Fish Observation, Detection, Recognition and Verification in The Real World

The purpose of this paper is to present fish observation, detection, recognition and verification for processing video stream data in the real world. A distributed real-time high-definition underwater video stream system has been demonstrated in Taiwan for long-term fish observation. End users can real-time observe the high-definition underwater ecological environment via Internet. These video data is preserved to form a resource base for marine biologists. Based on the video data, fish detection is implemented. However, it is complicated in the unconstrained underwater environment, due to the water flow causes the water plants sway severely. In this paper, a boundingsurrounding boxes method is proposed to overcome the problem. It efficiently classifies moving fish as the foreground objects and the swaying water plants as the background objects. It enables to remove the irrelevant information (without fish) to reduce the massive amount of video data. Moreover, we can acquire the images of multiple species of fish with varied angles, shapes, and illumination to construct a fish category database. Sparse representation-based classification (SRC) based on compressive sensing is shown to be robust for face recognition in recent years. We propose a maximum probability of parting ranking method based on the framework of SRC for fish recognition and verification. Experimental results show that the data volume is reduced greatly, and fish recognition and verification are able to achieve high accuracy.