• Tom Leeuw

    Tom Leeuw
    University of Maine
    Worked on the newly developed Cam-Trawl system, a method to improve fisheries acoustic surveys.

    As an intern at the Alaska Fisheries Science Center (AFSC) I had the opportunity to work with the newly developed Cam-Trawl system. The Cam-Trawl is a novel invention consisting of still image stereo-cameras mounted inside a mid-water trawl net. From its location near the cod-end, the Cam-Trawl can collect images of fish that pass through the net. In the context of fisheries acoustics, these images can be used to accurately identify acoustic targets on an echogram. My internship was focused on integration of optical data from the Cam-Trawl and acoustics data from the echo sounder.

    The Cam-Trawl solves an important issue in fisheries acoustics: how do you identify a sound scattering object observed on an echogram? The standard approach is to tow a net through the sound scattering layer and examine the contents. This method works well if the cod-end of the net only contains a single species. However, in the Gulf of Alaska this is rarely the case. Nets are often towed through multiple scattering layers and cod-ends often contain a variety of species. What the Cam-Trawl allows us to do is see in what order fish entered the net. This allows for identification of scattering targets by matching an image with any scattering target that is within the trawl path.

    To accomplish the task of matching images with acoustics, I wrote an algorithm that defines the area swept by the net on an echogram. It works by first measuring backscatter of light into the cameras by averaging pixel values from the images. The algorithm then measures the volume backscattering strength (Sv) from very small regions along the trawl path on the echogram. These regions are systematically shifted

  • Echogram of walleye pollock.

    in time along the echogram until the correlation between optical backscatter and acoustic backscatter (in terms of Sv) is at a maximum. The final trawl path is made up of regions that correspond to specific image numbers. Scattering targets that fall within these regions can be identified by viewing the corresponding images.

    During my internship I also had the chance to work with the Cam-Trawl first hand. For the first three weeks of July I joined the members of the midwater assessment and conservation engineering group (MACE) aboard the R/V Oscar Dyson in the Gulf of Alaska. These three weeks were part of the biennial Gulf of Alaska acoustic survey. The survey collects acoustic and trawl data that is used to estimate abundance of Walleye Pollock and other fishes in the Gulf of Alaska. As part of my duties aboard the Oscar Dyson I aided in collecting lengths, weights, and sexes of trawl-caught fish. In addition, we collected stomachs and otoliths from a subset of each catch. I also aided in the deployment of the Cam-Trawl, drop target-strength transducer, bottom typing drop-camera, and Expendable Bathythermographs (XBTs).

  • Image of walleye pollock taken with the Cam-Trawl.

    Back in Seattle, I worked on a number of different projects in addition to the correlation algorithm. To allow for measurements of fish in the images to be made, I calibrated the stereo cameras using a MatLab toolbox. I wrote an experimental program for object segmentation and stereo-correspondence of fish in the left and right images. Lastly, I examined transit time of fish through the net and compared it to the transit time of passive particles.

    My experience at the AFSC was exciting and informative. I had the opportunity to explore one of the fields of research I am most interested in. It was an added bonus that I was able to contribute to the Cam-Trawl system and in the process help improve the accuracy of acoustic surveys.