1. Introduction
The information of Earth’s gravity field is important in geophysics, geodesy, resource exploration, etc. [
1]. In order to accurately determine the information of the gravity field, there are several methods, such as satellite gravimetry, airborne gravimetry, shipborne gravimetry and ground gravimetry [
2]. The satellite gravimetry can achieve the task of determination high precision global gravity field, but it cannot obtain the gravity information in the high frequency band. The ground gravimetry is the initial and the most classical method to survey the gravity field, the shortcoming of which is inefficiency and confined to the terrain. Shipborne gravimetry, just as its name, uses a ship to measure the gravity field information. Among all the gravity measurement methods, airborne gravimetry is an effective method to collect high precision and high resolution large areas gravity field datum [
3].
There are different types of airborne gravimeters to achieve the task of airborne gravimetry and noted that the different principles are used in different airborne gravimetry systems. This means that the stable platform technology is used in GT series gravimeters, AIRGrav gravimeter and Cheken–AM gravity system [
4,
5], while a Strapdown Inertial Navigation System (SINS) is used in SISG and SGA-WZ [
6,
7]. Furthermore, the airborne gravimeter can be divided into scalar gravimeter and vector gravimeter according to the target of measurement. The GT gravimeter and Cheken–AM gravity system are scalar gravimeters, while the AIRGrav gravimeter, SISG and SGA-WZ can obtain the gravity vector information.
The platform gravimeter was firstly used in the gravimetry survey. The GT gravimeter was developed by the GT company (Moscow, Russia) and completed more than 200,000 kilometers scalar gravimetry task all over the world, with a precision of 0.6 mGal (mGal ≈ 10
−5 m/s
2) under the resolution of 3 km [
8]. The AIRGrav gravimeter can measure the three components of gravity field aided by Canadian Gravimetric Geoid 2005 (CGG05) [
9]. The repeatability of vertical components can reach the 0.2 mGal and the repeatability of horizontal was about 2 mGal [
9].
With the breakthroughs of inertial technology, the research on strapdown gravimetry stemmed from the 1990s. Using a math platform instead of physical platform, the strapdown gravimeter has a lower cost, smaller size and simpler structure than the platform gravimeters. In addition, the strapdown gravimeter can directly measure the gravity disturbance vector. The first airborne gravimeter, SISG, was developed by the University of Calgary [
7]. The presented results showed that accuracies of approximately 4 mGal and 6 mGal can be achieved for the vertical and horizontal gravity components by taking advantage of a new algorithm in the inertial frame and wavenumber coefficient filter (WCF) [
10]. Moreover, the navigation-grade SINS could be regarded as a strapdown gravimeter and show 0.5–3.2 mGal precision in scalar gravimetry [
11]. Through thermal calibration and correction, the accuracy of navigation-grade SINS was superior to 2 mGal when used as a gravimeter [
12]. The SGA-WZ series gravimeter is the strapdown gravimeter developed by a National University of Defense Technology. After optimizing the stability and the environment adaptability, the vertical gravity disturbance accuracy new generation of SGA-WZ gravimeter is better than 1 mGal under the resolution of 3 km [
13]. In addition, the France Aerospace Lab has developed an absolute shipborne gravimeter and the precision is superior to 1 mGal [
14].
The scalar gravimetry is almost mature, the accuracy of which can satisfy the requirements of most applications at present [
15]. However, the vector gravimetry still has a great potential in the accuracy promotion for researchers to strive. For example, wavelet decomposition was used to decline the error in the vector gravimetry and the accuracy was about 7 mGal [
16]. In addition, the artificial neural network was also applied into vector gravimetry [
17]. After algorithm optimization, the repeatability of the data collected by Calgary could directly reach 4–8 mGal in three directions, which was improved to 2–4 mGal after applying WCF [
18]. To reduce the negative effect of the gravity vector itself on the measurement, the iterative method was presented in the different research and showed promising results [
19]. The high precision gravity model, for example EGM2008, was introduced into vector gravimetry to correct the low frequency error [
20].
WCF is a widely used method to eliminate the measurement error in gravimetry and former research has shown its excellent performance. However, the application of WCF needs at least two repeated lines and an empirical threshold, which means low efficiency and restrictions. In this study, the backward strapdown inertial navigation algorithm was introduced into vector gravimetry. The backward algorithm was widely used in the fast initial alignment of autonomous underwater vehicle navigation [
21]. When compared with a forward strapdown inertial navigation algorithm, the backward algorithm has the same performance in an ideal condition and different error characteristics in actual sensor error conditions [
22]. The prerequisite of repeat lines in WCF is not needed because the data could be processed separately by the forward and backward algorithm. Based on the different characteristics, the weighted equation from optimal linear smoothing could be taken into application to fuse the results from the forward algorithm and backward algorithm [
23].
The rest of this paper is organized as follows.
Section 2 introduces the backward strapdown inertial navigation algorithm and the principle of strapdown gravimetry. Both the simulations about the different characteristics of two algorithms and the proposed accuracy improvement method are presented in
Section 3. An airborne gravimetry flight test using SGA-WZ02 is used to elaborate the actual performance of the method in
Section 4.
Section 5 discusses possible situations in the application of the method, which include using WCF instead of the variance weighted equation, the performance under different resolutions and the absence of gravity field model information. Finally, conclusions are drawn in
Section 6.
6. Conclusions
The research presented here developed an effective approach for strapdown gravimetry accuracy improvement. The advantages of the method can be summarized as follows:
The backward algorithm not only shows different characteristics in the standalone inertial calculation, but also has different convergence in the Kalman filter, which provides a backtracking result for the vector gravimetry.
Fusion of the data from forward and backward algorithms satisfies the precondition of error compensation and the repeat lines are not needed.
Based on an optimal linear filter, the data can be fused optimally rather than empirically.
The accuracy improvement method was given based on the promising simulation results. Applying to the data from an actual flight test, the improved gravity disturbance vector results can be obtained, especially for the horizontal components. The vertical components is 1 mGal under the resolution of 3 km. After applying the method, the north gravity disturbance component could reach the accuracy of 1.83 mGal and the east gravity disturbance component could reach the 1.80 mGal, both of which is under the resolution of 6 km. For considering the characteristics of the data itself, the optimal linear smoothing algorithm is better than WCF when fusing the results from two algorithms. Furthermore, the repeatability does not decline much with the improvement of resolution. To further test the method, the condition of no gravity field model correction is also presented, in which the method still shows good performance.
For further studies, the accuracy of the gravity disturbance vector obtained from the accuracy improvement method should be compared to the accurate gravity field control data. To fully achieve the goal of high accuracy airborne vector gravimetry, the excellent algorithm and high accuracy hardware system should be applied at the same time.