Multi-aspect Synthetic Aperture Radar (SAR) observations of complex targets often exhibits distinct anisotropic feature, which also contains rich information about target structural complexity. In practical scenarios, SAR images at certain aspects may be unsampled or not be available, in such case, interpolation of SAR images from other sampled or available aspects, known as multi-aspect interpolation, is needed. This paper conducts a preliminary study on the quantitative description of anisotropic scattering variation (ASV) and the relationship between ASV and multi-aspect interpolation based on typical aircraft target airborne multi-aspect SAR dataset. The aim is to clarify when it is more meaningful to perform multi-aspect interpolation and propose ASV operator to describe the variation of anisotropic scattering, which, to some extent, can also express target structural complexity. In the exploration process, it was found that certain expression of anisotropic feature can reflect the structural and textural feature of targets, which can represent changes of anisotropic scattering. Therefore, these expressions are incorporated into ASV operator, which makes the operator effective. Experimental results show that the proposed ASV operator can accurately describe changes of anisotropic scattering based on aircraft target data. Moreover, it can predict the effectiveness of interpolated SAR images based on the value of ASV operator. Consequently, multi-aspect interpolation of SAR images at required aspects can be performed based on prior predictions using sampled SAR images.

Due to the huge differences in Synthetic Aperture Radar (SAR) image features of the same target under different observing aspects, the demand for constructing multi-aspect SAR datasets of various typical targets is becoming increasingly urgent with the expansion of SAR technology application fields. Meanwhile, multi-aspect interpolation techniques for constructing multi-aspect SAR datasets, based on electromagnetic scattering features and on Generative Adversarial Networks (GANs), have some shortcomings that are difficult to address. The former method provide descriptions of the target scattering so overly idealized that they are not real, while the latter method suffers from incomplete amplitude information and a loss of phase information in multi-aspect interpolation results due to the SAR images input into GANs being phaseless and amplitude-quantized. In response to the above issues, this paper proposes the Multi-aspect Scattering Information Complex GAN (MS-CGAN) guided by the scattering information in observing aspects of SAR images to simulate the multi-aspect interpolation of SAR images from specific aspects. MS-CGAN provides a new approach for dataset construction and augmentation. Moreover, as a complex network, MS-CGAN does not require phase removal or amplitude quantization of the input SAR images; thus, the significant issue of the severe loss of scattering information in multi-aspect interpolation methods based on GANs is greatly addressed. In the experiments, assuming the absence of real SAR images from certain aspects, both the correlation coefficient and the phase correlation between interpolated SAR images from MS-CGAN and real SAR images achieve good results. In the case of a sampling aspect interval of 10°, the mean correlation of the amplitude and phase of the interpolated SAR images and the corresponding real SAR images both reach over 80%. In the case of a sampling aspect interval of 20°, the mean correlation of the amplitude and phase of the interpolated SAR images and the corresponding real SAR images remain above 75%. In the case of a sampling aspect interval of 30°, the mean correlation of the amplitude and phase of the interpolated SAR images and the corresponding real SAR images can reach around 70%. Energy integration curves are completed at specific aspects, demonstrating the effectiveness of the MS-CGAN multi-aspect interpolation method.