外文摘要：Monitoring and evaluating fish growth and health relies on accurate aquaculture body size and weight data. Pearl's gentian grouper has tender flesh and rapid growth, and the scale of aquaculture is increasing yearly. Moreover, it is inefficient to measure body size and weight data manually regularly, so in this paper, taking pearl gentian grouper as the research object, we propose a body size and weight estimation method based on machine vision and improve DeepLabV3 + algorithm, with the MIoU reaching 97.37% on the self-constructed dataset, to extract the closed contour of the fish body curve, then apply the algorithm for measuring body scale traits. This paper determines the critical points of body scale calculation. It uses the calibration coefficients to calculate the body scale parameters and then estimate fish body weight. In the established body weight estimation model, the coefficient of determination of the CatBoost regression model based on the total length, body height, body thickness, and area of the fish was R2 of 0.987 which realized the high-precision estimation of the body weight of the pearl gentian grouper and provided the basis for intelligent grading.
外文关键词：semantic segmentation;image processing;Machine vision;Pearl gentian grouper;Body measurement point extraction
作者：Cong, Xueqi;Tian, Yunchen;Quan, Jianing;Qin, Haijing;Li, Qingfei;Li, Ruipeng
作者单位：Tianjin Agr Univ;Tianjin Key Lab Aquat Ecol & Aquaculture
期刊名称：AQUACULTURE INTERNATIONAL
期刊影响因子：0.0
出版年份：2024
出版刊次：32(4)
原文传递申请：江苏省科技资源（工程技术文献）统筹服务平台