If you look up at the sky while visiting the quiet villages of Misungwi District in Tanzania you might just see an unmanned aerial vehicle (UAV) flying above you. Commonly known as a drone, UAV technology is being piloted with local farmers as an innovative way to gather information about their crops (Fig. 1 left).
Utilizing sweetpotato as a pilot crop, the International Potato Center (CIP), through a Remote Sensing project being implemented in Tanzania and Uganda is leading the efforts of developing cost-effective methods that utilize UAV as a platform to collect data for agricultural statistics. Crop statistics are important tools for planning, policy making and timely intervention to address food insecurity. Special cameras are attached to the UAV, and a pilot remotely flies the UAV over farmers’ fields to take images of the crops. Different crops reflect light in unique ways, and the optical characteristics of the crops are recorded by the camera (Fig. 1 right).
The on-going project began with the assembly of the UAV in Nairobi in January 2015 (the first in Sub-Saharan Africa) in an effort to decrease the cost of the technology and to make it more beneficial to smallholder farmers. To this end, aerial images of crops were taken during a field mission conducted in Misungwi district, Tanzania in April 2015. A team of CIP scientists are currently processing data to tell apart different crops from the images and estimate the area coverage of each crop. Sweetpotato is one of the major crops grown in the low lying region which partly borders Lake Victoria; the other crops include cassava, maize, sorghum, rice, and cotton.
Figure 1: Villagers watch the UAV flying over their fields in Tanzania (left), and an aerial image taken with a regular digital camera (right).
A wide range of methods are available, that may be used to discriminate crops from digital images, for example Maximum Likelihood, Spectral Information Divergence, Neural Network, etc. These conventional methods exploit the uniqueness of the reflective properties of crops in order to discriminate among them. However, special sensors needed to collect such kind of data are not easily available to normal users. For this reason, CIP is currently developing non-linear methods – based on wavelet transforms and multifractal analysis – to discriminate among crops using images taken with regular cameras (see comparison in Figure 2). Though the non-linear methods seem to be a bit complicated, CIP is working towards providing the technology to end-users in a user-friendly format that does not require expert knowledge to operate.
Figure 2: Identification of crops from images taken using different sensors – multispectral camera (left), and regular camera (right).
Beyond the simple discrimination of sweetpotato from other crops, CIP scientists are attempting to tell apart different sweetpotato varieties. This is a more challenging task because different varieties of the same crop reflect light in a very similar manner, so that conventional classification methods cannot discriminate them, even with highly detailed UAV-based images. “Even at this spatial resolution [5 cm] conventional processing methods do not seem to be the answer,” said the Project Leader, Dr. Roberto Quiroz. “Let us see what other less conventional processing techniques can do” He added. CIP is researching non-conventional processing methods to discriminate among sweet potato varieties.
So far, stakeholders have expressed great enthusiasm about the technology. “We look forward to seeing how this technology will improve the quality of our crop statistics”, said Mr. Lucas Kulliani, a District agricultural officer, as he received the team of scientists at Misungwi.
For more information on this technology please send an email to Dr. Roberto Quiroz (email@example.com), or see related stories in the following links: AgroTV–16; UAV Assembling; Community of Practice.