Lihong Kang

Accurate detection of small objects plays an important role in the application of unmanned aerial vehicles (UAV). However, current works mainly extract comprehensive features from unimodal images, which can obtain very limited distinguishable features for objects, especially those with small sizes. To address this issue, we propose a dynamic cascade cross-modal co-assisted network(DCCCNet), which integrates multimodal images fusion and fine-grained feature learning to generate powerful object semantic representations. Specifically, we design a multimodal high-order interaction module (MHl) to achieve collaborative interaction of spatial details and channel dependencies between modalities, thereby enhancing object discrimination. To preserve multimodal fine-grained details, we devise a scale-adaptive dynamic feature prompt (SADFP) module, which dynamically motivate the backbone network to capture feature degradation clues. Meanwhile, to maintain the spatial correlation of multi-modal cross-scale features and improve the quality of feature fusion, we derive an global collaborative enhancement (GCE) module into feature pyramid network(FPN) for enhancing the detection accuracy across multiple scales. Extensive experimental results on multimodal datasets have shown that our method achieves favorable performance, surpassing other state-of-the-art methods.

Multiclass geospatial object detection is a vital fundamental task for many remote sensing applications. However, it still faces several challenges in very high-resolution (VHR) images in remote sensing, such as the ambiguity of object appearance and the complexity of spatial distribution. In this letter, we propose a novel Siamese graph embedding network (SGEN) that leverages the spatial and semantic information to jointly extract the high-level feature representation for object detection. The main purpose of our SGEN is to learn an embedding discriminative feature space that strengthens the interclass compactness while alleviating the intraclass separability. Specifically, we first design a novel contrastive loss in terms of spatial dependence and semantic correspondence for graph similarity metric learning (ML). Then, the SGEN architecture is adopted for spatial and semantic similarity learning by training the novel contrastive loss function. The SGEN model contains two-stream graph convolutional networks (GCNs) for ML, which is helpful to capture the discriminative features. At last, these extracted features with high spatial and semantic discrimination are served to improve the performance of object detection. The comprehensive evaluations on a combined data set consisting of two public object detection data sets demonstrate the effectiveness of the proposed method.

The paper proposes a new method for measuring the azimuth pointing of spaceborne synthetic aperture radar (SAR) antenna beams based on the ground receiver, which can receive and record complex sampling data of the pulse signals transmitted from the spaceborne SAR. The center of the antenna pattern is extracted from the complex sampling data amplitude envelope to obtain the time when the beam main lobe center irradiates the ground receiver, and the range migration information is extracted from the complex sampling data to obtain the time when the satellite is over the top of the ground receiver. The results of Chinese civilian remote sensing GaoFen-3 SAR satellite experiment data processing show that the measurement accuracy of this method is better than 0.002°, which can be applied to the accurate measurement of azimuth pointing of various low Earth orbit (LEO) SAR antenna beams.

Compared with low-Earth orbit synthetic aperture radar (SAR), a geosynchronous (GEO) SAR can have a shorter revisit period and vaster coverage. However, relative motion between this SAR and targets is more complicated, which makes range cell migration (RCM) spatially variant along both range and azimuth. As a result, efficient and precise imaging becomes difficult. This paper analyzes and models spatial variance for GEO SAR in the time and frequency domains. A novel algorithm for GEO SAR imaging with a resolution of 2 m in both the ground cross-range and range directions is proposed, which is composed of five steps. The first is to eliminate linear azimuth variance through the first azimuth time scaling. The second is to achieve RCM correction and range compression. The third is to correct residual azimuth variance by the second azimuth time-frequency scaling. The fourth and final steps are to accomplish azimuth focusing and correct geometric distortion. The most important innovation of this algorithm is implementation of the time-frequency scaling to correct high-order azimuth variance. As demonstrated by simulation results, this algorithm can accomplish GEO SAR imaging with good and uniform imaging quality over the entire swath.