Preliminary tests of SciFish 2000 indicate that it is more than 85% accurate and its developers are working for 100 % accuracy.
The system is currently undergoing final testing and modification prior to its release as a commercial product. Key to the development of SciFish 2000 was a data acquisition processor (DAP), an analog-to-digital (A/D) converter with an onboard microprocessor, that permits the PC-based system to handle the massive amount of data generated by sonar transmissions. “If we had tried to use Windows for handling all this data, we wouldn’t have gotten real-time readings,” says Skip Denny, chief engineer of sonar systems at SciFish. “With the DAP, a PC becomes a viable platform for this demanding application.”
Existing fishery sonar systems emit narrowband energy in one or two frequencies.
They attempt to classify fish by characterizing their acoustic scattering by measuring the target strength of returned echoes. As such, available fish detection and classification systems work within the time-domain of the signal and focus on the integration of sonar returns (energy) over a given volume of water. Fish detection and classification systems of this type are labor-intensive to build, they require specialized training for operation, the narrowband frequencies are not adequate for all species, and their accuracy and robustness depends on a significant amount of operator training. As a result, they are able to identify fish to species only when large numbers of fish are available over long periods of time or their sizes are sufficiently different.
SciFish 2000 is designed to overcome these limitations. The system is the product of SciFish, an Anchorage, Alaska-based company working to develop technologies and products for more efficient fisheries. SciFish 2000 is based on research done by the Anti-Submarine Warfare (ASW) sector of the defense community that demonstrates that sonar is more effectively processed in the frequency domain, which is only possible using broadband sonar. SciFish’s founders had previously assisted in the development of several ASW systems for tactical, strategic and surveillance missions within the Navy and ARPA. SciFish 2000 represents the first effort toward converting these defense techniques to fisheries.
In addition to identifying fish in real time, it brings some important new capabilities to the fishing industry such as bottom fish detection, bottom type classification, and fish length estimation. The company believes the product will be useful in three sectors of the fishing industry. First, commercial fisheries can use the product to target only the species they wish to catch, improving the efficiency of their operations. Second, the product provides fisheries managers with an improved tool for fish stock and resource assessment. Third, the system will allow fisheries biologists to collect and process their own data.
Technology basis
SciFish combines three underlying technologies to make real-time identification possible. One is broadband sonar transmission. Broadband echoes contain more information because they encompass frequencies that provide greater backscatter within one fish species relative to others. The SciFish 2000 digitally processes the broadband echoes to produce frequency spectra. Spectral processing, the second underlying technology, provides a rich representation not available to existing fish finding sonar systems that use only time-domain processing that counts and integrates echoes. Using spectral decomposition, it is possible to determine which frequencies are most strongly reflected by the fish targets.
The spectral information is presented to a fuzzy neural network classifier, the third underlying technology, for identification. Signatures of fish (fuzzy neural network coefficients) are created by the user for those species of interest. These signatures are stored in a database where they are later recalled for classification. The system is built around a Pentium III PC running Windows NT. The spectral processing and fish identification applications are Windows-based for ease of use.
These technologies require the processing of huge amounts of data. A broadband sonar transceiver generates analog echoes, amplifies the echoes, tunes the echoes for the frequency response of the transducer, and transmits the resulting echo from the transducer.
Each “ping” represents 100,000 data points and typically the system sends one ping per second. The transducer collects the analog echo returns, applies amplification with adjustable gain to the echoes, and bandpass filters the echoes.
The transducer must pass the echoes to an A/D converter capable of sampling at 769k samples per second to satisfy the Nyquist sampling criteria and to achieve sufficient amplitude range and resolution. In addition to collecting all this data, the system must simultaneously process the information.
After the data leaves the A/D converter, the PC performs a number of complicated data analysis steps including producing digital echoes to determine whether the sonar pulse has bounced off an object and returned, performing feature extraction to measure specific characteristics of the echo to create the signature for the fish, and using the feature information in the fuzzy neural network to classify the fish.
SciFish 2000 has been tested in the Great Lakes and in Prince William Sound with a commercial fisherman, and in the Missouri River with research biologists. In these tests SciFish 2000 has demonstrated that it provides superior discrimination of targets at or near the bottom and better than 80% correct species identification on more than 20 species of fish. In some instances the identification performance exceeds 95% correct.
