Visual localization, the task of determining the position and orientation of a camera, typically involves three core components: offline construction of a keyframe database, efficient online keyframes retrieval, and robust local feature matching. However, significant challenges arise when there are large viewpoint disparities between the query view and the database, such as attempting localization in a corridor previously build from an opposing direction. Intuitively, this issue can be addressed by synthesizing a set of virtual keyframes that cover all viewpoints. However, existing methods for synthesizing novel views to assist localization often fail to ensure geometric accuracy under large viewpoint changes. In this paper, we introduce a confidence-aware geometric prior into 2D Gaussian splatting to ensure the geometric accuracy of the scene. Then we can render novel views through the mesh with clear structures and accurate geometry, even under significant viewpoint changes, enabling the synthesis of a comprehensive set of virtual keyframes. Incorporating this geometry-preserving virtual keyframe database into the localization pipeline significantly enhances the robustness of visual localization.

In synthetic aperture radar (SAR), sidelobes inevitably exist in the matched filtering-based imaging results, which severely affect the image interpretation. Unfortunately, under the near-field condition, the wavenumber spectrum is spatially variant, which means that there is no window function that can effectively weight the wavenumber spectra of all targets at the same time, making sidelobe suppression particularly challenging. To address this issue, an adaptive spatially variant apodization method is proposed in this article. In the proposed method, wavenumber domain weighting is equivalently implemented by a nine-point convolver in the image domain. By adjusting this image domain convolver with pixel position, the method achieves adaptive change of the weighted region, thereby effectively weighting the wavenumber spectra of different targets. In order to avoid the mainlobe broadening caused by weighting, multi-apodization technique is applied to the results after convolution. By transforming multi-apodization into solving a constrained minimization problem, the method achieves the goal of effectively suppressing the sidelobes without reducing the imaging resolution. Experiments on both numerical simulation and measured data verify that the proposed method can suppress the sidelobes by about 10 dB without sacrificing the image resolution.

Active recognition, which allows intelligent agents to explore observations for better recognition performance, serves as a prerequisite for various embodied AI tasks, such as grasping, navigation and room arrangements. Given the evolving environment and the multitude of object classes, it is impractical to include all possible classes during the training stage. In this paper, we aim at advancing active open-vocabulary recognition, empowering embodied agents to actively perceive and classify arbitrary objects. However, directly adopting recent open-vocabulary classification models, like Contrastive Language Image Pretraining (CLIP), poses its unique challenges. Specifically, we observe that CLIP's performance is heavily affected by the viewpoint and occlusions, compromising its reliability in unconstrained embod-ied perception scenarios. Further, the sequential nature of observations in agent-environment interactions necessitates an effective method for integrating features that maintains discriminative strength for open-vocabulary classification. To address these issues, we introduce a novel agent for active open-vocabulary recognition. The proposed method leverages inter-frame and inter-concept similarities to navigate agent movements and to fuse features, without relying on class-specific knowledge. Compared to baseline CLIP model with 29.6% accuracy on ShapeNet dataset, the proposed agent could achieve 53.3% accuracy for open-vocabulary recognition, without any fine-tuning to the equipped CLIP model. Additional experiments conducted with the Habitat simulator further affirm the efficacy of our method.

In the realm of large-scale industrial manufacturing, the precise detection of defective parts stands as a critical imperative. While current unsupervised anomaly detection algorithms exhibit commendable accuracy when applied to clean training datasets, their susceptibility to contaminated training data limits their real-world efficacy. In response to this challenge, this paper proposes a novel Outlier-Probability-Based Feature Adaptation (OPFA) network to realize robust unsupervised anomaly detection on contaminated training data. This method distinguishes itself by maintaining both high accuracy and robustness in the face of contaminated training data, enabling effective learning of discriminative features for anomaly detection. Specifically, the model enhances feature representations through the contraction of normal features and the contrast between normal and outlier features. Our methodology employs an iterative mechanism, featuring three core designs. First, outlier detection evaluates the outlier probabilities of current feature embeddings, providing a basis for subsequent improvements. Second, Gaussian Mixture Model (GMM) is leveraged to model the distributions of normal feature embeddings. Third, the adaptive network refines feature representations based on the GMM models and outlier scores of feature embeddings. Ablation experiments underscore the effectiveness of each component within our model. Furthermore, our approach outperforms other state-of-the-art methods on three benchmark datasets, demonstrating a notable advantage especially in scenarios with contaminated training data.

