﻿ 基于动态视频信息分析的海上舰船目标检测方法
 舰船科学技术  2022, Vol. 44 Issue (20): 169-172    DOI: 10.3404/j.issn.1672-7649.2022.20.035 PDF

A method of ship target detection based on dynamic video information analysis
YANG Fan
School of Computer, Qinghai Normal University, Xining 810008, China
Abstract: In offshore operations, environment video information is usually obtained through machine vision. Among them, the analysis of ship targets at sea is an important work. Therefore, this study proposes a method of ship target detection based on dynamic video information analysis. First, the difference processing method is used to remove the noise in the dynamic video image, and the fitness is used to enhance the background light. Then the moving object detection area is reduced, and the background model is updated in real time to reduce the impact of the background on the detected object. After dividing the dynamic video images of ships on the sea, the sky and sea antennas are detected to build a texture model of the sea background. Finally, by removing the pixels of the sky and non sea surface texture, the ship target on the sea can be obtained. The experimental results show that the method has high detection accuracy, fast response to dynamic video detection and good denoising effect.
Key words: dynamic video information     target detection     image difference     background model     marine ship target
0 引　言

1 海上舰船动态视频信息预处理和分析 1.1 动态视频图像差分处理

 $D\left( {x,y,\Delta t} \right) = \left| {f\left( {x,y,t} \right) - f\left( {x,y,t - 1} \right)} \right| 。$ (1)

 $D'\left( {x,y,\Delta t} \right) = {\left| {f\left( {x,y,t} \right) - f\left( {x,y,t - 1} \right)} \right|^2} 。$ (2)

 $BD\left( {x,y,\Delta t} \right) = \left\{ \begin{gathered} 255,{\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} D'\left( {x,y,\Delta t} \right) \geqslant t ，\\ 0,{\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} {\kern 1pt} D'\left( {x,y,\Delta t} \right) < t。\\ \end{gathered} \right.$ (3)

1.2 背景模型更新

 ${B_{n + 1}}\left( {x,y} \right) = \alpha {B_n}\left( {x,y} \right) + \left( {1 - \alpha } \right){I_n}\left( {x,y} \right) 。$ (4)

 ${B'_{n + 1}}\left( {x,y} \right) = \alpha {B_n}\left( {x,y} \right) + \frac{{\left( {1 - \alpha } \right){I_n}\left( {x,y} \right)}}{N} 。$ (5)

2 海上舰船目标检测 2.1 确定天空和海天线

 $\begin{gathered} F\left( {u,v} \right) = \\ \frac{{\displaystyle\sum\limits_{x=y = 0}^7 {uv \times } \cos \frac{{\left( {2x + 1} \right)u \text{π} + \left( {2y + 1} \right)v \text{π} }}{{16}}}}{{4f\left( {x,y} \right)}}。\\ \end{gathered}$ (6)

 $E = \frac{1}{{63}}\sum\limits_{i = 1}^{63} {\left| {{A_{{c_i}}}} \right|} 。$ (7)

2.2 提取海面背景纹理

 ${E_k} = {\sum\limits_{{R_k}} {\left( {{A_{ij}} - {{\bar A}_k}} \right)} ^2} 。$ (8)

2.3 海面背景纹理模型构建

 $\left\{ \begin{gathered} l\left( c \right) = \frac{{\displaystyle\sum\limits_{i = 1}^c {{{\left\| {{v_i} - \bar x} \right\|}^2}} }}{{\left( {c - 1} \right)}} \times \frac{{\left( {N' - c} \right)}}{{\displaystyle\sum\limits_{i = 1}^c {\displaystyle\sum\limits_{j = 1}^{N'} {u_{ij}^M} } \left\| {{X_j} - {v_i}} \right\|}} ，\\ \bar x = \frac{{\displaystyle\sum\limits_{i = 1}^c {\displaystyle\sum\limits_{j = 1}^{N'} {u_{ij}^M} } {X_j}}}{{N'}}。\\ \end{gathered} \right.$ (9)

 $u_{ij}^k = \sum\limits_{r = 1}^c {{{\left( {\dfrac{{d_{ij}^k}}{{d_{ri}^k}}} \right)}^{ - \frac{2}{{M - 1}}}}} 。$ (10)

 ${V^{k + 1}} = \sum\limits_{j = 1}^{N'} {{{\left( {u_{ij}^k} \right)}^M}} {X_j} 。$ (11)

2.4 舰船目标检测

3 实验结果与分析 3.1 动态视频图像去噪效果

 图 1 3种方法的去噪能力 Fig. 1 Denoising ability of three methods

3.2 检测效果

 图 2 不同方法的检测效果 Fig. 2 Test results of different methods

3.3 对动态视频图像变化的检测响应速度计算

4 结　语

 [1] 周安然, 谢维信, 裴继红. 傅里叶域内双背景模型的海上红外序列图像目标检测[J]. 信号处理, 2022, 38(1): 164-173. DOI:10.16798/j.issn.1003-0530.2022.01.019 [2] 赵浩光, 王平, 董超, 等. 结合多尺度视觉显著性的舰船目标检测[J]. 光学精密工程, 2020, 28(6): 1395-1403. [3] 杨龙, 苏娟, 黄华, 等. 一种基于深层次多尺度特征融合CNN的SAR图像舰船目标检测算法[J]. 光学学报, 2020, 40(2): 132-140. [4] 盖杉, 鲍中运. 基于深度学习的高噪声图像去噪算法[J]. 自动化学报, 2020, 46(12): 2672-2680.