Digital Agriculture Laboratory

Multispectral

A considerable amount of information lay beyond the visible range, where colors lose their meanings. Multispectral remote sensing takes advantage of the most informative bands within and outside the visible region. Multispectral sensors work similarly to RGB cameras, usually with 3-10 bands[1]. In fact, by some modifications, RGB cameras can be turned into a simple multispectral camera to acquire images at different wavelengths and bandwidths [2] even outside the visible region[3]. For example, to an RGB camera a filter can be added to block the red band and instead allow the NIR band [3]. However, most of multispectral cameras include several sensors and lenses and each sensor is sensitive to one spectral band. For example, the Micasense Red-Edge camera, that is a common multispectral camera in agricultural remote sensing, uses 5 separate sensors and lenses to capture NIR (842 nm center), red edge (717 nm center), red (668 nm center), green (560 nm center), and blue (475 nm center) bands with bandwidth from 12 to 57 nm[4]. Bands outside the visible region are not defined by color as they do in the visible region. In order to visualize and study these bands, different colors can be assigned to each band artificially. “Band combinations” is the term that is used in remote sensing to address the assignment of colors to represent the intensity in bands outside the visible range [5]. For instance, computers can visualize an image in which the NIR band is displayed in red, red in green, and green in blue. Such a color composite image is called a standard false-color composite that can effectively map healthy vegetation[6]. Similar band combinations can be used to estimate various plant parameters such as leaf area, leaf chlorophyll content, ground cover, and biomass [2].

Vegetation Indices (VI), ratios or linear combinations of spectral reflectance in two or more bands, are among the most used tools in multispectral remote sensing. VIs exploit vegetation's unique reflectance properties to infer biophysical properties related to plants maximizing sensitivity to the vegetation characteristics while minimizing confounding factors such as soil background, directional, or atmospheric effects[7]. Numerous VIs have been defined for various purposes, including the Normalized difference vegetation index (NDVI=(NIR-Red)/(NIR+Red)), which is the most distinguished index in remote sensing, derived from the combination of NIR and Red bands. A comprehensive review of the typical vegetation indices are presented at [8]. Although most VIs combine two or three bands, some of them use more bands. In this regard, unlike UAS-based multispectral sensors that use a small number of bands, some satellite-based spectral cameras are designed to measure more bands than regular multispectral cameras (more than ten bands) that are referred to as superspectral sensors such as Moderate Resolution Imaging Spectroradiometer (MODIS) [9]. In this paper, superspectral and hyperspectral will be used interchangeably.

The last point about multispectral cameras is that most of them, similar to RGB cameras, use a rectangular sensor plate, which allows capturing data from a two-dimensional scene at once as opposed to line scanners. However, unlike RGB cameras that has a single detector for all three bands, in multispectral cameras usually several receptive arrays are employed to capture images at different bands minimizing channel signal interference.

References

[1]       J. Qin, K. Chao, M. S. Kim, R. Lu, and T. F. Burks, “Hyperspectral and multispectral imaging for evaluating food safety and quality,” Journal of Food Engineering, vol. 118, no. 2, pp. 157–171, 2013.

[2]       B. Stark, B. Smith, and Y. Chen, “Survey of thermal infrared remote sensing for Unmanned Aerial Systems,” in 2014 International Conference on Unmanned Aircraft Systems (ICUAS), 2014, pp. 1294–1299.

[3]       T. Adão et al., “Hyperspectral imaging: A review on UAV-based sensors, data processing and applications for agriculture and forestry,” Remote Sensing, vol. 9, no. 11, p. 1110, 2017.

[4]       MicaSense, “RedEdge-MX | MicaSense,” 2020. https://micasense.com/rededge-mx/ (accessed Feb. 27, 2021).

[5]       J. B. Campbell and R. H. Wynne, Introduction to remote sensing. Guilford Press, 2011.

[6]       J. G. Liu, “REMOTE SENSING | Passive Sensors☆,” in Reference Module in Earth Systems and Environmental Sciences, Elsevier, 2013. doi: 10.1016/B978-0-12-409548-9.02956-0.

[7]       R. B. Myneni, F. G. Hall, P. J. Sellers, and A. L. Marshak, “The interpretation of spectral vegetation indexes,” IEEE Transactions on Geoscience and Remote Sensing, vol. 33, no. 2, pp. 481–486, 1995.

[8]       J. Xue and B. Su, “Significant remote sensing vegetation indices: A review of developments and applications,” Journal of Sensors, vol. 2017, 2017.

[9]       Soo Chin Liew, “Principles of Remote Sensing - Centre for Remote Imaging, Sensing and Processing, CRISP,” 2001. https://crisp.nus.edu.sg/~research/tutorial/optical.htm (accessed Jul. 25, 2020).