Abstract
A fish image automonitoring system for detecting acute toxicity in water purification plants was studied experimentally. A stable image taking technique for fish was developed based on a lighting method, in which a translucent backscreen behind the water tank was illuminated and then scattered by the illumination lamp. An effective fish image recognition method, regardless to the degree of turbidity, was found, which was based on the histogram of the gray image.
The image automonitoring technique, in which both the location profile and the velocity profile of a fish swimming in the water tank was applied to evaluation of fish movement patterns. The normal movement patterns of carp (Cyprinus carpio), crucian carp (Carassius carassius buergeri), and broadstriped biteling (Rhodeus osellatus osellatus) differed from those under toxic conditions (cyanide ion= 0.1 mg·l-1). Then toxicity could be detected on the basis of differences in fish movement patterns as determined by the image automonitoring system.
