Single-Frame Infrared Small Target Detection (SF-IRSTD) has grown in popularity due to its broad application. Several models based on Low-Rank and Sparse Decomposition (LRSD) have been proposed recently and have shown excellent performance. Nevertheless, these methods regard the non-low-rank sparse points as the targets, obscuring the distinction between the non-low-rank noise and the target in the infrared image. To address this issue, we consider that the targets usually have a high local salience compared to the noise and propose a novel method using High Local Variance, Low-Rank and Sparse Decomposition (HiLV-LRSD), identifying the sparse points with high local salience and non-low-rank as the targets and the remaining regions as the background. Specifically, we first use the local variance to represent local salience and propose an LV* norm to constrain the background’s low-rank and local variance. Then, we define an adaptively re-weighted L1 ( L _lv,1 ) norm to constrain the sparsity of the target and enhance the influence of local variance. Finally, we propose an optimization framework and solve it by a Partially Iterative Alternating Direction Method of Multipliers (PI-ADMM). We evaluate our proposed method on the publicly available dataset SIRST and compare it to 10 state-of-the-art SF-IRSTD methods. The results show that our proposed method outperforms these methods.

Semantic image segmentation, as one of the most popular tasks in computer vision, has been widely used in autonomous driving, robotics and other fields. Currently, deep convolutional neural networks (DCNNs) are driving major advances in semantic segmentation due to their powerful feature representation. However, DCNNs extract high-level feature representations by strided convolution, which makes it impossible to segment foreground objects precisely, especially when locating object boundaries. This paper presents a novel semantic segmentation algorithm with DeepLab v3+ and super-pixel segmentation algorithm-quick shift. DeepLab v3+ is employed to generate a class-indexed score map for the input image. Quick shift is applied to segment the input image into superpixels. Outputs of them are then fed into a class voting module to refine the semantic segmentation results. Extensive experiments on proposed semantic image segmentation are performed over PASCAL VOC 2012 dataset, and results that the proposed method can provide a more efficient solution.

Most object detection methods based on remote sensing images are generally dependent on a large amount of high-quality labeled training data. However, due to the slow acquisition cycle of remote sensing images and the difficulty in labeling, many types of data samples are scarce. This makes few-shot object detection an urgent and necessary research problem. In this paper, we introduce a remote sensing few-shot object detection method based on text semantic fusion relation graph reasoning (TSF-RGR), which learns various types of relationships from common sense knowledge in an end-to-end manner, thereby empowering the detector to reason over all classes. Specifically, based on the region proposals provided by the basic detection network, we first build a corpus containing a large number of text language descriptions, such as object attributes and relations, which are used to encode the corresponding common sense embeddings for each region. Then, graph structures are constructed between regions to propagate and learn key spatial and semantic relationships. Finally, a joint relation reasoning module is proposed to actively enhance the reliability and robustness of few-shot object feature representation by focusing on the degree of influence of different relations. Our TSF-RGR is lightweight and easy to expand, and it can incorporate any form of common sense information. Sufficient experiments show that the text information is introduced to deliver excellent performance gains for the baseline model. Compared with other few-shot detectors, the proposed method achieves state-of-the-art performance for different shot settings and obtains highly competitive results on two benchmark datasets (NWPU VHR-10 and DIOR).

The widespread applications of remote sensing image scene classification-based Convolutional Neural Networks (CNNs) are severely affected by the lack of large-scale datasets with clean annotations. Data crawled from the Internet or other sources allows for the most rapid expansion of existing datasets at a low-cost. However, directly training on such an expanded dataset can lead to network overfitting to noisy labels. Traditional methods typically divide this noisy dataset into multiple parts. Each part fine-tunes the network separately to improve performance further. These approaches are inefficient and sometimes even hurt performance. To address these problems, this study proposes a novel noisy label distillation method (NLD) based on the end-to-end teacher-student framework. First, unlike general knowledge distillation methods, NLD does not require pre-training on clean or noisy data. Second, NLD effectively distills knowledge from labels across a full range of noise levels for better performance. In addition, NLD can benefit from a fully clean dataset as a model distillation method to improve the student classifier’s performance. NLD is evaluated on three remote sensing image datasets, including UC Merced Land-use, NWPU-RESISC45, AID, in which a variety of noise patterns and noise amounts are injected. Experimental results show that NLD outperforms widely used directly fine-tuning methods and remote sensing pseudo-labeling methods.

Object detection in optical remote sensing images has made remarkable progress on computer vision. Unlike objects in natural images, objects in remote sensing are generally more complex and diverse. Therefore, most methods design deeper networks to capture more distinguishing features for each object region while ignoring the semantic correlation between scenes and ground objects. In this work, we introduce a novel multiknowledge learning module (MKLM), which combines two kinds of knowledge to adaptively enhance the original feature representation of the basic detector by interacting with information. Specifically, the internal knowledge module is utilized to learn global and local contextual relationships (e.g. feature similarity, spatial location) given object proposals, and the external knowledge module learns human common-sense knowledge (e.g. co-occurrence, inherent properties) to impose constraints. Then, it achieves the purpose of updating the object state under the supervision of different knowledge guidance. Our MKLM is easy to extend and migrate and can be embedded in any detection paradigm. Reliable experiments on two challenging remote sensing image object detection datasets (DOTA and RSOD) show that the proposed method achieves superior detection performance compared to many state-of-the-art methods and baselines.