面向智能船舶的水面小目标检测算法

删除或更新信息，请邮件至freekaoyan#163.com(#换成@)

本站小编 Free考研考试/2024-01-16

-->

梁月翔,冯辉,徐海祥.面向智能船舶的水面小目标检测算法[J].,2021,61(3):255-264

面向智能船舶的水面小目标检测算法

Detection algorithm for small objects on water surface for intelligent ships

DOI：10.7511/dllgxb202103005

中文关键词:智能船舶目标检测卷积神经网络

英文关键词:intelligent ship object detection convolutional neural network

基金项目:国家自然科学基金资助项目(5187921051979210)；中央高校基本科研业务费专项资金资助项目(2019Ⅲ0402019Ⅲ132CG).

作者	单位
梁月翔,冯辉,徐海祥

摘要点击次数:308

全文下载次数:465

中文摘要:

针对智能船舶中基于视觉传感器的水面小目标识别具有识别区域分辨率低、图像模糊、信噪比低等问题，提出了一种新的基于卷积神经网络的水面小目标检测算法——自注意力特征融合检测算法．首先，为了提高视觉信息处理的效率与准确性，在网络模型中引入了自注意力模块，更多关注小目标的细节信息．其次，在网络模型中采用了结构化的特征融合算法，通过多尺度语义信息融合提升对小目标的检测性能．最后，为了解决目标检测的定位问题，在smooth L1损失函数的基础上设计了一种大梯度定位损失函数．通过与传统的Faster R-CNN目标检测算法在船舶数据库上进行仿真对比，验证了所提算法在解决水面小目标检测问题上的有效性．

英文摘要:

Aiming at the difficulty in recognizing small objects on the water surface based on vision sensors in intelligent ships, which are characterized by low resolution of recognition area, blurred image and low signal to noise ratio, a new detection algorithm, self attention feature fusion detection algorithm is proposed for small objects on the water surface based on convolutional neural network. Firstly, in order to improve the efficiency and accuracy of visual information processing, the self attention module is introduced into the network model to pay more attention to the details of small targets. Secondly, a structured feature fusion algorithm is adopted in the network model to improve the detection performance of small targets through multi scale semantic information fusion. Finally, in order to solve the localization problem of object detection, a large gradient positioning loss function is designed based on the smooth L1 loss function. Compared with the traditional Faster R-CNN target detection algorithm in the ship database simulation, the effectiveness of the proposed algorithm in solving the problem of small object detection on the water surface is verified.

查看全文查看/发表评论下载PDF阅读器

关闭

面向智能船舶的水面小目标检测算法

本站小编 Free考研考试/2024-01-16

相关话题/

领限时大额优惠券,享本站正版考研考试资料!