New algorithm makes use of satellite tv for pc photos to differentiate olive grove varieties with out area visits

A research by the schools of Cordoba and Seville develops a way that makes it doable to confirm, simply and rapidly, whether or not an olive plantation is conventional, intensive, or super-intensive.

The standard olive grove, with massive and well-separated bushes, is being changed by plantations with larger density, reminiscent of intensive or super-intensive ones, the place the area between them is lowered. Productiveness will increase with these kind of intensive and super-intensive plantations, however so does their use of assets, reminiscent of water.

As a result of this transformation’s agronomic and environmental penalties, but additionally its financial and socio-cultural ones, public administrations implement coverage and difficulty support to modulate it. To do that, they should have up to date info always to know what number of hectares of every sort of plantation there are and how briskly they’re altering.

This want is met by a way primarily based on convolutional networks, a sort of neuron community, developed by a crew on the College of Cordoba in collaboration with the College of Seville, which routinely identifies these patterns in olive groves utilizing free open satellite tv for pc photos from Sentinel-2. The work is printed within the journal Computer systems and Electronics in Agriculture.

“The issue we confronted is that, till now, we had this info, due to the pictures of the PNOA (Nationwide Aerial Orthophotography Plan), which function nice spatial decision, however they’re up to date each 3 years, so we had very outdated info,” explains Isabel Castillejo, a researcher within the Division of Graphic and Geomatic Engineering of the UCO.

To beat this drawback, the crew turned to using Sentinel-2 photos (an Earth statement mission developed by ESA) which can be obtainable totally free each 5 days. These photos, nevertheless, function decrease spatial decision, which makes it tougher to establish patterns, because the treetops can’t be straight noticed within the photos. That is the place Convolutional Neural Networks (CNN) are available in, that are a sort of Deep Studying evaluation methods used for superior sample recognition duties in information.

“We skilled 3 totally different studying algorithms to detect planting programs utilizing these satellite tv for pc photos and located that the very best of the three (strategy B) was 80% correct, a really excessive proportion contemplating the problem of the issue and the decision of the enter photos,” explains Cristina Martínez, a researcher within the UCO’s Division of Digital and Laptop Engineering.

The researchers level to a different essential benefit of this technique, which is that all the things is automated. “Simply by getting into a textual content file with the code of the plot or the cadastral reference, the plot is outlined, with its boundaries, and with that info the satellite tv for pc photos for the interval requested are recognized, downloaded and entered into the community, which determines the kind of olive grove, all in an automatic means.”

This progressive technique eliminates dependence on conventional strategies, which often contain area visits and random sampling, thus providing a extra environment friendly and exact different for the administration and monitoring of olive groves. The crew is already conducting analysis to use one of these processing utilizing neural networks and satellite tv for pc photos to the research and prediction of water stress in olive groves.