Object detection is a challenging task in aerial images, where many objects have large aspect ratios and are densely arranged. Most anchor-based rotating detectors assign anchors for ground-truth objects by a fixed restriction of the rotation Intersection-over-Unit $(IoU)$ between anchors and objects, which directly follow horizontal detectors. Due to many directional objects with a large aspect ratio, the object-anchor $IoU$ is heavily influenced by the angle, which may cause few anchors assigned for some ground-truth objects. In this study, we propose an anchor selection method based on sample balance assigning anchors adaptively, which we name the Self-Adaptive Anchor Selection ($A^2$S-Det) method. For each ground-truth object, $A^2$S-Det selects a set of candidate anchors by horizontal $IoU$. Then, an adaptive threshold module is adopted on the set of candidate anchors, which calculates a boundary of these candidate anchors aiming to keep a balance between positive and negative anchors. In addition, we propose a coordinate regression of relative reference ($C{R}^3$) module to precisely regress the rotating bounding box. We test our method on a public aerial image dataset, and prove better performance than many other one-stage detectors and two-stage detectors, achieving the $mAP$ of 70.64. An efficiency anchor matching method helps the detector achieve better performance for objects with large aspect ratios.

Remote-sensing scene classification (RSSC) is crucial for remote-sensing image interpretation and has become a research hotspot in recent years. However, the high complexity of remote-sensing scenes causes most RSSC models to fail to accurately capture key objects, resulting in low classification accuracy. Meanwhile, it is intractable to effectively distinguish similar scenes, such as forest and meadow, whose semantic labels are mainly determined by wide-scale features. In addition, existing remote-sensing attention mechanisms are heuristic settings, which require expert knowledge and extensive experiments. To solve the above problems, a novel plug-and-play hyperparameter-free attention module (HFAM) based on feature map mathematical calculation is proposed in this work. HFAM uses statistical indicators to quantitatively characterize the fluctuations of feature maps that can accurately locate key features and distinguish different scenes, alleviating the problems of intraclass diversity and interclass similarity. Moreover, HFAM adaptively acquires attention weights by performing simple mathematical calculations on the feature maps, which solves the problem of difficult adjustment of hyperparameters. Our proposed HFAM can be expediently inserted into the existing ConvNet models without increasing the number of model’s parameters. Extensive contrast experiments with several famous plug-and-play attention modules on three mainstream datasets reveal the superiority of our HFAM in accuracy, number of parameters, and calculation amount. Moreover, compared with state-of-the-art methods, it also demonstrated considerable competitiveness.

Oriented object detectors have achieved great success in aerial detection task with the help of ample labeled data. Unlabeled images are easier and less expensive to obtain than labeled aerial images. Therefore, semi-supervised oriented object detection (SSOOD) is becoming a hot task, which can leverage both labeled and unlabeled data to train oriented detectors. Most SSOOD approaches focus on well-designed approaches to generate high-quality pseudo labels or positive learning regions, which are limited for complex and variable aerial scene. This study first analyzes key factors influencing the performance of SSOOD and proposes a global focal learning method (termed as Focal Teacher) without artificial priori design. It rely on global region and soft regression approaches to blur the boundaries between positive and negative samples, mainly through Localization Focal Loss to achieve. It leverages the localization consistency between teacher and student model to focus more on hard regions. Moreover, we organize a large remote sensing unlabeled dataset to exploit the performance potential of oriented detectors on mainstream aerial detection datasets (DOTA and DIOR). Adequate experiments reveal that the proposed method achieves the best performance compared with other mainstream SSOOD methods, including partly, fully and additionally data settings on DOTA and DIOR datasets. Semi-supervised mechanism without preset learning regions can be better applied in dense and complex aerial scenes.

Oriented object detection plays an important role in aerial image interpretation. Image processing speed is also essential due to massive amounts of aerial images. Anchor-free oriented detectors with fast processing speed are generally accepted despite the absence of pre-set anchors, contributing to their performance gap with anchor-based detectors. Most anchor-free oriented detectors are carefully designed by defining samples according to target characteristics, which require substantial prior knowledge, to realize improved performance. This study proposes an anchor-free oriented detector (termed as rfpoint) that requires minimal prior knowledge. Moreover, this study mainly aims to introduce a dynamic sample definition strategy. This strategy is modeled as a dynamic regulating process, wherein the classification and box regression interact until the model converges. A rotating quality-driven loss (RQDL) and adaptive-weight box loss (AWBL) are also proposed to realize the aforementioned process. RQDL redefines positive and negative attributes of samples according to the distribution of rotating Intersection-over-Unit (IoU) between predictions and ground truth. AWBL adjusts the importance degree of candidate samples in the box regression through classification scores. The proposed method is then tested on three mainstream aerial image datasets (DOTA, DIOR, and HRSC2016). Results reveal that the proposed method achieves the best performance compared with other oriented detectors, whose mAP are 79.92%, 70.88%, and 90.67%. Moreover, the method maintains the inference speed advantage of anchor-free detectors. An effective sample definition method can bridge the performance gap of anchor-free oriented detectors without minimizing inference speed.

Object recognition, as one of the most fundamental and challenging problems in high-resolution remote sensing image interpretation, has received increasing attention in recent years. However, most conventional object recognition pipelines aim to recognize instances with bounding boxes in a supervised learning strategy, which require intensive and manual labor for instance annotation creation. In this paper, we propose a weakly supervised learning method to alleviate this problem. The core idea of our method is to recognize multiple objects in an image using only image-level semantic labels and indicate the recognized objects with location points instead of box extent. Specifically, a deep convolutional neural network is first trained to perform semantic scene classification, of which the result is employed for the categorical determination of objects in an image. Then, by back-propagating the categorical feature from the fully connected layer to the deep convolutional layer, the categorical and spatial information of an image are combined to obtain an object discriminative localization map, which can effectively indicate the salient regions of objects. Next, a dynamic updating method of local response extremum is proposed to further determine the locations of objects in an image. Finally, extensive experiments are conducted to localize aircraft and oiltanks in remote sensing images based on different convolutional neural networks. Experimental results show that the proposed method outperforms the-state-of-the-art methods, achieving the precision, recall, and $F_1$-score at 94.50%, 88.79%, and 91.56% for aircraft localization and 89.12%, 83.04%, and 85.97% for oiltank localization, respectively. We hope that our work could serve as a basic reference for remote sensing object localization via a weakly supervised strategy and provide new opportunities for further research.