Image sequence processing to supervise fish passes
applications, fish passage, identification, image sequences, real time, store, tracking process
Theories and Applications
European Assoc. for Signal Processing
This paper presents an automatic system of fish identification and counting using image sequences taken in fish passes. At first, real time is not considered and, so, images have to be stored. We present the electronic system designed to code and store them in real time, on optical disks in files of sequences. Then, we describe the tracking process implemented to count the fish when they cross the pass from down-river to up-river, the recognition process and the overlap problem. A result file can be edited by a human operator. Obtained results are satisfactory with regard to static recognition and are improved by the temporal redundancy generated by the tracking process. The processing time is, on average, lower than the storage time and thus there is no data accumulation.