[S-TT39] Synthetic Aperture Radar and its application

[STT39-05] Making SLC images in steep terrain by FMCW SAR-equipped drone observation and evaluating their accuracy

*Yutaro Shigemitsu¹, Kazuya Ishitsuka¹, Weiren Lin¹, Tomoyuki Sugiyama², Munemaru Kishimoto², Takeharu Takahashi² (1.Kyoto University, 2.Nittetsu Mining Consultants Co., Ltd)

Keywords:drone, FMCW SAR, SLC images, coherence

In recent years, several studies have been conducted on drones equipped with synthetic aperture radar owing to the improvement in the accuracy of drone position estimation. For example, in Bekar et al. (2022), a 77-GHz frequency-modulated continuous-wave (FMCW) radar system was used to produce highly accurate Single Look Complex (SLC) images. Drones are effective in steep terrain because they have more orbital flexibility than any other platforms for SAR data acquisitions. Therefore, SAR-equipped drones are expected to play an important role in areas that are difficult to observe with satellite observations. However, while many previous studies have conducted experimental observations in areas where it is easy to obtain microwave reflections, there is insufficient discussion on the application of SAR-equipped drones in steep terrain. In this study, observation experiments using a SAR-equipped drone were conducted at a mine, and SLC images were produced and evaluated to verify the validity of the SAR-equipped drone observation system in steep terrain. The observation area of the mine included mountainous and bare areas, and radar reflections from the bare areas were expected. The radar used was a 12.875-GHz FMCW radar system, which has the advantage of requiring only a Fourier transform for SAR processing in the range direction, and for SAR processing in the azimuth direction, which is the key to SAR analysis, a convolution integral of reference waves was applied on the time axis. The drone used was a DJI Matrice 300. Real Time Kinematic (RTK) method was used to estimate the drone's position, and it succeeded in reducing the orbit error to less than a few 10 cm compared to the preset orbit. In a total of four observation experiments, the speed of the drone was varied in the order of 1, 3, 5, and 3 m/s. The effect of different drone speeds on the SLC images was examined. Coherence calculated from the two images was used to quantitatively evaluate the coherence of the SLC image.
As results, high radar reflection intensity results were obtained from the bare areas of the mine, and SLC images with a shape similar to that of the bare terrain were obtained. In addition, the data from all four experiments showed a high degree of agreement with the bare terrain, indicating the high reproducibility of the observation experiments. On the other hand, the radar reflection intensity was low in the mountainous area, which was considered to be a result of random scattering. Next, the relationship between drone speed and resolution was examined. It was confirmed that the resolution of the observed object increased as the drone speed increased. In particular, when the drone speed was 5 m/s, the resolution in the azimuth direction was approximately 10 m or less. As the resolution is expected to improve as the accuracy of the drone's position estimation, we plan to improve the accuracy of orbit error removal in SAR analysis in the future. Finally, to quantitatively evaluate the phase coherence, the coherence of the two images at a drone speed of 3 m/s was calculated. Under the conditions of azimuthal resolution of 0.9 m and a search window size of 5 × 5, the coherence distribution took the form of a normal distribution with a mean value of approximately 0.5 and a standard deviation of 0.2, indicating a certain degree of phase coherence. It is desirable for the coherence to be close to 1, but it is expected that better coherence will be obtained in the future by removing the drone's orbital error. These results suggest that SAR analysis using a drone as a platform is applicable to steep terrain.