Establishment of a Canola Emergence Assessment Methodology Using Image-based Plant Count and Ground Cover AnalysisManual assessment of emergence is a time-consuming practice that must occur within a short time-frame of the emergence stage in canola (Brassica napus). Unmanned aerial vehicles (UAV) may allow for a more thorough assessment of canola emergence by covering a wider scope of the field and in a more timely manner than in-person evaluations. This research aims to calibrate the relationship between emerging plant population count and the ground cover. The field trial took place at the University... K. Krys, S. Shirtliffe, H. Duddu, T. Ha, A. Attanayake, E. Johnson, E. Andvaag, I. Stavness |
3. Mapping Marginal Crop Land on Millions of Acres in the Canadian PrairiesCrop fields cover more than 250,000 km2 of the Canadian Prairies, and many of these contain areas of marginal soil condition that are farmed annually at a loss. Setting aside these unprofitable areas may represent savings for growers as well as reductions in GHG emissions, while restoring them with perennial vegetation could create new natural carbon sinks. There is high potential for these in-field marginal zones to act as a nature-based climate solution in Alberta, Saskatchewan and Manitoba.... S. Shirtliffe, T. Ha, K. Nketia |
4. Digital Agriculture Driven by Big Data Analytics: a Focus on Spatio-temporal Crop Yield Stability and Land ProductivityIn the ever-evolving landscape of agriculture, the adoption of digital technologies and big data analytics has ushered in a transformative era known as digital agriculture. This paradigm shift is primarily motivated by the pressing imperative to address the growing global population's food requirements, mitigate the adverse effects of climate change, and promote sustainable land management. Canada, a significant player in global food production, has made a substantial commitment to reducing... K. Nketia, T. Ha, H. Fernando, S. Shirtliffe, S. Van Steenbergen |
5. Autonomous Edge-AI–Enabled Drone Systems for Real-Time Agricultural Inference and Decision-MakingHigh-throughput, low-latency phenotyping and field surveillance remain critical bottlenecks in precision agriculture and environmental monitoring due to delayed data turnaround, large data volumes, computationally intensive preprocessing, and expertise-heavy analysis workflows. These constraints hinder timely crop improvement, pest and disease management, and informed agronomic decision-making. To address these challenges, we present an integrated, end-to-end autonomous drone system that enables... |
6. GAIG: High‑Resolution Wall‑to‑Wall Modelling of Within‑Field Spatial Variability in Crop YieldQuantifying the temporal stability and causes of within‑field variation in crop yield is fundamental to precision‑agriculture research, particularly as agricultural areas seek to identify lands with persistently low productivity that may constitute marginal cropland. What is needed is a modelling framework capable of using spatial patterns in yield to reveal stability zones, diagnose sources of variability, and enable consistent comparison across farms and years. Accordingly, the objective... |
7. Semi-Automatic Plot Segmentation for Crop Phenotyping Using Adaptive Spectral Indices and SAM3Yield trials and hill plots are widely used in plant breeding to evaluate large numbers of genotypes simultaneously. Extracting per-plot canopy boundaries from drone imagery is a key step in this process, but manual delineation is time-consuming, and rigid grid overlays do not account for true canopy boundaries. This paper presents an annotation-free pipeline for segmenting individual plots from multispectral drone imagery, requiring only approximate plot dimensions as input. The pipeline first... A. Lotfi, S. Shirtliffe, A. Carter, T. Ha, M. Eramian, S. Neupane |