Deep learning-based aircraft detection methods have been increasingly implemented in recent years. However, due to the multi-resolution imaging modes, aircrafts in different images show very wide diversity on size, view and other visual features, which brings great challenges to detection. Although standard deep convolution neural networks (DCNN) can extract rich semantic features, they destroy the bottom-level location information. The features of small targets may also be submerged by redundant top-level features, resulting in poor detection. To address these problems, we proposed a compact multi-scale dense convolutional neural network (MS-DenseNet) for aircraft detection in remote sensing images. Herein, DenseNet was utilized for feature extraction, which enhances the propagation and reuse of the bottom-level high-resolution features. Subsequently, we combined feature pyramid network (FPN) with DenseNet to form a MS-DenseNet for learning multi-scale features, especially features of small objects. Finally, by compressing some of the unnecessary convolution layers of each dense block, we designed three new compact architectures: MS-DenseNet-41, MS-DenseNet-65, and MS-DenseNet-77. Comparative experiments showed that the compact MS-DenseNet-65 obtained a noticeable improvement in detecting small aircrafts and achieved state-of-the-art performance with a recall of 94% and an F1-score of 92.7% and cost less computational time. Furthermore, the experimental results on robustness of UCAS-AOD and RSOD datasets also indicate the good transferability of our method.
Quantum dots | Machine learning | Materials genome initiative | Neural networks | On-demand With the rapid developments in the field of information technology, the material research society is looking for an alternate scientific route to the traditional methods of trial and error in material research and process development. Machine learning emerges as a new research paradigm to accelerate the application-oriented material discovery. Quantum dots are expanded as functional nanomaterials to enhance cutting-edge photonic technology. However, they suffer from uncertainty in industrial fabrication and application. Here, we discuss how machine learning accelerates the development of quantum dots. The basic principles and operation procedures of machine learning are described with a few representative examples of quantum dots. We emphasize how machine learning contributes to the optimization of synthesis and the analysis of material characterizations. To conclude, we give a short perspective discussing the problems of combining machine learning and quantum dots.
The positioning accuracy is critical for satellite-based topographic modeling in cases of exterior orientation parameters with high uncertainty and scarce ground control data. The integration of multi-sensor data can help to ensure precision topographical modeling in such situations. Presently, research on the combined processing of optical camera images and laser altimeter data has focused on planetary observations, especially on the Moon and Mars. This study presents an endeavor to establish a combined adjustment model with one constraint in image space for integration of ZY3-02 stereo images and laser altimeter data for improved Earth topographic modeling. The geometric models for stereo images and laser altimeter data were built first, and then, the laser ranging information was introduced to construct a combined adjustment model on the basis of the block adjustment model. One constraint that minimized the back-projection discrepancies in image space was incorporated into the combined adjustment. Datasets in several areas were collected as experimental data for the validation work. Experimental results demonstrated that the inconsistencies between stereo images and laser altimeter data for the ZY3-02 satellite can be reduced, and the elevation accuracy of stereo images can be significantly improved after applying the proposed combined adjustment. Experiments further proved that the improved height accuracy is insensitive to the number and relative position of laser altimeter points (LAPs) in stereo images. Moreover, additional plane control points (PCPs) were incorporated to achieve better planimetric accuracy. Experimental results in the Dengfeng area showed that the adjustment results derived by using LAPs and additional four PCPs were only slightly lower than those for the block adjustment with four ground control points (GCPs). Generally, the proposed approach can effectively improve the quality of Earth topographic model.
The Luojia1-01 satellite provides high-resolution, high-sensitivity nighttime light data at a resolution of 130 m. To effectively use the Luojia1-01 nighttime light data for global applications, the problems of relative and absolute positioning accuracy should be solved. This paper proposes a high accuracy regional geometric processing method of nighttime light imagery. We utilized a nighttime light image matching algorithm to obtain tie points, which are used in the planar block adjustment with ground control points. Then, orthorectification of all images is implemented. Finally, we obtain the nighttime light map of China by mosaicking all the nighttime light orthoimages. According to the experimental results for 275 Luojia1-01 images, the root mean square error of the tie points is 0.983 pixels and the root mean square error of independent checkpoints is 195.491 m (less than 1.5 pixels) after the planar block adjustment. The experimental results prove the validity and feasibility of the proposed method.
Illegal open-pit mining causes environmental harm and undermines sustainable development. Conventional monitoring approaches such as field research and unmanned aerial vehicle (UAV) imagery are time-consuming and labor-intensive, making large-scale monitoring difficult. In comparison, optical remote sensing imagery can cover large areas but is vulnerable to adverse weather conditions and is not sensitive to vertical ground changes. As open-pit excavation causes sudden changes in the scattering properties of ground objects along with dramatic vertical deformation, we evaluated the feasibility of using interferometric synthetic aperture radar (InSAR) coherence to identify illegal mining activities. Our method extracts the coherence coefficient from two SAR images taken on different dates, applies thresholding and filtering to extract a decorrelation map, and then overlays this with legal mining boundaries and optical satellite images to identify illegal mining activities. For three test cases in southwestern Inner Mongolia, China, 49 legal mining sites were correctly detected (with an accuracy of 90.74%) as well as six illegal mining sites. Ground truthing confirmed the presence of ongoing activity at one of these sites. Our study shows that InSAR coherence is suitable for the identification of mining activities, and our method provides a new approach for the detection and monitoring of illegal open-pit mining.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.