Weakly-supervised Temporal Action Localization (WTAL) aims to classify and localize action instances in untrimmed videos with only video-level labels. Existing methods typically use snippet-level RGB and optical flow features extracted from pre-trained extractors directly. Because of two limitations: the short temporal span of snippets and the inappropriate initial features, these WTAL methods suffer from the lack of effective use of temporal information and have limited performance. In this paper, we propose the Temporal Feature Enhancement Dilated Convolution Network (TFE-DCN) to address these two limitations. The proposed TFE-DCN has an enlarged receptive field that covers a long temporal span to observe the full dynamics of action instances, which makes it powerful to capture temporal dependencies between snippets. Furthermore, we propose the Modality Enhancement Module that can enhance RGB features with the help of enhanced optical flow features, making the overall features appropriate for the WTAL task. Experiments conducted on THUMOS’14 and ActivityNet v1.3 datasets show that our proposed approach far outperforms state-of-the-art WTAL methods.

Based on Alberta Infant Motor Scale (AIMS), a questionnaire that tracks an infant’s motor function, an infant’s mental development can be evaluated by recording poses a baby can achieve. Therefore, it is meaningful to propose a systematic image-based pose classifier to classify infant actions based on AIMS to provide early diagnosis of a potential develop-mental disorder such as Autism. This paper presents a hierarchical pose classifier, given a baby image frame that com-bines the benefits of 3D human pose estimation and scene context information. Due to privacy policies, we cannot collect enough real infant images/videos for experiments. In-stead, we generate synthetic baby images with the help of the Skinned Multi-Infant Linear (SMIL) model. Images are first fed into a ResNet-50 for coarse-level pose classification. A stacked hourglass CNN and a hierarchical 3D pose estimation scheme are used for 2D/3D pose estimation. Finally, an innovative Hierarchical Infant Pose Classifier (HIPC) takes the estimated 3D keypoints and coarse-level pose classification confidence scores to give the fine-level baby pose classification results. Our experimental results show that our hierarchical pose classifier achieves accurate and stable performance on infant pose recognition.

The inverse synthetic aperture radar (ISAR) imaging technique of a moving target with sparse sampling data has attracted wide attention due to its ability to reduce the data collection burden. However, traditional low-rank or 2D compressive sensing (CS)-based ISAR imaging methods can handle the random sampling or the separable sampling data only. When the specific data collection condition cannot be satisfied, low-rank or 2D CS-based methods cannot provide satisfactory imaging results any more. To remedy this problem, in this paper, we proposed a joint low-rank and sparsity priors' constrained model for ISAR imaging with various sparse data patterns. This model is inspired by the facts that the received radar data have a low-rank property and the ISAR image is sparse on the specific dictionary. Two reconstruction algorithms to solve the double priors' constrained optimization problem are developed under the alternative direction method of multipliers (ADMM) framework with the help of augmented Lagrange multipliers (ALM). Results on simulation data and real data show that the proposed methods are quite effective in recovering missing samples and focused image and perform better than the matrix completion-based method and the sparse representation-based method when dealing with the various kinds of sparse sampling data.

In this letter, a 3-D range migration algorithm for multiple-input multiple-output synthetic aperture radar imaging is proposed. The accurate expression of the signal spectrum is derived by utilizing the spherical wave decomposition. This method compensates for the curvature of the wavefront in wavenumber domain and achieves the 3-D Fourier transform of the reflectivity map through a dimension-reducing accumulation operation. Its fast Fourier transform-based imaging scheme provides high imaging efficiency. In addition, this method does not take the plane wave approximation and can be applied in near-field imaging scenes. Both theoretical analysis and experimental results show that the proposed method can reconstruct the 3-D image as accurately as the backprojection algorithm but with much lower computational load.

