磁异常探测具有跨介质传播稳定、不受制于环境介质(如水、空气、沙石,泥土)的干扰、不受水文气象条件限制的优势,在工业、军事等领域一直作为研究的热点。依托于无人机平台的磁异常探测具有机动灵活的特点,极大提升了工作效率,适用于大范围区域、复杂地形等极端环境下的探测作业。未爆弹、地下线缆与管道、地下金属废弃物等埋藏物小目标所产生的磁异常有效信号幅值较小,常受到环境噪声的干扰,且无人机动态噪声的引入进一步降低了数据质量。传统的目标检测算法主要应用于数据质量较好、信噪比较高的条件下,在复杂作业环境中无法有效检测出掩埋小目标的存在。文中提出了一种改进型遗传支持向量机(GSVM),对低信噪比下掩埋小目标磁异常信号进行检测,其采用遗传算法对支持向量机模型参数进行优化,提升模型训练效率与泛化性能。在数值实验与现场实验对比中,GSVM 能够有效提升航磁探测中掩埋小目标探测与识别的正确率。

Magnetic anomaly detection (MAD) has the characteristic of stable propagation across different medium. It has advantages that are not influenced by hydrometeorological conditions and environmental medium (such as water, air, sand, soil). As a result, it has been a research hotspot in industrial and military fields. With the combination of Unmanned Aerial Vehicle platform, the working efficiency of MAD is greatly improved, especially in extreme environments. However, the magnetic anomaly signal for small buried targets (such as unexploded bombs, underground cables and pipelines, underground metal wastes) is usually weaker compared with environmental noise, and the effective signal is usually disturbed by environmental noise. Data quality is further affected by dynamic noise introduced by drone platforms. Traditional target detection algorithms are effective in conditions with good data quality and high signal-to-noise ratio. These algorithms failed to detect the existences of buried targets under complex conditions. An improved genetic support vector machine (GSVM) algorithm is proposed to resolve this problem. To improve the efficiency of the proposed algorithm, the hyperparameters are optimized by using a genetic algorithm. The computational efficiency and accuracy of the algorithm are compared with traditional algorithms in the numerical experiments and field data applications. The results prove that the accuracy of detection and identification of buried small targets has been effectively improved by the proposed GSVM algorithm.

TN911.72

国家重点研发计划资助项目(2022YFC3104000)