"The second intrinsic problem... is the analysis of motion. 

 Continuous motion can be reduced to discrete events by a 

 still camera, [but] . . .meaningful analysis of changes in 

 position requires a more dynamic record, such as that 

 provided by video or motion picture cameras... A scientist 

 with a movie camera can record in a few seconds more raw 

 information on film about the motion of a single object 

 than he can quantify and analyze by hand in months..." 



"A third intrinsic problem is that of the third spatial 

 dimension. A frequent simplifying assumption is that 

 oceanic organisms can be studied as a two dimensional 

 problem just as one might study land-based animals. But 

 except in the shallowest of water the spatial arrangement of 

 [oceanic animals] ... is inherently three-dimensional .. .Thus, 

 the proper study of [oceanic animals] ... involves determining 

 the positions of individual organisms in three dimensions." 



"A fourth problem involves the necessity on occasion to 

 analyze each picture completely for its full information 

 content. In classical signal-processing-based computer 

 image analysis a photograph or video image must be digitized 

 point for point or pixel by pixel. For a standard JVC video 

 chip this requires some 500,000 entries, each evaluated in 

 terms of pixel density (commonly 98 shades of grey). The 

 image then must be processed for noise removal, edge 

 detection and edge linking, statistics and shape metrics, 

 and image classification. Obviously such a task is 

 impossible to accomplish manually because of the time 

 required to enter such an enormous data set physically into 

 the computer. At present only the largest and most 

 expensive computers such as the Cray can do full -image 

 analysis in real time. Minicomputers can accomplish all 

 of the tasks required but must deal with them sequentially 

 and much more slowly because of the magnitude of the 

 computations involved." 



Automated Tracking of Objects in 3-D Space in Real Time: 

 Sacrifice of Information for Time 



Historically, two computing systems have been used to track 

 behavior of marine or fresh water organisms. The system in use 

 at the University of Chicago, called the Galatea system, can 

 resolve all of the problems noted previously (Potel et al., 1979; 

 1980), but the Galatea system is limited by the fact that the 

 data still must be manually digitized. (Also it is one of a kind 

 and not available commercially. ) However, one cannot afford to 

 rely on operator data input. If one were to manually digitize 

 the movements of 20 fish in three-dimensional space, and record 

 their positions via stop-frame analysis at 30 frames per second, 

 it would require over a half-million file entries to digitize 10 

 minutes of data. At one entry every three seconds (far faster 



313 



