The scale and resolution of the imagery gathered by UAS is ideal for crop monitoring and precision agriculture, whether the crop is broad acre variety such as poppies or barley, an intensive crop such as lettuce, an orchard crop such as apples, or a vineyard. The size of the area to be monitored dictates which UAV platform is most appropriate and the characteristics of the crop that need to be assessed dictates the type of sensor. The TerraLuma team have been evaluating various aircraft carrying a range of sensors to assess the potential applications of UAS for crop monitoring. In this case study we will summarise those studies through examples of datasets that have been produced using the equipment listed here.
Image acquisition with a consumer grade compact camera
The first agricultural images were taken with a Canon G10 and a modified G10 that could acquire images in the near infrared (NIR). These images can give insight into the relative vigour of vegetation but the wavelength range is broad and the NIR reflectance is difficult to quantify making it difficult to get anything more than a qualitative understanding of crop health.
This example shows a false colour infrared image taken with the modified Canon G10 camera (taken from approximately 100 m using the Maja UAV) draped over a digital surface model derived using structure from motion techniques.
Image acquisition with the 6-band multispectral camera, the thermal infrared camera and a digital SLR camera
The MiniMCA camera provided our next multispectral view of crops. In this case a barley crop was imaged that had had varying levels of nitrogen applied and the areas of high nitrogen application showed significant differences in the red edge and NIR bands.The narrow (10 nm) bands acquired by the MiniMCA provides the ability to derive spectral indices such as the the Normalised Vegetation Difference Index (NDVI) and Photochemical Reflectance Index (PRI). The FLIR thermal camera acquires coarser resolution data that can provide insight into plant stress and irrigation efficiency. The UAV captures an image every second and we have developed automated mosaicking tools to provide high resolution mosaics of crops from both multispectral and standard colour photography. As can be seen below and in other case studies these multiple views of an area can be used to derive high resolution digital surface models (DSMs).
This MiniMCA image is a standard false colour composite (Red: Green, Green: Red, Red: Near Infrared)).
In this example the NDVI is derived and again the areas that have had increased nitrogen application are clearly identifiable in the upper left (along with the weed infestation in the lower portion of the image).
The FLIR thermal camera was flown over the same crop and again the trial areas stand out as lush (and cool) barley.
We then trialled the system over a poppy crop under irrigation and we can see a marked difference between bare ground, irrigated and non-irrigated poppies.
In this mosaic image a lettuce crop is clearly visible from a flying height of 80m.
When this image is viewed more closely individual lettuce plants are easily visible in the 2cm resolution image and a farmer can assess visible characteristics.
Diseased lettuce plants are clearly visible.
To maximise lettuce plant size the lettuce seedlings must be planted in a diagonal planting pattern, maximising space for plant growth. In this image poor planting is easily discernible.
Vineyard monitoring via UAV provides vineyard managers with valuable insights into the health and vigour of their precious grapes. This mosaic of the Frogmore Creek Vineyard was produced using multiple images taken from around 40m above ground level.
As we zoom in it is clear that the resolution provides plant level detail.
As in the example above we can acquire multispectral data and evaluate vine health and vigour when we capture near infrared data.
Once again, the narrow bands enable indices to be derived. In this case the Photochemical Reflectance Index (PRI) has been calculated and we are able to clearly distinguish between the vines and the grass ground-cover between rows. Portions of the vineyard that need attention can be identified.
Thermal imagery shows the temperature differences between vines and the high temperature of bare ground.
Image acquisition with the hyperspectral pushbroom scanner
The most recent experiments have been with a small hyperspectral sensor and the results are promising. As can been seen in the examples below the high number of narrow bands provide the ability to derive biophysical and biochemical characteristics from the continuous spectrum in the visible and NIR wavelengths.
We are able to create products that can provide a farm manager with valuable insights into the health of a crop.