A full-polarimetric GPR system and its application in ice crack detection

The warming season of Antarctica, where sea ice melts and breaks, is often the golden time for Antarctic research. But this season is also a high frequency season for the formation of ice crack, the existence of which poses a serious threat to Antarctic scientific research. Therefore, it is of important significance to monitor the trend, location, and depth of Antarctic sea ice crack in real time. Ice crack detection often uses ground penetrating radar (GPR) technology. Compared with the single-polarimetric GPR, full-polarimetric GPR can obtain more comprehensive polarization information. Therefore, a full-polarimetric GPR system based on the horn antenna is built in this paper to obtain more abundant ice crack information. Then we apply it to detect a small-scale ice crack atop frozen lakes. The analysis of the experiments and data fusion processing results can verify the effectiveness of the system for the detection of ice crack and lay the foundation for subsequent actual detection.


Introduction
The warming season of Antarctica, when sea ice melts and breaks, is often the golden time for Antarctic research, but this season is also a high-frequency season for the formation of ice crack [5]. For more than 30 years, China's Antarctic scientific research has used snowmobiles to replenish supplies almost every year, but many dangerous situations have occurred because of the ice crack [3]. Therefore, the rapid and real-time monitoring of the trend, location, and depth of the Antarctic ice crack is of important significance. In addition, the prediction of the location and depth of the ice crack is also important for understanding the disintegration process of Antarctic glaciers [1].
GPR is a new technology that has been approved by the scientific community, and it is very suitable for ground-based surveys and ice crack detection [9]. GPR is also an effective non-destructive tool for detecting shallow underground based on electromagnetic reflection information. Polarization technology is one of the most important technologies in the field of microwave remote sensing during recent decades [4,13]. With the rapid development of polarization technology, it has also been introduced into the field of GPR to improve its detection capability [2,18]. In order to obtain richer information about ice crack, we choose the full-polarimetric GPR for research. In recent years, major research institutions and universities have constructed various GPR systems and applied them in different fields [16]. Kong et al. developed a GPR system which uses step frequency signals [14]. Daniels et al. developed multi-channel GPR systems for mine detection [6]. Plettemeier et al. developed a full-polarimetric GPR antenna system for the mission of water ice and subsurface deposit observations on Mars [17]. Li Lili developed a full-polarimetric GPR acquisition system and calibration technique [15]. Dong Zehua et al. developed a vehicle-mounted GPR system for rapid detection of asphalt pavement thicknesses [7]. Gianluca et al. developed a low frequency airborne GPR system for wide area geophysical surveys [12]. While continuously developing and improving the performance of the hardware system, it also performs GPR antenna performance analysis, target body attribute analysis, target body recognition and etc. [10,11]. Therefore, we built a full-polarimetric GPR system in this paper to achieve a faster and more accurate assessment of ice crack.

Composition of full-polarimetric GPR system
The full-polarimetric GPR system built in this paper is mainly composed of a PC control unit, a vector network analyzer, a switch driver, and two horn antennas, which is shown in Figure 1. The PC control unit is mainly responsible for controlling the vector network analyzer and the switch driver. The vector network analyzer is used to generate electromagnetic wave signals and receive echo signals from the antennas. The switch driver sends opening and closing instructions of the corresponding channel to the coaxial switch by receiving a command from the PC control unit. The horn antennas are responsible for transmitting the electromagnetic wave signal and receiving the echo signal from the target. The hardware composition of the system is shown in Figure 2.

Horn antenna
Generally, antennas suitable for GPR systems can be roughly divided into two categories: dielectriccoupled antennas and air-coupled antennas. The former includes various dipole elements, bow ties and helical antennas. The latter mainly includes horn antennas of various bandwidths. Due to the large volume and weight of air-coupled antennas relative to dielectric-coupled antennas, they are usually used in high-frequency bands above 1 GHz [8]. The system built in this paper is mainly used for the detection of shallow targets, so the horn antenna ranges from 1 GHz to 4 GHz is selected as the transmitting and receiving antenna of the system, which is shown in Figure 3.

Composition of full-polarimetric GPR data
There are four polarization modes for full-polarimetric GPR, as shown in Figure 5: horizontal transmission-horizontal reception (HH), horizontal transmission-vertical reception (HV), vertical transmission-horizontal reception (VH) and vertical transmission-vertical reception (VV). The measurement matrix at each measuring point can be obtained from these four polarization modes (Yu, 2016), which is shown in formula (1): where, , , is the data measured in HH polarization mode, , , is the data measured in HV polarization mode, , , is the data measured in VH polarization mode, , , is the data measured in VV polarization mode. Because the system's transmitting and receiving antennas are interchangeable, HV data is equal to VH data [11]. Therefore, we only need to measure the data in the three polarization modes: HH, HV, and VV.

Data fusion processing
Data fusion technology is an effective method for full-polarimetric GPR data processing. It fuses the acquired polarized data to make the fused data have the characteristics of various polarized data, which can improve the recognition ability of target. The data fusion processing technique used in this paper is weighted average method, as shown in formula (2) [11].
, , = , , × , , + , , × , , + , , , where, , , is the data fusion result, , , is the weight parameter of , , , , , is the weight parameter of , , . To simplify the problem, we use mean fusion, and set , , = , , = 1. Then we combine the three-polarization data into one fusion data, which can be used for subsequent experimental analysis and processing.

Ice crack experiments and data analysis
In order to demonstrate the overall performance of the constructed full-polarimetric GPR system, an ice crack detection experiment was performed on a frozen lake as shown in Figure 6. The length of the survey line is 1m, and the interval between two measuring points is 1cm. The survey line runs perpendicular to the ice crack, and the ice crack is located at the line of 0.5 m. The width of the upper end of the ice crack is about 1.5cm, and the width of the ice crack is basically unchanged when it extends downward. The frequency band we choose ranges from 1 GHz to 4 GHz with the center frequency of 2.5 GHz, and the time window ranges from 0 ns to 15 ns, with 1024 frequency sweep points in total. The GPS antenna in Figure 6 can record latitude and longitude information, thereby giving an approximate trajectory of the survey line.
The acquired experimental data is subjected to inter-channel equalization processing to obtain radar profile of three polarization modes, as shown in Figure 7   coincidence axis can be seen in the black ellipse which is due to the ice crack, and the position of the coincidence axis coincides with the actual ice crack location. Compared with Figure 7 (a) and 7 (b), the series of hyperbolas can be seen more clearly and completely in the black box in Figure 7 (d), and the hyperbolas have no significant intensity decay. Also, clear hyperbolas can still be seen at the three arrows in Figure 7 (d), which reflects deeper information of ice crack, but only fuzzy hyperbolas can be seen at the corresponding positions in Figure 7 (a) and 7 (b). In addition, a hyperbola with strong intensity can be seen in the range of 0.35 m to 0.75m on the abscissa and 13.8ns to 14ns on the ordinate in figure 7 (d), which we speculate may be caused by the bottom tip of the ice crack. In summary, the data fusion processing results of the three polarization modes can reflect the information of deep ice crack, and can also improve the recognition ability of ice crack.

Conclusions
In this paper, a full-polarimetric GPR system is built. The entire system is integrated and controlled by the PC control unit, which can realize the automatic measurement of the target's fully polarized information. Experiments of a small-scale ice crack on frozen lakes were performed using the constructed system and good results were obtained. In addition, we use the data fusion technology of weighted average method to process the obtained fully polarized data, and the processing results can better reflect the deep information of the target, improving the recognition ability of the ice crack.