基于红外特征的空间目标点面一体化智能识别方法
Intelligent Recognition Method for Space Target Point-Surface Integration Based on Infrared Features
刘九齐 1黄海晨 1杨浩东 1梁海朝1
作者信息
- 1. 中山大学 航空航天学院·深圳·518107
- 折叠
摘要
针对空间飞行器博弈场景中飞行器采用传统识别算法并不能很好识别出空间目标群中的高价值目标的问题,提出了一种基于红外辐射特征的空间目标点面一体化智能识别方法.首先,建立空间飞行器博弈的一般场景,并对影响目标飞行器红外特征的所有因素如三维外形、运动特性、表面温度等进行建模.然后,利用计算机计算常见的空间目标的红外辐射强度,并对其进行适当处理后量化和显示为较为真实的红外特征数据.针对传统识别算法难以适用复杂的空间博弈场景,导致对空间目标的识别准确率不高的问题,采用一种基于深度学习的点面一体化智能识别算法,根据红外数据特征的特点对空间目标进行分段识别.最后对提出的算法进行了仿真验证,结果表明,该识别算法具有更高的识别准确率,还避免了传统方法中复杂的特征提取过程,具有更强的鲁棒性.
Abstract
In order to solve the problem that the traditional recognition algorithm is not good at identifying the high value targets in the space vehicle game scene,a point surface integrated intelligent recognition method of space tar-gets based on infrared radiation features is proposed. First,a general scenario for spacecraft games is established,and all factors affecting the infrared characteristics of the target spacecraft,such as their three-dimensional shape,motion characteristics,and surface temperature,are modeled. Then,using a computer,common space target infra-red radiation intensities are computed,quantified,and displayed as more realistic infrared feature data after appro-priate processing. Aiming at the problem that traditional recognition algorithms are difficult to apply to complex spatial game scenarios,resulting in low recognition accuracy of spatial targets,an intelligent recognition algorithm based on deep learning is adopted to segment spatial targets according to the characteristics of infrared data features. Finally,the proposed algorithm is simulated and verified. The results show that the recognition algorithm has higher recognition accuracy,and also avoids the complex feature extraction process in traditional methods,which has stronger robustness.
关键词
红外辐射/深度学习/卷积/图像识别/姿态运动/空间目标Key words
infrared radiation/deep learning/convolution/image recognition/attitude motion/space target引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据