中文摘要：

采用能综合反映水域水质状况的鱼类作为水质生物指示器，通过计算机视觉实时检测生物式水质传感器内的鱼体，采用动态图像理解技术分析鱼类的生态状况指标，如存活量、活动量以及投放药物后鱼类所产生的异常行为等，以判断水质的安全程度；通过传感器的特殊结构设计，自动区分存活的、濒临死亡的、已死亡的鱼，以方便获得鱼类生态状况的视频图像。最后，给出了一种集生物水质检测、机械设计、计算机视觉、动态图像理解及无线网络通信等技术于一体的生物式水质检测传感器的设计方法，并给出实验结果。

英文摘要：

Fish were chosen as a water quality bio-indicator because of their ability to reflect local water quality. The fish bred in a bio-sensor for monitoring water quality were continuously observed by the computer vision system, and their ecological conditions, such as their survival, activity, and abnormal behaviors resulting from drags, were analyzed with the dynamic image understanding technology, so that the safe level of water was investigated. With the special sensor structure designed, living fish, dying fish and dead fish were automatically separated, so that the video images reflecting fish＇ s ecological conditions were easily captured. Finally, the method of designing bio-sensors for monitoring water quality was proposed, which integrated biological water quality monitoring, mechanism design, computer vision, dynamic image understanding, as well as wireless network communication technologies. The experiment results were also given.