Underwater footages in bottom trawling
Underwater footages in bottom trawling
I am involved in a project focused on understanding, through underwater footage in bottom trawling, the efficacy of selection grids and gaining basic information on the behaviour of fish/shrimps towards them. Hauls will be performed at depths ranging from 100 to 200 m.
In my institute, we have always worked with GoPros and deepwater housings associated with lights (see the images attached), but sometimes, especially in dark and/or high turbidity environment, the footages are not usable.
Do you have any alternative or advice on the best solution to obtain useful footages during trawl hauls to monitor escapees from selection devices and some behaviour information?
We use these devices on commercial vessels, so we need easy-handling equipment.
Andrea Petetta, CNR IRBIM
In addition to the main post above, I would like also to receive any information on the software that you use to analyze fish behavioral data from underwater footage.
Thank you in advance.
Regarding your second question, I describe here a method we successfully applied in trawl gears and pots studies:
To collect and characterize behavioural events in underwater video recordings taken in fishing gears, we use the software BORIS (http://www.boris.unito.it/). The collected data is subsequently analysed using a model-free tool that estimates probabilities for a given fish behaviour to happen, enabling at the same time the estimation and visualization of behavioural tree diagrams representing behavioural patterns in relation to the subject of study (the fishing gear or specific device). In case you are interested to know more, do not hesitate to contact us for further information. We will be also happy to share software!
here the links to relevant papers using this method:
We have done some underwater videos in bottom trawl footgear and Nordmore grids, among others. We also use GoPros and usually, 2 DIV08W diving lights from Brinyte Technology Ltd. in the red light setting to illuminate the camera’s field of view. We use CamDo and GroupBinc deepwater housings.
We recently published a manuscript using this system: https://doi.org/10.3389/fmars.2022.920429
Also, for this particular manuscript, we used BORIS software (http://www.boris.unito.it/) but I know other researchers prefer to use Adobe Premier Pro. Both will work well to track behaviours frame by frame, Adobe Premier Pro has better video quality but you will have to manually record the data in excel spreadsheets, while BORIS software can record the data within the software and export it to spreadsheets when done.
Our company SafetyNet Technologies have been carrying out research on fish behavior in trawls for a number of years now. Like most researchers, we began with GoPro / action Cam to see and learn what was actually happening in the gear. The results were fascinating, however the process of obtaining good quality footage was not reliable and there were too many compromises, such as battery life, damage and lost sd cards to name just a few. We decided to make a solution specifically for the task and developed CatchCam
The system is turnkey and comprises of a camera, a LED lamp, a remote control, a wireless charging cradle and a tablet and App for configuration the camera and accessing the footage. The whole system is wireless and therefore stays dry all the time. The battery life is 24 hours on a single charge and the footage can be reviewed when the camera is on deck and attached to the gear, vastly improving the usability. We would be more than happy to share experience / advice.
Did you check upon Sonars? Take a look at the AIS or DIDSON stuff https://www.adfg.alaska.gov/index.cfm?adfg=sonar.didson .. We wish we had something like that, but no: We also play mainly with simple GoPro’s.
Good luck in your work!
The main problem that you will encounter with turbidity is ‘backscatter’, when your lighting illuminates particles in the water between your camera and your subjects. This has been compared to driving through a snowstorm. The conventional remedy for backscatter is to separate your lights and cameras, aiming both toward your area of interest with minimal overlap in the space close to the camera. I would expect the rig in your picture to have fairly substantial backscatter. Unfortunately, backscatter gets worse when using red or infrared lights, often used to minimize effects on fish behavior, as these frequencies are absorbed quickly and illuminate close objects more brightly. We used a fairly extreme system (expensive, pre-action-camera system) to achieve such separation in:
Olla, B.L., M.W. Davis and C.S. Rose 2000. Differences in orientation and swimming of walleye pollock Theragra chalcogramma in a trawl net under dark and light conditions: concordance between field and laboratory observations: concordance between field and laboratory observations. Fisheries Research 44: 261-266.
More common arrangements for backscatter reduction put the camera and lights at the same range, but more than 30 cm. apart.
Another arrangement to reduce turbidity in the back ends of trawls, often coming from sediment suspension by the footrope, is to supress that ‘mud cloud’ with a canvass behind the footrope. This novel idea was reported by Ludvig Krag at a recent FTFB meeting.
A description of the sediment suppression using canvas mentioned by Craig can be found here: https://doi.org/10.1016/j.fishres.2022.106323
We used the DIDSON sonar in this paper https://doi.org/10.1016/j.fishres.2005.07.015 and also applied it in trawls (unfortunately not published other than as a Project Report, which I can provide). The main downsides are that they are very expensive and we had to develop an autonomous recording/power system, which had reliability problems. It does solve the illumination issue and increases range ofver camera systems, but both turbidity and net panels can interfere with the images. The geometery of the images takes some adjustment for those used to camera imagery.
While searching on this topic, I came across https://doi.org/10.1016/j.ohx.2020.e00149 , which seems relevant. (I note that their camera and lights are in separate housings, allowing separation)
Some experience presented in these papers:
Madsen, N., Pedersen, M., Jensen, K. T., Møller, P. R., Andersen, R. E., & Moeslund, T. B. (2021). Fishing with C-TUCs (Cheap Tiny Underwater Cameras) in a sea of possibilities. The Journal of Ocean Technology, 16(2), 19-30. https://www.researchgate.net/publication/352350751_Fishing_with_C-TUCs_Cheap_Tiny_Underwater_Cameras_in_a_Sea_of_Possibilities
Pedersen, M., Madsen, N., & Moeslund, T. B. (2021). No Machine Learning Without Data: Critical Factors to Consider when Collecting Video Data in Marine Environments. The Journal of Ocean Technology, 16(3), 21-30. https://www.thejot.net/archive-issues/?id=73
Pedersen, M., Haurum, J. B., Gade, R., Moeslund, T. B., & Madsen, N. (2019). Detection of Marine Animals in a New Underwater Dataset with Varying Visibility. I IEEE Conference on Computer Vision and Pattern Recognition Workshops IEEE. IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) http://openaccess.thecvf.com/content_CVPRW_2019/html/AAMVEM/Pedersen_Detection_of_Marine_Animals_in_a_New_Underwater_Dataset_with_CVPRW_2019_paper.html
Perhaps a bit dated now but here is our research titled, “Development and observations of a spiny dogfish Squalus acanthias reduction device in a raised footrope silver hake Merluccius bilinearis trawl”. We used cameras to categorize fish behaviors around a grid.
We published a manuscript using underwater video camera to record the behavior of the catches inside the codend during towing.
Keigo Ebata (Kagoshima University, Japan)