基于探测参数的海上目标识别场景生成方法

doi:10.16182/j.issn1004731x.joss.25-0373

摘要/Abstract

摘要：

为解决传统海上目标识别场景生成方法仅考虑了不同环境对所生成场景数据的影响，而忽略了传感器探测参数带来的海上目标识别场景信息变化，使生成的场景缺乏准确性和真实性，提出了一种基于探测参数的海上目标识别场景生成方法。面向海上目标识别任务，分析影响场景生成质量的关键探测参数和场景中的关键场景特征；提出一种基于贝叶斯网络的关联关系建模方法，构建场景特征和探测参数间的映射关系模型，并通过敏感性分析揭示影响探测参数要素的关键场景特征；提出了基于DeepFillv2与泊松融合的场景生成方法，结合得到的关键场景特征，实现了不同探测参数条件下的场景增广。仿真实验验证了所提方法的有效性。

关键词: 探测参数, 场景生成, 贝叶斯网络, 场景特征, 敏感性分析, 泊松融合

Abstract:

Traditional scene generation methods for maritime target recognition consider only the effects of different environments on the generated scene data, while overlooking the changes in scene information caused by sensor detection parameters, resulting in a lack of accuracy and authenticity in generated scenes. To address this issue, a detection parameter-based scene generation method for maritime target recognition was proposed. For the task of maritime target recognition, key detection parameters affecting scene generation quality and essential scene features were analyzed. An association relationship modeling method based on Bayesian networks was proposed to construct a mapping relationship model between scene features and detection parameters. The key scene features affecting detection parameter elements were revealed through sensitivity analysis. Furthermore, a scene generation method based on DeepFillv2 and Poisson blending was proposed, which, combined with the obtained key scene features, realized scene augmentation under different detection parameters. Simulation experiments have verified the effectiveness of the proposed method.

Key words: detection parameter, scene generation, Bayesian network, scene feature, sensitivity analysis, Poisson blending

中图分类号:

TN911.73

润雨瑄,杨德真,刘烨炀等 . 基于探测参数的海上目标识别场景生成方法[J]. 系统仿真学报, 2025, 37(8): 1951-1964.

Run Yuxuan,Yang Dezhen,Liu Yeyang,et al . Scene Generation Method for Maritime Target Recognition Based on Detection Parameters[J]. Journal of System Simulation, 2025, 37(8): 1951-1964.

图/表 14

图1

图2

图3

图4

图5

表1

图6

图7

图8

图9

图10

表2

目标距离参数的映射关系区间覆盖率计算结果

$(f 1) k$	$N (f 1) k$	$N (X 1, f 1) k$	$P k$ /%
平均覆盖率			86.05
(0, 0.001)	50	45	90
(0.001, 0.002)	341	290	85.04
(0.002, 0.003)	293	222	75.77
(0.003, 0.004)	264	229	86.74
(0.004, 0.006)	325	279	85.85
(0.006, 0.008)	279	244	87.46
(0.008, 0.012)	245	205	83.67
(0.012, 0.028)	245	230	93.88

