Processing video images in real time has become feasible but still requires comparatively expensive technology in regard to both image acquisition and image processing. This type of processing would be inconsistent with our overall objective of facilitating technology transfer by constructing comparatively inexpensive samplers with largely off-the-shelf components. Using an approach we have termed "smart-sampling", however, we have now incorporated video imaging into our strobe silhouette-illuminated zooplankton sampling. Resolution at comparable fields of view are reduced (ca. 40 Ám for the deck unit and 100 Ám for the in situ unit) but still useful.
In collaboration with Drs. Checkley, Scripps Institution of Oceanography, and Herman, Bedford Institute of Oceanography, we have used the Optical Particle Counter (OPC) invented by Dr. Herman (Herman et al., 1988) to flag possible targets before they pass through the field of view of a video silhouette strobe system. Drs. Davis and Gallagher of the Woods Hole Oceanographic Institution (see GLOBEC NEWS No. 3 for a description of their Video Plankton Recorder-VPR), greatly assisted us in the initial selection of video components. In the smart-sampler, the OPC detects and sizes incoming particles. If they fall within a user-specified size range the individual video frame number is then recorded and a digital flag encoded on the video tape. Using an inexpensive framegrabber, a desktop microcomputer and commercial image-processing software, images of interest can be processed, their features extracted, compressed and stored at rates up to 0.5 Hz upon optical disks.
Integrating the instruments reduces the data flow often more than a hundredfold and simplifies data processing proportionally. The OPC (not the video camera) is the primary sampler. It counts and sizes every single particle encountered (up to 200 Hz), while the video images are used predominately for target identification (taxonomy) of the OPC size classes. Comparatively large amounts of data can be efficiently processed. Figure 2 depicts a transect of ca. 32 km that reaches an acute midshelf front. At the front the abundance of small copepods (the dominant particle type) plummets. Using the computer-encoded video tape and the synchronized OPC datafile we can regenerate another set of images for a size class not originally selected.
As in our in situ photographic silhouette system, strobe illumination is absolutely essential in eliminating blurred images because the organisms move rapidly through the field of view. Moreover it faciliates seeing near transparent objects like teleost eggs. Because we are actually imaging the shadow of objects comparatively near the light source rather than the light reflected from distant objects, we can use a low power strobe of our own design which has facilitate packaging the unit for in-situ application. Finally, since the video silhouettes are used primarily for target identification with the "smart-sampling" approach it is unnecessary for many purposes to image every particle - merely a random subsample. This allows us to strobe at the video frame interval rate (30 Hz) rather than the field rate (60 Hz) so that the two fields representing a single exposure can be integrated which doubles the number of independent horizontal lines in the pixcell array.
A far better technical solution for a "smart-sampler" would be to use a "snapshot" CCD camera/strobe (akin to our original photographic system but substituting asynchronous for synchronous frame acquisition). Such a fully digital system would free us from the inherent constraints of video technology (the digital to analogue and back to digital conversions). We hope to test this alternative within the next year. Regardless of how plankton images are obtained, it is essential for the various workers in this field to collaborate in establishing a standard protocol for efficient image analysis of plankton "cartoons" (see Figure 1). Purely statistical classification of randomly oriented objects has proven to be less than fully satisfactory. Various alternatives are currently being explored including the applicability of deformable template image analysis techniques.
Herman, A.W. 1988. Simultaneous measurement of light attenuance with a new optical particle counter. Cont. Shelf Res., 8, 205-221.
Ortner, P. B., L. C. Hill, and H. E. Edgerton. 1981. In-situ silhouette photography of Gulf Stream zooplankton. Deep Sea Res., 28, 1569-1576.