A Novel Approach for Species Detection from Oceanographic Video

Today for oceanographic research, many computer vision applications can be used in data analysis, indexing of underwater objects and the estimation of the population statistics of marine animals has been done with the help of the Remotely Operated Underwater Vehicles (ROV´s). Scientific observation of the undersea environment is a challenging problem as it offers a potential for continuous observations of more data, but it is restricted by the time and effort. Hence, there is tremendous potential benefit in automating analysis portions. For scientific research, the manual processing of such large amounts of video had been major bottleneck. So, the detection of objects in recorded video is potential interest for human video annotators. To overcome the bottleneck in analyzing ROV drive videos and to anticipate the emerging field represented by fixed underwater observatories, an automated system for detecting animals (objects) visible in the videos has been developed. This paper focuses on the design and development of automated system for detecting the species from the ocean videos and putting the different result based on different parameters as RR, FRR, and FR.