Reza Ehsani,Dvoralai Wulfsohn,Jnaneshwar Das,Ines Zamora Lagos 2016-06-29 04:35:46
A low-hanging fruit for application of small UAS Small unmanned aerial systems (sUAS) are powered aerial vehicles that weigh less than 55 lb (25 kg) and can operate autonomously or be operated remotely by a human pilot. A conservative estimate by an aviation economist projected a $13.6 billion impact of this technology in the first three years of approved commercial use, and precision agriculture was projected to comprise as much as 80% of the commercial market. The ability of sUAS to collect high-fidelity spatial, spectral, and temporal aerial data at relatively low cost can provide new opportunities for growers. In fact, sUAS may change the way we collect field data, monitor field equipment, and even operate agricultural machinery. Among the many potential applications of sUAS, crop scouting, inventory management, tree canopy management, and yield estimation will probably be the most highly used applications in commercial fruit and vegetable production. Software for optimal sampling Early and accurate yield forecasting is important for many crops, but traditional techniques are time-consuming, labor-intensive, and often inaccurate. As an alternative, sUAS can be used for yield estimation, either by representative sampling or direct count. For sampling, sUAS can be used to get an accurate count of trees and determine the distribution of tree size, health, and other canopy features. This information can then be used to make a near-optimal selection of sample trees for manual fruit counts and measurement of fruit quality and size. The Pronofrut sampling system, developed by Dayenú Limitada, an agriculture technology and consulting firm in San Fernando, Chile, takes this approach and uses sUAS and software algorithms to provide near-optimal tree selection. Manual counts are supported by handheld software for direct sampling in the field, and the within-tree sampling algorithm can be adapted to many plant structures and architectures. The advantage of manual counting is that systematic errors due to fruit masking and other artifacts in images are avoided, allowing consistent yield estimates from small, statistically representative samples. Typically, hundreds of fruit counts are distributed spatially over a grove. The handheld software indicates the location of the next sample (row, tree, branch, or branch segment), always moving forward in the orchard, and records the sample location, time stamp, and measurements. The Pronofrut system provides yield estimations with a known absolute error range (the goal is typically <10% error) and substantially reduces human resource requirements compared to traditional sampling. The Pronofrut sampling methodology has been validated at the commercial scale for fruit, nut, berry, and vegetable crops in Chile, Argentina, the U.S., and Spain, including grapes, sweet cherries, apples, pears, maize, and hybrid cucumber seeds. Image-based yield estimates Another technique is to use sUAS imagery for counting fruit. In 2015, the USDA funded a project for the University of Pennsylvania and the University of Florida to investigate the use of sUAS for yield estimation of citrus, tomatoes, and blueberries. With a focus on data-driven techniques to improve estimation accuracy, this project will incorporate direct fruit counts through close-range imaging, followed by a correction based on ground-truth fruit count data. A specialized sensor suite, called the Intelligent Remote Imaging System (IRIS), was designed and tested at the University of Pennsylvania in collaboration with specialty crop growers to enable high-fidelity multi-spectral and multimodal data acquisition. Additionally, information on canopy size and health, such as the normalized difference vegetation index (NDVI), acquired by IRIS will allow statistical models to predict yield based on both direct sampling of the fruit count as well as the overall aerial observations. There are two sources of error with image-based yield estimation that we are addressing. First, image-based fruit counting during daylight hours suffers from shadows and reflections that introduce errors. Second, the camera can only see fruit that are not hidden by leaves, and site-specific calibration with a ground-truth fruit count is needed to ensure an accurate count with the current technique. We are working on new technologies, such as use of controlled illumination, as well as advanced methods such as backscatter x-ray imaging, that can resolve these issues and result in higher accuracy in fruit counts. With the large amount of research and development currently being conducted by universities and private companies, sUAS will be much more than toys. They will be essential tools for growers in the not-too-distant future. ASABE member Reza Ehsani, Associate Professor, Department of Agricultural and Biological Engineering, University of Florida, Institute of Food and Agricultural Sciences, Citrus Research and Education Center, Gainesville, USA, email@example.com. ASABE member Dvoralai Wulfsohn, Director, Geco Enterprises Center for R&D, El Tambo, Chile, firstname.lastname@example.org. Jnaneshwar Das, Postdoctoral Researcher, General Robotics, Automation, Sensing and Perception Laboratory, University of Pennsylvania, Philadelphia, USA, email@example.com. ASABE member Ines Zamora Lagos, General Manager, Dayenú Limitada, San Fernando, Chile, firstname.lastname@example.org.
Published by ASABE. View All Articles.
This page can be found at http://bt.e-ditionsbyfry.com/article/Yield+Estimation/2522850/316822/article.html.