Estimating Mass of Harvested Asian Seabass Lates calcarifer from Images  

作　　者：Dmitry A. Konovalov Alzayat Saleh Jose A. Domingos Ronald D. White Dean R. Jerry 

机构地区：[1]College of Science and Engineering, James Cook University, Townsville, Australia [2]James Cook University Singapore, Singapore [3]Centre for Sustainable Tropical Fisheries and Aquaculture, James Cook University, Townsville, Australia

出　　处：《World Journal of Engineering and Technology》2018年第3期15-23,共9页世界工程和技术（英文）

摘　　要：Total of 1072 Asian seabass or barramundi (Lates calcarifer) were harvested at two different locations in Queensland, Australia. Each fish was digitally photographed and weighed. A subsample of 200 images (100 from each location) were manually segmented to extract the fish-body area (S in cm2), excluding all fins. After scaling the segmented images to 1mm per pixel, the fish mass values (M in grams) were fitted by a single-factor model (M=aS1.5, a=0.1695 )achieving the coefficient of determination (R2) and the Mean Absolute Relative Error (MARE) of R2=0.9819 and MARE=5.1%, respectively. A segmentation Convolutional Neural Network (CNN) was trained on the 200 hand-segmented images, and then applied to the rest of the available images. The CNN predicted fish-body areas were used to fit the mass-area estimation models: the single-factor model, M=aS1.5, a=0.170, R2=0.9819, MARE=5.1%;and the two-factor model, M= aSb, a=0.124, b=0.155, R2=0.9834, MARE=4.5%.Total of 1072 Asian seabass or barramundi (Lates calcarifer) were harvested at two different locations in Queensland, Australia. Each fish was digitally photographed and weighed. A subsample of 200 images (100 from each location) were manually segmented to extract the fish-body area (S in cm2), excluding all fins. After scaling the segmented images to 1mm per pixel, the fish mass values (M in grams) were fitted by a single-factor model (M=aS1.5, a=0.1695 )achieving the coefficient of determination (R2) and the Mean Absolute Relative Error (MARE) of R2=0.9819 and MARE=5.1%, respectively. A segmentation Convolutional Neural Network (CNN) was trained on the 200 hand-segmented images, and then applied to the rest of the available images. The CNN predicted fish-body areas were used to fit the mass-area estimation models: the single-factor model, M=aS1.5, a=0.170, R2=0.9819, MARE=5.1%;and the two-factor model, M= aSb, a=0.124, b=0.155, R2=0.9834, MARE=4.5%.

关 键 词：AQUACULTURE ASIAN SEABASS BARRAMUNDI Lates calcarifer Computer Vision Image Processing WEIGHT Estimation 

分 类 号：R73[医药卫生—肿瘤]