表2

表3

表 4

参考文献 32

[1]	吕闫, 李理, 孙略, 等. 基于抽样概率区间修正的典型源荷时序场景生成方法[J]. 电力系统自动化, 2024, 48(17): 141-150.
	Yan Lü, Li Li, Sun Lüe, et al. Typical Source-load Temporal Scenario Generation Method Based on Modification of Sampling Probability Interval[J]. Automation of Electric Power Systems, 2024, 48(17): 141-150.
[2]	Felski Andrzej, Zwolak Karolina. The Ocean-going Autonomous Ship-challenges and Threats[J]. Journal of Marine Science and Engineering, 2020, 8(1): 41.
[3]	郭虹静, 陶传奇, 黄志球. 意外充分性引导的深度神经网络测试样本生成[J]. 计算机研究与发展, 2024, 61(4): 1003-1017.
	Guo Hongjing, Tao Chuanqi, Huang Zhiqiu. Surprise Adequacy-guided Deep Neural Network Test Inputs Generation[J]. Journal of Computer Research and Development, 2024, 61(4): 1003-1017.
[4]	沈惠康. 基于卷积神经网络的海面弱目标检测方法研究与实现[D]. 扬州: 扬州大学, 2024.
	Shen Huikang. Research and Implementation of Weak Target Detection Method on Sea Surface Based on Convolutional Neural Network[D]. Yangzhou: Yangzhou University, 2024.
[5]	Hampali S, Stekovic S, Sarkar S D, et al. Monte Carlo Scene Search for 3D Scene Understanding[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Los Alamitos: IEEE Computer Society, 2021: 13799-13808.
[6]	Che Yushuai, Ouyang Linhan, Ma Yizhong, et al. A Two-stage Active Learning Kriging Method Based on Portfolio Allocation and Importance Sampling for Structural Reliability Analysis[J]. Structural and Multidisciplinary Optimization, 2025, 68(6): 111.
[7]	Speagle J S. A Conceptual Introduction to Markov Chain Monte Carlo Methods[EB/OL]. (2020-03-07) [2025-04-25]. .
[8]	Zhang Luyao, Zhu Mengtao, Zhang Ziwei, et al. Disentangled Representation Learning for Robust Radar Inter-pulse Modulation Feature Extraction and Recognition[J]. Remote Sensing, 2024, 16(19): 3585.
[9]	Yang Dezhen, Liu Yeyang, Ren Yi, et al. A High-dimensional Dynamic Characteristics-based Boundary Test Scenario Generation Method for UAVs in Dynamic Target Tracking[J]. IEEE Systems Journal, 2025, 19(2): 553-564.
[10]	Jia Lulu, Yang Dezhen, Ren Yi, et al. A Dynamic Test Scenario Generation Method for Autonomous Vehicles Based on Conditional Generative Adversarial Imitation Learning[J]. Accident Analysis & Prevention, 2024, 194: 107279.
[11]	Genovese Giuseppe. Minimax Formula for the Replica Symmetric Free Energy of Deep Restricted Boltzmann Machines[J]. The Annals of Applied Probability, 2023, 33(3): 2324-2341.
[12]	Eslami S M A, Jimenez Rezende D, Besse F, et al. Neural Scene Representation and Rendering[J]. Science, 2018, 360(6394): 1204-1210.
[13]	Vinicius Luis Trevisan de Souza, Bruno Augusto Dorta Marques, Harlen Costa Batagelo, et al. A Review on Generative Adversarial Networks for Image Generation[J]. Computers & Graphics, 2023, 114: 13-25.
[14]	王鑫鑫. 基于图像解纠缠的上色方法研究[D]. 大连: 大连理工大学, 2020.
	Wang Xinxin. Image Colorization Methods Based on Image Disentanglement[D]. Dalian: Dalian University of Technology, 2020.
[15]	杨志林, 尹勇, 张如凯, 等. 面向无人船视觉感知的几何结构保护视差图像拼接[J]. 系统仿真学报, 2024, 36(8): 1749-1763.
	Yang Zhilin, Yin Yong, Zhang Rukai, et al. Parallax-tolerant Image Stitching with Geometric Structure Protection for Unmanned Ship Visual Perception[J]. Journal of System Simulation, 2024, 36(8): 1749-1763.
[16]	Goodfellow I, Pouget-Abadie Jean, Mirza Mehdi, et al. Generative Adversarial Networks[J]. Communications of the ACM, 2020, 63(11): 139-144.
[17]	Verma Gyanendra K. Hybrid DCGAN-CNN Architecture for EEG Based Emotion Recognition in 3D Valence-arousal-dominance Space[J]. SN Computer Science, 2025, 6(5): 463.
[18]	鹿强, 吴琳, 陈昭, 等. 海上目标多源轨迹数据关联综述[J]. 地球信息科学学报, 2018, 20(5): 571-581.
