Fish species identification from acoustic broadband data

Project

Fish species identification from acoustic broadband data

Through collaboratioin between the pelagic industry and acoustic fisheries scientists this project aims to improve fishing selectivity through acoustic species identification and therefore reduce bycatch by use of a novel broadband echosounder.

Introduction

The broadband technique is at the very forefront of technical developments in fisheries acoustics. The project will deliver a demonstrator imaging software for classification and analysis of broadband data. Focus will be on identification of 3-4 fish species (for example: herring, mackerel, horse mackerel, boarfish) from the collected broadband acoustic data. These classifications will be based on test data collected on a trawler during fishing operations, modelled broadband sound scattering and algorithms for different fish species.

Background

Echosounders are important instruments commonly used on pelagic fishing vessels to detect fish schools, quantitatively asses their size and facilitate catch operations. The same instruments, but then calibrated, are also used for scientific purposes to quantitatively assess fish biomass and estimate stock sizes.

Figure 1. Echogram showing broadband data of a mackerel school observed on a pelagic freezer-trawler
Figure 1. Echogram showing broadband data of a mackerel school observed on a pelagic freezer-trawler

Through the use of echosounders operating sound pulses simultaneously at different frequencies, fish species may be separated from each other using objective scientifically derived species identification algorithms. These algorithms rely on specific sound reflection properties of different fish species (Figure 2a). They evaluate the echo strength measured between different single frequencies and use these distinct frequency signatures to classify fish species (Figure 2b). However, these multifrequency techniques take measurements only at the limited number of single frequencies that are available. Due to the variability around the measured echo strength values at these different frequencies and the sometimes minimal differences between species, classification is sometimes difficult and results not usable for species separation in practice.

Broadband techniques make use of so-called chirp pulses thereby covering a wide frequency band in every single ping. They can provide echo strength measurements over a wide (broad) frequency range rather than at just a number of distinct frequencies (Figure 2c). Consequently, the information can be provided much more accurately.

Figure 2. Schematic description of frequency-specific identification properties of different fish species. a) Echo strength of the fish species (solid lines) and measurements taken at four distinct frequencies (filled circles) (mackerel: blue; herring: red; horse mackerel: green; boarfish: orange). b) Multi-frequency echo characteristics of the species used for identification algorithms. c) Broadband technique covering a wide frequency band (e.g. 50-300 kHz) and measuring ehco strengths of the species at essentially 250 individual frequencies.
Figure 2. Schematic description of frequency-specific identification properties of different fish species. a) Echo strength of the fish species (solid lines) and measurements taken at four distinct frequencies (filled circles) (mackerel: blue; herring: red; horse mackerel: green; boarfish: orange). b) Multi-frequency echo characteristics of the species used for identification algorithms. c) Broadband technique covering a wide frequency band (e.g. 50-300 kHz) and measuring ehco strengths of the species at essentially 250 individual frequencies.

Aims

Results

Figure 3. Classification of 3 fish species based on 2 frequency response features of acoustic broadband data.
Figure 3. Classification of 3 fish species based on 2 frequency response features of acoustic broadband data.

European fisheries.png