﻿ 基于大数据分析的港口停泊船舶识别方法
 舰船科学技术  2023, Vol. 45 Issue (13): 158-161    DOI: 10.3404/j.issn.1672-7649.2023.13.032 PDF

Methods of berth ship identification based on big data analysis
LING Hai-sheng
Zhejiang Institute of Communication, Marine Department, Hangzhou 311112, China
Abstract: In order to solve the problems of complex background and large data volume of port image, which affect the efficiency of ship identification, a method of berth ship identification based on big data analysis is studied. Select Hadoop cloud computing framework and use distributed processing method to process massive port data in parallel. Hadoop uses the data analysis platform to run the Top-hat watershed segmentation method, which uses the local minimum value of the port image to determine the watershed position, calculates the gradient function of the watershed position, and divides the port image according to the threshold value set. The segmented image is used as the input of the convolutional neural network. The convolutional neural network performs convolution and pooling operations on the image to output the results of berth ship recognition in port. Experimental results show that the proposed method can accurately identify ships in port under different backgrounds, such as sunny, rainy and so on, and the recognition time is less than 40 ms.
Key words: big data analysis     to lay in port     ship identification     Hadoop     watershed segmentation     convolutional neural network
0 引　言

 图 1 Hadoop的港口停泊舰船识别结构图 Fig. 1 Structure diagram of port berth ship identification in Hadoop

1.2 基于Top-hat分水岭分割的港口图像分割

 $\begin{split} G\left( {x,y} \right) =& {\rm{grad}}\left( {f\left( {x,y} \right)} \right) = \\ & {\left\{ {{{\left[ {f\left( {x,y} \right) - f\left( {x - 1,y} \right)} \right]}^2} + {{\left[ {f\left( {x,y} \right) - f\left( {x,y - 1} \right)} \right]}^2}} \right\}^{\frac{1}{2}}} \text{，} \end{split}$ (1)

 $G\left( {x,y} \right) = \max \left( {{\rm{grad}}\left( {f\left( {x,y} \right)} \right),T} \right)\text{，}$ (2)

 $G \circ b = \left( {G\Theta b} \right) \oplus b\text{。}$ (3)

1.3 卷积神经网络的港口停泊船舶识别

 $X_j^l = f\left( {\sum\limits_{i = 1}^d {X_i^{l - 1}} w_{ij}^l + b_j^l} \right)\text{。}$ (4)

 图 2 卷积神经网络结构图 Fig. 2 Convolutional neural network structure diagram

 $X_j^l = f\left[ {B_j^lC\left( {X_j^{l - 1}} \right) + b_j^l} \right]\text{。}$ (5)

2 实例分析

 图 3 原始港口图像 Fig. 3 Original port image

 图 4 港口图像分割结果 Fig. 4 Port image segmentation results

 图 5 港口停泊船舶识别结果 Fig. 5 Results of berth ship identification in port

3 结　语

 [1] 赵龙飞, 姜晓轶, 孙苗, 等. 面向海运统计的AIS大数据挖掘分析研究[J]. 海洋科学, 2021, 45(12): 55-64. ZHAO Longfei, JIANG Xiaoyi, SUN Miao, et al. AIS big data mining for maritime statistics[J]. Marine Sciences, 2021, 45(12): 55-64. [2] 付哲泉, 李尚生, 李相平, 等. 基于高效可扩展改进残差结构神经网络的舰船目标识别技术[J]. 电子与信息学报, 2020, 42(12): 3005-3012. [3] 马啸, 邵利民, 卢惠民, 等. 一种基于视觉感知的舰船目标智能化识别方法[J]. 电讯技术, 2020, 60(10): 1133-1141. MA Xiao, SHAO Limin, LU Huimin, et al. An Intelligent Ship Targets Recognition Method Based on Visual Perception[J]. Telecommunication Engineering, 2020, 60(10): 1133-1141. [4] 孙嘉赤, 邹焕新, 邓志鹏, 等. 基于级联卷积神经网络的港口多方向舰船检测与分类[J]. 系统工程与电子技术, 2020, 42(9): 1903-1910. [5] 关欣, 国佳恩, 衣晓. 基于低秩双线性池化注意力网络的舰船目标识别[J]. 系统工程与电子技术, 2023, 45(5): 1305-1314. GUAN Xin, GUO Jiaen, YI Xiao. Ship target recognition based on low rank bilinear pooling attention network[J]. Systems Engineering and Electronics, 2023, 45(5): 1305-1314.