copepods in the laboratory) to 5 meters (6-gill sharks in the 

 Atlantic at 6,000 ft.). This is a major breakthrough in 

 underwater imaging, but the nature of our approach to image 

 analysis (which sacrifices information content to track in real 

 time) limits our use of the original stereopaired video tapes. 

 We cannot now extract photogrammetric data from the tapes for 

 automated measuration because our motion analysis program must 

 reduce successive images of each individual to a series of points 

 (centroids) moving in 3-D space. For example, although we can 

 track the movements of the 6-gill sharks around baits of tuna at 

 6,000 ft. with great precision using 3-D cameras mounted on the 

 Pisces submarine, we do not have the computer capability for 

 automated measurements of body lengths, shapes, and surfaces 

 needed in order to compare the individual sharks recorded on the 

 tapes. In short, the digitized data taken from the stereometric 

 paired tapes cannot distinguish whales from submarines nor can 

 one measure the sizes or plot the shapes of those objects, which 

 is the starting point for all modern 3-D image analysis. We 

 intend to develop this capability in the future because 

 measurement of objects is absolutely critical for all aspects of 

 underwater science, from analysis of the behavior of fish to 

 mapping the shape of natural or man-made objects. 



Image analysis is now accomplished automatically by a 

 computer through almost instantaneous "frame grabbing" 

 techniques. Still, each data set is very large and the digitized 

 information from the original images quickly fills the memory 

 even of large computers. Intermediate sized minicomputers 

 commonly frame-grab 2 to 4 video images at once, but complete 

 analysis of each image can last minutes or hours depending on the 

 analytical task required. The result is that if one wants 

 complete image analysis ( as is required for any photogrammetric 

 application) one must sacrifice real-time processing in order to 

 obtain a complete pixel by pixel analysis. It is important to 

 stress that the original video sequences from which we obtain 

 information on movements of marine animals in 3-dimensional 

 space are recorded under water with stereometric video cameras, 

 aligned and calibrated in such a way that we could also use the 

 full information content of the tapes for photogrammetric 

 reconstruction. Several image analysis computers are now 

 available that apparently can do these tasks. 



Information Analysis 



Most biologists are unfamiliar with the mathematics involved 

 in 3-D statistics and 3-D reconstruction, but these are commonly 

 used in the discipline of x-ray crystallography. We will use 

 these tools for reconstruction and predictive modelling of 

 aggregation and behavior in three-dimensional space, and use 

 computer animation to test the applicability of our models. We 

 will discuss this aspect of the subject in a later publication 

 ( Prewitt and Hamner, in preparation). 



315 



