基于水平卷积核的“张衡一号”EFD数据水平状电磁波扰动自动检测

Automatic detection of “horizontal” electromagnetic wave disturbance in the data of EFD on ZH-1

海量的"张衡一号"观测数据中隐藏了大量的空间扰动事件,不同的空间扰动在电磁数据的时频图上呈现不同的形状.比如人工发射的甚低频电波、电力线谐波以及卫星平台自身的干扰等均在时频图中呈现高于背景的水平谱形态.为了统计分析这些空间扰动事件,进一步分析其物理机制,亟需快速、准确地实现从海量数据中自动提取空间扰动事件的功能.本文利用计算机视觉技术提出了一种水平状空间扰动事件的自动检测算法.首先对电场探测仪(EFD)数据进行傅里叶变换得到时频图;其次,根据时频图像特征,取RGB色彩空间的红色通道作为目标图像;再根据水平状形态特点提出一种新的水平卷积核,自动提取目标图像的水平边缘特征;然后水平特征提取后的图像进行二值化处理,并对二值化结果进行逐行点密度统计,把密度值超过给定阈值的行判定为一条直线,最终实现自动检测水平状空间扰动事件的功能.实验通过在100张时频图上开展大量的实验,结果表明本文提出的算法时间消耗最少,准确率达到99.12%,为进一步分析空间扰动及地震前兆研究奠定基础.

A large number of spatial disturbance events are hidden in the massive “zhangheng-1” observation data. Different spatial disturbances show different shapes in the time-frequency diagram of electromagnetic data. For example, the artificially launched VLF radio waves, power line harmonics and the interference of satellite platform itself all present the horizontal spectrum shape higher than the background in the time-frequency diagram. In order to statistically analyze these spatial disturbance events and further analyze their physical mechanism, it is urgent to realize the function of automatically extracting spatial disturbance events from massive data quickly and accurately. This paper presents an automatic detection algorithm of horizontal spatial disturbance events by using computer vision technology. Firstly, the time-frequency diagram is obtained by Fourier transform of the data of the Electric Field Detector(EFD);Secondly, according to the characteristics of time-frequency image, the blue channel of RGB color space is selected as the target image;Then, a new horizontal convolution kernel is proposed to extract the horizontal edge features of the target image automatically;Then the image after horizontal feature extraction is binarized, and the line by line point density statistics of the binarization results is carried out, and the line whose density value exceeds the given threshold is determined as a straight line, and finally the function of automatic detection of horizontal spatial disturbance events is realized. A large number of experiments are carried out on 100 time-frequency maps, and the results show that the proposed algorithm consumes the least time, and the accuracy rate reaches 99.12%, which lays the foundation for further analysis of spatial disturbance and earthquake precursor.