基于分割掩膜和双目视觉的鱼类体长测量方法研究  

Study on Fish Body Length Measurement Method Based on Segmentation Mask and Binocular Vision

作　　者：朱铭劼 廖薇 徐震 田伊林 ZHU Mingjie;LIAO Wei;XU Zhen;TIAN Yilin(School of Electric and Engineering,Shanghai University of Engineering Science,Shanghai 201620,China;School of Mechanical and Automotive Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)

机构地区：[1]上海工程技术大学电子电气工程学院,上海201620 [2]上海工程技术大学机械与汽车工程学院,上海201620

出　　处：《中国农业科技导报(中英文)》2025年第2期116-124,共9页Journal of Agricultural Science and Technology

基　　金：国家自然科学基金项目(62001282)。

摘　　要：为了实现更高精度的水下鱼类体长测量,提出基于分割掩膜和双目视觉的鱼类体长测量方法。通过双目相机进行水下标定后获得水下图像;再使用U2-Net分割出鱼类目标的掩膜,经过立体校正以及SGBM(semi-global block matching)立体匹配算法后生成鱼体表面的点云数据;最后通过在归一化以及点云拟合过程中增加噪声过滤,降低体长数据误差。结果表明,该方法经过水下标定后,精度提升2%左右;增加噪声过滤后,测量结果的误差在4%以内,相较于现有方法降低2%以上。并且,鱼体与相机夹角在60°以内时都能达到较高的测量精度,明显优于现有方法。以上研究结果为水产养殖中的鱼类无接触尺寸测量提供了可行的思路。To achieve higher precision underwater fish body length measurement,a new method based on segmentation mask and binocular vision was proposed.The underwater image was obtained after underwater calibration by the binocular camera,and then U2-Net was used to segment the mask of the fish target.After stereoscopic correction and the SGBM(semi-global block matching)stereoscopic matching algorithm,the point cloud data on the surface of the adult fish body was generated.Finally,the noise filtering was added in the normalization and point cloud fitting process,which was for reducing the error of body length data.The results showed that the accuracy of this method was improved by about 2%after underwater calibration.After noise filtering,the error of measurement results was about 3%,which decreased more than 2%compared with the existing other methods.Moreover,when the angle between the fish body and the camera was less than 60°,a high measurement accuracy could be achieved,which was better than the existing other methods.Above results provided a feasible idea for non-contact size measurement of fish in aquaculture.

关 键 词：双目视觉 U2-Net 三维重建 鱼体测量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]