Automated Image Processing for Fisheries Applications
Pacific States Marine Fisheries Commission
03/16/2015 - 06/30/2019
Scientists doing fisheries research are increasingly relying on image-based methods for assessing fish populations and studying fish behavior. These methods currently require manual review of video and still image data to extract meaningful information. Automated image processing has the potential to greatly reduce the large amount of time necessary for manual analysis, further improving the value of image-based sampling approaches. While automated image processing methods are well established in biomedical and security applications, software packages capable of automated target detection and identification of fish are not commercially available. This project requires automatic detection, sizing, and classification of fish targets from stereo-video imagery of fish passing on a conveyor belt or sliding on a chute. The project involves controlling image acquisition, developing and applying computer algorithms for image processing, and providing user interfaces and suitable data outputs for operation of software by fisheries biologists. Tasks can be accomplished by applying and modifying classification algorithms developed in computer vision industry, with improvements and adjustments for the specific challenges of fish imagery.