摘要
抛撒地雷是一种布设于地表的杀伤性武器,如何精准地检测出抛撒地雷具有重要的军事意义。考虑到高光谱图像具有丰富的空谱信息,因此基于高光谱图像进行地雷检测能够很大程度上提高检测的准确性。然而高光谱数据往往存在采集困难、成本高以及抛撒地雷真实数据少等特点,严重制约了基于高光谱抛撒地雷检测的研究。因此,提出了一种高光谱抛撒地雷的图像仿真方法并将其应用到异常检测任务中:首先,通过仿真软件选择的抛撒地雷型号确定仿真数据的尺寸、布设范围以及插雷的个数,并人工设定光谱融合系数;其次,分别从空间维度和光谱维度出发对仿真数据进行数据增强,并将其融合到不同背景的高光谱图像中,获得抛撒地雷仿真数据集;最后,采用RXD异常检测算法分别对仿真数据集和真实数据图像进行异常点检测,验证仿真的准确性和真实性。实验表明,仿真数据集异常检测的准确率约为82.67%,真实数据图像的准确率约为90.28%。
This paper proposes an image simulation method for hyperspectral scattered mines and applies it to anomaly detection tasks.Firstly,the size,deployment range and the number of inserted mines of simulation data are determined by the type of scattered mines selected by the simulation software and spectral fusion ratio is manually set.Secondly,the simulation data is enhanced from the spatial dimension and spectral dimension respectively and fused into the hyperspectral images with different backgrounds to obtain the simulation dataset of scattered mines.Finally,the RXD algorithm is used for anomaly detection both in the simulation dataset and real data images to verify the accuracy and authenticity of the simulation.The experiment shows that the accuracy of anomaly detection is 82.68%in the simulation dataset and 90.28%in the real data image.
出处
《工业控制计算机》
2023年第8期103-104,107,共3页
Industrial Control Computer
关键词
高光谱异常检测
抛撒地雷
仿真数据集
hyperspectral anomaly detection
scattered mines
simulation dataset
