Researchers at the University of California, Davis have developed a system that combines large datasets of aerial imagery with artificial intelligence to acquire per-plant analytics and predict crop yields. The system is a scalable per-tree yield prediction model for nut crops, provides large-scale canopy profile analytics in 3D, and the next generation of aerial image analytics for agriculture.
Growers need improved, early estimates of plant/tree data in order to optimize crop health and overall field/orchard yields. Analyzing aerial images obtained from unmanned aerial vehicles (UAVs/”drones”) on a per-tree level allows for more precise yield predictions, optimum irrigation, and best nutrient management practice by leveraging machine learning algorithms. Targeted interventions involving water, and nutrient inputs are especially critical in regions (such as California) that limit total nitrogen applications for environmental and sustainability reasons. These types of per-plant/tree yield analyses allow growers to have an insight into their orchard’s within-field variability in productivity both more accurately and earlier in the growing season.
Techniques such as using a lightbar platform on a utility vehicle for measuring fractional photosynthetically active radiation (fPAR) are used currently to assess and predict certain crop yields. This technology can provide insights into canopy light interception; however, such method is time consuming, expensive, and require human presence. These methods are also imprecise. While they may provide an overall approximation of the performance of a field/orchard, no detailed information on a per-tree basis is generated by this method. Thus, improved/superior methods of gathering per-plant field data would have a significant impact on agricultural productivity.
Researchers at the University of California Davis have integrated aerial imaging, sensing technologies, and soil-sensing datasets with traditional plant and soil monitoring. Processing these data using algorithms improves harvest management practices and generates 3D models of plants. Light Detection and Ranging (LiDAR) sensors allow data collection to optimize nitrogen and irrigation scheduling for both individual plants and entire fields. These data can provide growers more insights into the yield variations in trees of similar size, providing opportunities to improve overall acreage productivity. Canopy optimization can also be estimated by interpreting the canopy profile data captured. Analyzing the shading effect of branches and neighboring trees has been validated by three years of precise ground data. Compared to alternative sensing methods, this technology provides a less expensive, more accessible, higher precision (per-plant) and accuracy and faster method of measuring canopy profiles, potential productivity and other, critical, field/crop details.
crop yield optimization, orchard management, artificial intelligence, big data, 3D modeling, LiDAR, drones, UAVs, remote sensing, sustainability, environmental stewardship, Smart Farm, photogrammetry