The continuity of flowing water is in demand and the fish bypasses are considered to be important points for fish migration and the entire ecosystem. Current solutions, however, do not fulfill the goal of natural fish migration. Within the interdisciplinary concept Ethohydraulics, which stands for Ethology and Hydraulics, the functional capability of fish bypasses should be improved. The goal of this research is to find a reasonable ethohydraulic signature using computer-vision technology.
For this purpose, a three-dimensional tracking system is developed for fish behavior analysis. First, the depth images are generated using an industrial camera (Fotonic P60UA) applying technology of structured light. This camera emits infrared light into space with a pattern template and then receives the deformed pattern back. After decoding, the distance between the camera and the object can be measured as depth value. Second, the depth images are preprocessed for extract fishes as 3D objects. Under some circumstance, the fishes can be only captured partly. Therefore, we used a three-points-definition model for a fish.
After extracting the fishes, we proposed a multi-tracking method based on Kalman Filter (KF). We set the state of the system with 18 elements, which response to the positions and velocities in three dimensions of the three points. For each fish object, a KF is set and traced in the living period. To associate the candidates of KFs, Hausdorff-Distance is applied. In the end, the tracking results will be combined with 3D current data in order to to find out the relation between fish behavior and water flows.
This project is supported by the FORIN-Project from the Department of Civil and Environmental Engineering Sciences and cooperated with Institute for Hydraulic and Water Resources at TU Darmstadt.