	Lu Qiang, Wu Lin, Chen Zhao, et al. A Review of Multi-source Trajectory Data Association for Marine Targets[J]. Journal of Geo-Information Science, 2018, 20(5): 571-581.
[19]	关键. 雷达海上目标特性综述[J]. 雷达学报, 2020, 9(4): 674-683.
	Guan Jian. Summary of Marine Radar Target Characteristics[J]. Journal of Radars, 2020, 9(4): 674-683.
[20]	张梦妮, 王祎鸣, 吴勇剑. 基于改进Mask R-CNN的船载地波雷达目标检测方法[J]. 海洋科学进展, 2025, 43(3): 693-705.
	Zhang Mengni, Wang Yiming, Wu Yongjian. Target Detection Method of Shipborne Surface Wave Radar Based on Improved Mask R-CNN[J]. Advances in Marine Science, 2025, 43(3): 693-705.
[21]	Huai Songyao, Tim Van de Voorde. Which Environmental Features Contribute to Positive and Negative Perceptions of Urban Parks? A Cross-cultural Comparison Using Online Reviews and Natural Language Processing Methods[J]. Landscape and Urban Planning, 2022, 218: 104307.
[22]	Wei Jiahua, Qiu Jun, Li Tiejian, et al. Cloud and Precipitation Interference by Strong Low-frequency Sound Wave[J]. Science China Technological Sciences, 2021, 64(2): 261-272.
[23]	Lu Peng, Chen Yuze, Chen Ming, et al. An Improved Stacking-based Model for Wave Height Prediction[J]. Electronic Research Archive, 2024, 32(7): 4543-4562.
[24]	董云龙, 刘洋, 刘宁波, 等. 基于雷达方程修正的目标探测距离评估方法[J]. 信号处理, 2022, 38(10): 2102-2113.
	Dong Yunlong, Liu Yang, Liu Ningbo, et al. A Method for Evaluating Target Detection Range Based on Radar Range Equation Modification[J]. Journal of Signal Processing, 2022, 38(10): 2102-2113.
[25]	Yu Jiahui, Lin Zhe, Yang Jimei, et al. Free-form Image Inpainting with Gated Convolution[C]//2019 IEEE/CVF International Conference on Computer Vision (ICCV). Piscataway: IEEE, 2019: 4470-4479.
[26]	伍江江, 李政宏, 沙志超, 等. 结合掩码图像建模和对比学习的道路提取方法[J]. 系统仿真学报, 2025, 37(4): 922-932.
	Wu Jiangjiang, Li Zhenghong, Sha Zhichao, et al. A Method for Road Extraction Using Masked Image Modeling and Contrastive Learning[J]. Journal of System Simulation, 2025, 37(4): 922-932.
[27]	Ma Weizhao, Yin Mengxiao, Li Guiqing, et al. PCMG:3D Point Cloud Human Motion Generation Based on Self-attention and Transformer[J]. The Visual Computer, 2024, 40(5): 3765-3780.
[28]	Sun Yueli, Pan Xingbo, Wu Lin, et al. A Moving-target Imaging Method for Azimuth Multi-channel HRWS SAR System Based on Spectrum Energy Distribution[J]. Remote Sensing Letters, 2025, 16(1): 100-107.
[29]	Saleh M, Elhajj Imad H, Asmar Daniel. Virtual Testing Methods of Cyber-physical Systems: A Framework for Testing Instrumentation and Measurement Systems[J]. IEEE Instrumentation & Measurement Magazine, 2024, 27(8): 11-15.
[30]	Lu Yan, Huang Qingxiang. Underground Coal Mine Target Tracking via Multi-feature Joint Sparse Representation[J]. Journal of Computer and Communications, 2021, 9: 118-132.
[31]	Todd M D. Sensor Data Acquisition Systems and Architectures[M]//Lynch J P, Hoon Sohn, Wang Ming L. Sensor Technologies for Civil Infrastructures. 2nd ed. Cambridge: Woodhead Publishing, 2022: 19-49.
[32]	Srijithesh P P, Gijo E V, Raja P, et al. Leveraging Lean Six Sigma Principles in an Indian Tertiary Care Hospital: A Case Study[J]. International Journal of Quality & Reliability Management, 2025, 42(2): 600-630.

场景特征	归一化敏感性系数	变异系数敏感性系数
像素占比	4.625 0	0.679 506
对比度	4.036 4	0.644 874
偏心率	3.437 5	0.454 068
均匀性	3.784 5	0.591 327

场景来源	NIQE	误差/%	IS	误差/%
真实场景	39.937 5	—	4.452 2	—
消融实验	40.671 1	1.84	3.557 9	20.09
本方法	39.697 9	0.60	4.257 9	4.36
CCGAN	48.357 3	21.08	3.325 3	25.31
DCGAN	63.305 2	58.51	2.660 9	40.23

方法	1σ范围内	2σ范围内	3σ范围内
本方法	79.45	93.15	98.63
CCGAN	72.28	88.04	96.20
DCGAN	59.24	73.91	82.61
消融实验	76.69	94.32	98.24