CropRadar – Technical background and pursued innovations

After the introduction of the new CropRadar portal in the previous issue of Kleffmann Insights, this article is now focusing on the general technical background and the pursued innovation of the recently launched CropRadar solution (figure 1).

Figure 1: The recently launched CropRadar solution of Kleffmann Group

The whole analytical method of the KLEFFMANN crop detection is based on an algorithm that utilizes multispectral and radar data which are both provided by the European Space Agency (ESA).

The actual crop detection workflow always starts with the download of the appropriate remote sensing data to the Kleffmann servers, which enables local use as well as a permanent archiving of the required data.

Once this is done, a team of geoinformatic experts start to train or rather to feed the numerical algorithm with new data from the current growing season. During this process, the deployed software receives information of how fairly advanced growth of the different crops is at the time of the data recording (ESA data) and consequently, which specific characteristics support the algorithm to differentiate between the diverse crop types. An example of a set of delineated training fields can be seen in figure 2.

Figure 2: A set of delineated fields that function as training data for the CropRadar algorithm

However, figure 2 shows a typical example based on multispectral images that shows no unfavorable weather effects. Unfortunately, especially in countries with a predominantly maritime climate, like Great Britain, persistent cloud coverage makes this procedure rather difficult. For this reason, a part of the CropRadar team currently concentrates on the recognition of arable land on SAR radar images. For this purpose, the radar data must be initially transformed to a visual image, so that the Kleffmann geoinformaticians are able to recognize different terrain structures. In a second step, the Kleffmann experts consider the varying phenological developments of the single plant types that can be gathered from the newly developed radar images. Finally, by means of comparison of the phenological developments of the different crops, it is soon-to-be possible, to avoid the cloud problem and to implement a worldwide exhaustive crop detection, regardless of the atmospheric conditions.

Nevertheless, until now, it is not feasible to detect the full range of crop types with this method and crop detection exclusively with radar data is still in the development phase. However, the Kleffmann Group is confident that this approach can soon be actively used for the CropRadar calculation and the development team is looking forward to presenting their clients new technological solutions in the near future.

Figure 3: Comparison of the usability of multispectral (left) and radar data (right). It is obvious that, in this case, only the radar data enables the plotting of the new training fields.

Author:

André Hinsenhofe, Junior Project Manager

Are you interested to receive further information?

More information about Kleffmann CropRadar can be found here.