UNDERWATER FISH-VIDEO IMAGES - IMAGE QUALITY AND EDGE-DETECTION TECHNIQUES

Aquaculture needs accurate, economical, manageable and noninvasive health, condition, size, and inventory monitoring systems. To this end, underwater video cameras and recording equipment were used to videotape fish in sea cage and tank environments. Backlit and non-backlit fish images were recorded. The absence of fish handling and the use of existing light rendered this technique noninvasive to fish. The captured images can be used to count fish and assess fish condition. To automate monitoring methods, the fish images need to be separated from their background. Edge detection is difficult because the optical properties of water environments often produce poor quality fish images. Backlit images, however, produced edge information that is distinct. Backlit images were more completely segmented with a LoG operator than with other techniques. Robert's edge detector on backlit fish images has potential to be used on an economical PC-based system in near real-time (video rates: 30 frames per second). Non-backlit fish images are variably-contrasted and/or poorly-resolved. The application of first derivative edge detectors and contrast enhancers did not produce enough significant edge information to segment non-backlit fish images from their associated background. This study made clear the need for further application of image processing strategies or a combination of operators on non-backlit fish images.