摘要
针对深海光线分布不均匀导致鱼类识别检测困难的问题,提出了符合视觉认知的多维度深海鱼类识别算法.该方法从时间维度优化传统的高斯混合模型(GMM)初步确定变化区域,从空间维度构建目标特征,完整提取运动目标,从时空关联维度建立深度学习的鱼类识别框架,试验结果表明:本算法可在多种复杂条件下准确提取运动目标,面积交迭度(AOM)达到80%以上,优于当前主流算法.
To solve the difficulty of fish recognition and detection due to the nonuniformly distributed deep-sea light,the multi-dimensional deep-sea fish recognition algorithm was proposed based on visual cognition.The traditional GMM was optimized to initially determine the changing area from time dimension and construct the target features from space dimension for extracting the moving target completely.The fish recognition framework based on deep learning was established from spatio-temporal correlation dimension.The results show that the proposed algorithm can accurately extract moving objects under variously complex conditions.The AOM is more than 80%,which is better than that of current mainstream algorithms.
作者
李晨
刘怡丹
孙科林
李勃
全向前
刘凯斌
LI Chen;LIU Yidan;SUN Kelin;LI Bo;QUAN Xiangqian;LIU Kaibin(Institute of Deep-sea Science and Engineering,Chinese Academy of Sciences,Sanya,Hainan 572000,China;School of Ocean and Earth Science,Tongji University,Shanghai 200092,China)
出处
《江苏大学学报(自然科学版)》
EI
CAS
北大核心
2021年第3期303-308,共6页
Journal of Jiangsu University:Natural Science Edition
基金
三亚市院地科技合作项目(2018YD09)。
关键词
鱼类
认知
运动目标提取
深度学习
识别
fish
cognition
moving target extraction
deep learning
recognition