Jellyfish and drone

 

This project was in Pruth Bay, British Columbia, Canada. 

 

Unmanned aerial vehicles (i.e. drone) have become a popular recreational and photography tool. They also enabled the remote sensing community by offering high resolution images, and extremely flexible flights in comparison to conventional airborne and space-borne platforms (Figure 1). 

 Figure 1.  A spatio-temporal comparison between common sampling options in ocean studies (Schaub et al., 2018).

Figure 1.

A spatio-temporal comparison between common sampling options in ocean studies (Schaub et al., 2018).

This project was Jessica's honorable thesis that I was lucky enough to get my hands on those fine drone images (Figure 1). The goal was to aerially survey and monitor jellyfish aggregations and calculate their percent coverage and wet weight mass. 

 Figure 2.   A  DJI Phantom 3  partying with a school of jellyfish. 

Figure 2. 

A DJI Phantom 3 partying with a school of jellyfish. 

Jellyfish are known to form dense aggregations and have a list of impacts on marine ecosystem
processes. The density and formation of these aggregations can be used as indicators for monitoring larger scale environmental issues such as carbon fluxes, microbial loops, and organic matter pool in the nearby water body.  

My role in this project was to assist on how to extract and count these jellyfish aggregations from the raw drone images. Raw images were captured by a DJI Phantom 3 equipped with a 12 mega-pixel camera, capturing images at a 90 degree angle with a sub-meter pixel resolution. 

I applied a texture analysis on the raw RGB images in ENVI 5.2. Given the lack of spectral information, texture analysis accounted for both the variation and data range for all three bands (i.e. RBG).  The output consisted of 9 bands in total: original, variance, and data range for each of red, green, and blue. 

The next step is to count how dense the those jellyfish aggregations were in the images. A k-means clustering analysis was applied to delineate pixels according to differences and similarities in the 9 bands of the stacked image. 

Figure 2 shows a comparison between raw RGB and the results from texture analysis. 

 Figure 3.   An example of how well texture analysis is able to extract the jellyfish aggregations. a) raw RBG images, b) results from texture analysis on all three colors. The red dots in Figure 2b are indicative of jellyfish aggrogations. 

Figure 3. 

An example of how well texture analysis is able to extract the jellyfish aggregations. a) raw RBG images, b) results from texture analysis on all three colors. The red dots in Figure 2b are indicative of jellyfish aggrogations. 

Our results suggested that the use of UAV facilitates rapid and reliable assessment of jellyfish aggregation which is otherwise costly and inaccurate due to the lack of horizontal view.

The fine resolution georeferenced imagery enables the detection of even smaller differences in the extent of the jellyfish aggregations e.g. within the same tidal period and through the day). This will open up the opportunity for accurate high-frequency measurement and monitoring of aggregation size, shape, and movement. Comparing to conventional manned aircraft for aerial surveys, such possibility offered by UAV's is a major advantage and should be applied more in ocean samplings. The lower cost and ease of use of UAVs allow high-frequency repeated passes of a sampling area, enabling high temporal resolution monitoring and measurement of aggregations.

 

Full article can be accessed here

Jessica's lab is here. You should check it out. It's super cool. 

This research was funded by the Tula Foundation. We thank the staff of Hakai Institute for assistance in all phases of this research. Bryn Fedje, Nelson Roberts, and Emma Myers assisted in field data collection.