Land-based measurement method of underwater fish body length based on binocular vision
The traditional manual measurement method not only makes the fish suffer from stress,but also has low measurement efficiency,and the non-contact measurement method can serve these shortcomings well.In this study,the top view images of underwater fish are collected by a binocular camera placed above the water surface.The GrabCut algorithm combined with contrast adaptive adjustment is used to achieve selective segmentation.Skeleton extraction method is used to straighten the fish bending of the fish body,and correct the influence of water refraction.Finally,the fish's body length measurement points after segmentation are extracted and measured.The experimental results show that compared with the original GrabCut algorithm,the proposed image segmentation algorithm improves the segmentation accuracy by 14.9％,reduces the oversegmentation rate by 23％,and improves the time efficiency by 11.6％.The average relative error in the final measurement of fish body length is 0.91％,which is an improvement of 0.37％over the accuracy without considering refraction.The non-contact land-based measurement method proposed in this study meets the measurement accuracy requirements.It is more accurate and convenient than the underwater fish body measurement method,and is more suitable for engineering implementation and application.
binocular visionfish body length measurementGrabCut segmentationfish bending straighteningwater refraction
万方数据
维普
