Object detection in optical remote-sensing images is an important and challenging task. In recent years, the methods based on convolutional neural networks (CNNs) have made good progress. However, due to the large variation in object scale, aspect ratio, as well as the arbitrary orientation, the detection performance is difficult to be further improved. In this article, we discuss the role of discriminative features in object detection, and then propose a critical feature capturing network (CFC-Net) to improve detection accuracy from three aspects: building powerful feature representation, refining preset anchors, and optimizing label assignment. Specifically, we first decouple the classification and regression features, and then construct robust critical features adapted to the respective tasks of classification and regression through the polarization attention module (PAM). With the extracted discriminative regression features, the rotation anchor refinement module (R-ARM) performs localization refinement on preset horizontal anchors to obtain superior rotation anchors. Next, the dynamic anchor learning (DAL) strategy is given to adaptively select high-quality anchors based on their ability to capture critical features. The proposed framework creates more powerful semantic representations for objects in remote-sensing images and achieves high-performance real-time object detection. Experimental results on three remote-sensing datasets including HRSC2016, DOTA, and UCAS-AOD show that our method achieves superior detection performance compared with many state-of-the-art approaches. Code and models are available at https://github.com/ming71/CFC-Net.
1 October 2021 in IEEE Geoscience and Remote Sensing Letters
Arbitrary-oriented objects exist widely in remote sensing images. The mainstream rotation detectors use oriented bounding boxes (OBBs) or quadrilateral bounding boxes (QBBs) to represent the rotating objects. However, these methods suffer from the representation ambiguity for oriented object definition, which leads to suboptimal regression optimization and the inconsistency between the loss metric and the localization accuracy of the predictions. In this letter, we propose a representation invariance loss (RIL) to optimize the bounding box regression for the rotating objects in the remote sensing images. RIL treats multiple representations of an oriented object as multiple equivalent local minima and hence transforms bounding box regression into an adaptive matching process with these local minima. Next, the Hungarian matching algorithm is adopted to obtain the optimal regression strategy. Besides, we propose a normalized rotation loss to alleviate the weak correlation between different variables and their unbalanced loss contribution in OBB representation. Extensive experiments on remote sensing datasets show that our method achieves consistent and substantial improvement. The code and models are available at https://github.com/ming71/RIDet to facilitate future research.
Object detection in aerial images has received extensive attention in recent years. The current mainstream anchor-based methods directly divide the training samples into positives and negatives according to the intersection-over-unit (IoU) of the preset anchors. This label assignment strategy assigns densely arranged samples for training, which leads to a suboptimal learning process and cause the model to suffer serious duplicate detections and missed detections. In this paper, we propose a sparse label assignment strategy (SLA) to select high-quality sparse anchors based on the posterior IoU of detections. In this way, the inconsistency between classification and regression is alleviated, and better performance can be achieved through balanced training. Next, to accurately detect small and densely arranged objects, we use a position-sensitive feature pyramid network (PS-FPN) with a coordinate attention module to extract position-sensitive features for accurate localization. Finally, the distance rotated IoU loss is proposed to eliminate the inconsistency between the training loss and the evaluation metric for better bounding box regression. Extensive experiments on the DOTA, HRSC2016, and UCAS-AOD datasets demonstrate the superiority of the proposed approach.
29 May 2020 in IEEE Transactions on Geoscience and Remote Sensing
Currently, reliable and accurate ship detection in optical remote sensing images is still challenging. Even the state-of-the-art convolutional neural network (CNN)-based methods cannot obtain very satisfactory results. To more accurately locate the ships in diverse orientations, some recent methods conduct the detection via the rotated bounding box. However, it further increases the difficulty of detection because an additional variable of ship orientation must be accurately predicted in the algorithm. In this article, a novel CNN-based ship-detection method is proposed by overcoming some common deficiencies of current CNN-based methods in ship detection. Specifically, to generate rotated region proposals, current methods have to predefine multioriented anchors and predict all unknown variables together in one regression process, limiting the quality of overall prediction. By contrast, we are able to predict the orientation and other variables independently, and yet more effectively, with a novel dual-branch regression network, based on the observation that the ship targets are nearly rotation-invariant in remote sensing images. Next, a shape-adaptive pooling method is proposed to overcome the limitation of a typical regular region of interest (ROI) pooling in extracting the features of the ships with various aspect ratios. Furthermore, we propose to incorporate multilevel features via the spatially variant adaptive pooling. This novel approach, called multilevel adaptive pooling, leads to a compact feature representation more qualified for the simultaneous ship classification and localization. Finally, a detailed ablation study performed on the proposed approaches is provided, along with some useful insights. Experimental results demonstrate the great superiority of the proposed method in ship detection.
Oriented object detection has a very wide range of application scenarios. In recent years, a lot of rotation detectors have been designed to achieve high-performance oriented object detection. Intersection-over-Union (IoU) is the commonly used indicator to evaluate the accuracy of detection performance. Many methods introduce IoU into the bounding box regression loss to achieve the aligned training and evaluation process for better performance. However, in this paper, we demonstrate several drawbacks of rotated IoU loss through both experiments and theoretical derivation: 1) There is a negative correlation between the loss gradient and the angular error. 2) The optimization process of rotated IoU loss suffers from scale sensitivity, which is not conducive to the model convergence. To solve the problems, we propose a Gradient Calibration Loss (GCL) that optimizes the rotated IoU loss via gradient analysis and correction. We construct the optimized gradient in GCL to avoid IoU loss oscillation and scale sensitivity, thereby accelerating model convergence. Models supervised by GCL have a more stable training process, faster convergence, and better performance. Moreover, GCL can be easily introduced into the existing rotation detectors to achieve performance gains without extra inference overhead. Extensive experiments on multiple oriented object detection datasets and models demonstrate the superiority of our method. Our method achieves state-of-the-art performance on the mainstream benchmark datasets. The source code and models are available at https://github.com/ming71/GCL.