Object detection from aerial and satellite remote sensing images has been an active research topic over the past decade. Thanks to the increase in computational resources and data availability, deep learning-based object detection methods have achieved numerous successes in computer vision, and more recently in remote sensing. However, the ability of current detectors to deal with (very) small objects still remains limited. In particular, the fast detection of small objects from a large observed scene is still an open question. In this work, we address this challenge and introduce an enhanced one-stage deep learning-based detection model, called You Only Look Once (YOLO)-fine, which is based on the structure of YOLOv3. Our detector is designed to be capable of detecting small objects with high accuracy and high speed, allowing further real-time applications within operational contexts. We also investigate its robustness to the appearance of new backgrounds in the validation set, thus tackling the issue of domain adaptation that is critical in remote sensing. Experimental studies that were conducted on both aerial and satellite benchmark datasets show some significant improvement of YOLO-fine as compared to other state-of-the art object detectors.
This paper deals with the reconstruction of two-dimensional objects from laboratory-controlled data in microwave tomography. This inverse problem is commonly ill-posed and nonlinear, therefore we propose to solve it in a Bayesian estimation framework using an iterative scheme to solve the optimization problem. This approach allows us to introduce a priori knowledge about the object function to be reconstructed. The experimental data were obtained in a controlled environment at Institut Fresnel (Marseille, France). The considered targets are either metallic or dielectric homogeneous cylinders. In this paper, the authors have only considered the data corresponding to the transverse magnetic polarization case. For these targets, the presented results show the potentiality of the proposed regularization scheme and the interest of these experimental data for testing inverse algorithms.
In this contribution, an adaptive multiscale approach for the localization and characterization of buried objects in a half-space is proposed. The main goal of the approach is to reduce the number of elements to be estimated and so the degrees of freedom in the unknown profile. This leads to improvement of the robustness of the inversion and to an increase in the quality of reconstruction. The proposed inversion scheme is based on an adaptive, coarse-to-fine iterative strategy using spline pyramids. The global procedure consists of sequences of non-linear inversions separated by refinement steps, which overall produces an accurate, low-order representation of the sought object.
Abstract-In this contribution a model based on asymptotic methods is proposed to compute the scattered field from complex objects on a sea surface. The scattering model combines the geometrical optics, the physical optics and the method of equivalent currents. It includes the shadowing effects and multiple-bounce up to order 3. This model is used, in the following, for Radar Cross Section (RCS) estimation and to generate Synthetic Aperture Radar (SAR) raw data for imaging applications. The theoretical aspects are reviewed in this paper and the proposed model is detailed. Numerical results are provided to validate the approach through the computation of RCS for canonical objects and complex scenes. Both the bistatic and the monostatic configurations are studied in this work. Finally some first results dealing with SAR imaging of objects on a sea surface are provided. These images are constructed from the simulated raw data thanks to a chirp scaling-based algorithm.
Convolutional neural networks (CNNs) have rapidly become the state-of-the-art models for image classification applications. They usually require large groundtruthed datasets for training. Here, we address object identification and recognition in the wild for infrared (IR) imaging in defense applications, where no such large-scale dataset is available. With a focus on robustness issues, especially viewpoint invariance, we introduce a compact and fully convolutional CNN architecture with global average pooling. We show that this model trained from realistic simulation datasets reaches a state-of-the-art performance compared with other CNNs with no data augmentation and fine-tuning steps. We also demonstrate a significant improvement in the robustness to viewpoint changes with respect to an operational support vector machine (SVM)-based scheme.
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.