Hyperspectral imagery

The narrow bands of hyperspectral imagery allow it to capture more information about the materials that are being imaged.

Fig Figure 1. Number of bands and associated wavelengths typically represented in RGB, Multispectral and Hyperspectral imagery (image source: wikipedia, created by Lucasbosch).

This can be used to identify different materials, even if they are very similar in appearance. For example, hyperspectral imagery can be used to identify different types of vegetation, even if they all appear to be the same colour. This feature of hyperspectral imagery makes them ideal for detection of weeds, even if they are located in areas of dense vegetation.

HS imagery captures data in the NIR, SWIR, MIR, and LWIR bands, while multispectral imagery typically only captures data in the visible and NIR bands (Table 1). Hyperspectral imagery has more but narrower bands providing more detail than the wider bands of multispectral imagery, however, hyperspectral imagery is more expensive and  complex to acquire and process.

In the context of weed detection, it is important to consider that both hyperspectral (HS) and multispectral (MS) imagery typically have lower spatial resolution than RGB imagery. This is because RGB imagery uses a smaller number of bands, which allows for a higher spatial resolution.

Spatial resolution refers to the amount of detail that can be seen in an image. A higher spatial resolution image will have more detail than a lower spatial resolution image. Hyperspectral imagery has been useful for plants in vegetative stages or with plants that appear similar to surrounding foliage, where green foliage looks very similar to surrounding green vegetation. The high number of bands allows for greater spectral resolution than multispectral sensors can achieve.

Table 1. the approximate wavelength ranges for some common hyperspectral and multispectral bands