U.S. DEPARTMENT OF THE INTERIOR U.S. GEOLOGICAL SURVEY Experiments in In Situ Fish Recognition Systems Using Fish Spectral and Spatial Signatures

This report is preliminary and has not been reviewed for conformity with U.S. Geological Survey editorial standards (or with the North American Stratigraphic Code). Any use of trade, product or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. FIGURE Number Page 1. Comparison of spectral (high uniform reflectivity of one color) and spacial (reflectivities of different colors varying spatially like spots and stripes) signatures of four marine Illustration of a bar pattern and how a Fourier transform power spectrum displays the vertical patterns to frequency bands 10. Illustration of a bar pattern and sine wave showing how a Fourier transform power spectrum displays the vertical patterns to frequency bands along the Y axis.. The objective of this study was to determine if inter-species spectral or spatial variations could be used to develop an automated fish identification system. Comparative tests were conducted using twelve fish species, (chinook salmon, striped bass, American shad, rainbow trout, northern pikeminnow, Sacramento sucker, channel catfish, hardhead, white catfish and white sturgeon) common to the Sacramento River, California. Experiments were performed under optimum laboratory conditions at the Bureau of Reclamation's Fish Research Facilities located near Red Bluff Diversion Dam, Red Bluff, California. Spectral reflectance information was gathered from subject fish in the visible wavelengths range of 400 to 700 nm, using a CCD camera equipped with narrow band pass filters. Images were captured, digitized and stored on computer. The twelve fish species could be divided into three broad (high, medium, and low) reflectivity categories, however, additional independent signature data was needed for in situ specific identification. Additional signatures were developed using basic morphometrics and spatial reflectivities (i.e., spots, stripes, color and reflectance gradient patterns). Morphometric measurements included fish size, pixel number within a mask, body length and depth ratios, and calculations of color moments. Spatial signatures showing promise either singly or in combination for species identification were: moment of color (quantified color gradient changes across the length or depth of the fish), center of color (quantified head and tail colors differing from the central fishes body), Fourier transformation signature of line profiles, and histogram signatures (pixel intensity-number distributions) across the entire fish. These experiments suggest that the technology already exists for an automated, in-situ fish identification system to work under laboratory conditions. Graphic analysis of morphometrics, spectral and spatial signature curves suggested there was differentiation between 10 of the …