A new technology developed by researchers at the University of Minnesota will allow largeholders to see important plant species earlier in the season than ever before.
Satellite images have long been used by agricultural organizations to tell which plants are grown in the field. This allows manufacturers to predict supply chains, evaluate crop damage due to environmental factors and organize supply-chain logistics.
While this information is important, currently available mapping products cannot provide this information at the beginning of the agricultural season. For example, the crop database (CDL), a mapping product by the USDA National Agricultural Statistics Service, is not released frequently until four to six months after collection. Because of the long field data collection process required to learn the backend algorithm for separating plants from satellite images.
In a study published in Distance Vision of the EnvironmentResearchers at the University of Minnesota explain that they are developing a new way that farmers can see where wheat and soybean will grow in early July, with precision. such as the USDA CDL, and the lack of global surveillance.
With the availability of fast -growing satellite data and the advancement of labor knowledge and globalization, the bottleneck of labeling products has shifted due to the lack of authentic labels. , Which is the history of fruit varieties in specific places. In such cases, scientists have tried to use old labels to identify plants of the expected year.
For example, to document plant species in 2022, scientists will develop a model using labels collected in 2021, 2020, or earlier in order to develop a model. for example, if a new survey is not available or not available. However, this model often fails because it can change with changes in soil, season and driving practices over the year given the nature of the plants in the satellite images.
To avoid the need to collect country labels, the research team created pseudo-labels (they are called “pseudo” because these labels are not collected from months. Food) each year in question in relation to the plant species.
This method simulates how a person sees things in relation to their relative positions (also called relation topology) in an image and uses a computer model to see the corn and the soybean because of their topology relationship in a two -dimensional space obtained from satellite images. These pseudo-labels are designed to be similar to collected labels and can be used for the critical task of labeling product types in the first place.
“This is a revolutionary approach that uses computer -assisted technology to model how a person sees different things in images. The types of plants in early July,” said Zhenong Jin, Ph.D., assistant professor in the Department of Bioproducts and Biosystems Engineering at the University of Minnesota.
“We have had stable topology relationships for different plants in different years and in different countries, showing how we can extend a comprehensive design that works for the same features. No, ”said Chenxi Lin, a Ph.D. candidate and original author of the work mentioned by Jin.
The study also found:
- The method allows to create pseudo labels of similar quality as the labels collected for different plants grown in different years and in different places.
- In the U.S., the accuracy of labeling crop types is based on pseudo-labels compared to the cropland land product data (CDL) about six months ago.
- In the North of France, this type of work can help significantly reduce the amount of land labels needed to produce suitable crops, which can be difficult due to the large number. of the plants that grow in the land.
In addition, high -quality maps of the past were created the way they were supposed to be used for other purposes.
A full and timely tour of the designated countries is an advantage for insurance companies to better plan their products. In addition, product acreage and customer counting can help customers get better project costs, and fence as such.
Looking ahead, researchers find that the implementation of this approach, relying on authentic historical labels, is not a problem for large -cap countries like the United States, but a source of income. border for places like Africa.
However, implementing an approach to underdeveloped countries like many in Africa could have far -reaching consequences for the ultimate goal of achieving a food security world. The company plans to increase the bandwidth reported in this study in those areas by incorporating in -depth learning algorithms that reduce the usefulness of historical labels.
Cultivation and covering the effects of making flour
Chenxi Lin et al, Recording early and periodic product types without the accuracy of the current year: Creating labels from historical information by topology method. Distance Vision of the Environment (2022). DOI: 10.1016 / j.rse.2022.112994
Presented by the University of Minnesota
Directions: Using technology to detect plant species in the first place, without entering the field (2022, March 31) Retrieved 1 April 2022 from https: // phys .org/news/2022-03-technology-crop-early-season-field .html
This document is subject to copyright. Except for appropriate work for the purpose of personal inquiry or research, no piece may be reproduced without permission. Information is provided for informational purposes only.