Distributed acoustic sensing (DAS) has been widely applied in railway safety monitoring, perimeter security, seismology, and other fields. The high precision target source multi-dimensional localization is important for these applications. However, most implementations of DAS provide the position of detected sources as a function of distance within the one-dimensional axial space along the sensing fiber, and the transversal distance between the detected sources and the sensing fiber is unclear, which hinders the process of DAS practical applications.
The current target source localization methods can be divided into two categoriesone is based on time difference of arrival (TDOA) algorithm, and the other is based on array signal processing (ASP) method. The ASP methods include beamforming and spatial spectrum estimation. The positioning accuracy of TDOA algorithm is poor, and the beamforming method often requires high signal-to-noise ratio, large array aperture, and a priori knowledge of the environmental noise and target source which is difficult to obtain accurately.
The spatial spectrum estimation method is based on the orthogonal property of signal subspace and noise subspace, and has high estimation accuracy and angle resolution. Cao et al. used multiple signal classification (MUSIC) algorithm to locate the underwater near-field target source with an error of 0.7 m, and the ratio of signal wavelength to detection aperture was 25∶1. Liang et al. realized sound source location in the air medium by wrapping the optical fiber densely around the cylindrical cavity structure, and the ratio of signal wavelength to detection aperture was 100∶1. In these studies, the channel detection aperture is much smaller than the signal wavelength and sensor can be regarded as a point sensor. However, DAS is limited by the spatial resolution, and the single-channel detection aperture of the existing optical cable is 10 m, which is comparable with the target signal wavelength, so it is difficult to directly use the spatial spectrum estimation method to achieve multi-dimensional target source localization. In addition, the channel aperture compression requires a special design of the sensor unit, and the structure is complex, so it is not easy for large-scale application.
In this paper, we propose a multi-dimensional target source localization method for DAS by correcting fiber array phase deviation. The proposed method can eliminate the influence of DAS large detection aperture, and the high precision target source multi-dimensional localization can be obtained by common optical cables.
To eliminate the influence of DAS large detection aperture, the phase correction method is proposed. First, the DAS sensing channel response is analyzed and the phase deviation between DAS equivalent array and distributed uniform linear array (ULA) is calculated. Then, the TDOA algorithm is used to obtain pre-estimation location of target source for array phase correction. The effects of sensing channel number and position on TDOA estimation are studied. Multiple sensing channel groups are used for TDOA estimation, and the final pre-estimation location is the average value of estimation results of all those groups. After that, the corrected signal is used for spatial spectrum estimation by MUSIC method, and a higher precision target source localization can be obtained. Then, the array phase is corrected according to the MUSIC estimation and the MUSIC algorithm is iterated. The effect of MUSIC algorithm iterations on the root-mean-square error (RMSE) is studied.
The proposed method can realize multi-dimensional localization of target source, and the preliminary experiment verifies that the minimum RMSE of localization result is 1.1 m. The proposed target source localization method contains three major processing stages: localization pre-estimation, array phase correction, and high precision localization. The pre-estimation accuracy of TDOA is uncorrelated with the number of sensing channels, and the TDOA pre-estimation of the sensing channels at different locations is quite different, which may be related to the uneven transmission medium and inconsistent cable deployment conditions. Multiple sensing channel groups are used for TDOA pre-estimation, and the final pre-estimation location is the average value. The result of TDOA pre-estimation is (29.7 m, 58.1°), and the RMSE is 5.7 m. The DAS detected phase is corrected according to the pre-estimation location, and the corrected DAS detected phase is used for spatial spectrum estimation by MUSIC method to obtain a multi-dimensional target source localization. In the experiment, the RMSE of localization result can be effectively reduced by increasing the iterations of MUSIC algorithm. When the iteration number is increased to three, the RMSE reaches a minimum value, and a high precision target source multi-dimensional localization result can be obtained. The final localization result is (28 m, 67.5°), and the RMSE is 1.1 m. The proposed method enables that the detected signal of DAS sensing channel can be accurately located using ASP method directly. Moreover, compared with TDOA pre-estimation, the localization accuracy is greatly improved.
In the present study, a multi-dimensional target source localization method for distributed acoustic sensing is proposed, which is suitable for common communication fiber in a wide range of applications. Due to the large detection aperture of DAS, there is a phase deviation between DAS equivalent array and uniform linear array. The DAS detected phase is corrected by the proposed phase correction method, and the target source can be accurately located using ASP method without shrinking the sensing channel aperture. The principle of array phase deviation is analyzed and the feasibility of the proposed localization method is preliminarily verified. Compared with previous DAS target multi-dimensional localization studies, the proposed method does not require special structures to wind the optical fiber and shrink the sensing channel aperture, greatly simplifying the system complexity. The RMSE of localization result can be effectively reduced by increasing the iterations of MUSIC algorithm. The proposed method provides a simple and effective means for DAS target source multi-dimensional localization. It is believed that the proposed method will improve DAS localization performance in actual applications, such as intrusion detection and earthquake monitoring.
