Feasibility study on an automated intruder detection for tropical fish farm

In this paper, an automated intruder detection system for sea cage fish farm is introduced. Optical imaging method is used to detect possible predator or theft invasion in the fish net area or in the vicinity. In order to accurately alert the operator on an invasion event from tens of camera installed around the feeding nets, an high speed rule based algorithm is tested to identify possible intruders that trespass into the sea cage net area of a local fish farm. A camera system will be installed below the net level of the feeding area. In the early stage, a recorded mode camera system is used to record down the images captured for analysis purpose. The objective is to identify predator, non-fish and the fish categories from looking down position. We employ rule based algorithm that show high tolerances to in-plane rotation, scale variation and out of plane rotation. Various testing images from different scenarios are used in the experiment. The results show that low cost system can be installed using this algorithm to identify the targets in least image processing resources.