In this letter, a novel frequency-domain imaging method is proposed for a single-input-multiple-output array, which avoids the frequency-domain interpolation. This method transforms the measurements into the wavenumber domain for compensation. The spectrum data at each frequency are proved to be the Fourier transform of the phase-modulated reflectivity function; therefore, a subimage at a specific frequency can be produced by inverse fast Fourier transform and phase demodulation. The final image is obtained by coherent accumulation of all subimages. This method does not take the plane wave approximation and can be applied in short-range imaging scenes. It has the same imaging accuracy as the backprojection algorithm but greatly reduces the computational load. Both two- and three-dimensional imaging experiments verify its performance.

In this letter, we propose a fast reconstruction algorithm for 3-D turntable microwave imaging from sparse measurements. A conventional Fourier-transform-based 3-D microwave imaging method collects data over densely azimuth-elevation samples and needs a large amount of data storage and long collection time. To reduce the cost of data acquisition, the proposed method exploits the sparsity in the image domain to achieve 3-D microwave imaging by utilizing sparse measurements. For this aim, the signal model is first represented as a tensor array, and then, a novel sparse reconstruction algorithm called 3-D-SL0 is applied to recover the 3-D scattering reflectivity, i.e., a 3-D image. Simulation results are finally shown to investigate the validity of the proposed method.

A methodology to reconstruct the 3-D scattering center model from the data with wide azimuthal aperture at a single elevation, such as those collected in turntable or circular synthetic aperture radar configurations, is proposed in this paper. The wide azimuthal aperture is divided into overlapped subapertures, and the 2-D scattering centers are extracted in each subaperture. These local scattering centers are rotated and mapped into the ground plane in the target coordinate system, where they are associated according to their location and amplitude consistency. Three-dimensional position of the scatterer is then estimated from the location variation of the 2-D points at different azimuths. The theoretical performance of the position estimator is analyzed, which reveals how the scatterers' azimuthal directivity and persistency affect the precision of the position estimates. The reconstructed model consisted of scattering centers described by their 3-D positions and scattering coefficient profiles at the specific depression angle. Examples using both point scatterers and computer-aided design models not only verify the validity of the methodology but also manifest the applicability of the reconstructed model in scattering analysis, data regeneration, and elevation extrapolation.

A 2-D range/cross-range radar image of a target is always sparse since only a few strong scattering centers occupy the whole image plane, and thus, it is quite suitable to apply the compressive sensing (CS) theory to obtain inverse synthetic aperture radar (ISAR) images. In this paper, a novel fully polarimetric ISAR imaging method based on CS is proposed. First, a definition of joint sparsity is given by exploiting the scattering characteristics of a target in fully polarimetric channels. Then, fully polarimetric ISAR images are constructed by means of the sparse recovery algorithm under the constraint of the joint sparsity. This proposed imaging method combines the merits of a full-polarization technique and CS theory, and hence, it has two main advantages: it can provide high-resolution ISAR images with limited measurements, which is a promising technique for reducing data storage; it generates fully polarimetric ISAR images with the number and the positions of the scattering centers aligned in polarimetric channels, which allows for further polarimetric scattering characteristic analysis. Finally, both simulation and experimental results are shown to demonstrate the validity of the proposed approach.

This paper proposes a synthetic aperture radar (SAR) automatic target recognition approach based on a global scattering center model. The scattering center model is established offline using range profiles at multiple viewing angles, so the original data amount is much less than that required for establishing SAR image templates. Scattering center features at different target poses can be conveniently predicted by this model. Moreover, the model can be modified to predict features for various target configurations. For the SAR image to be classified, regional features in different levels are extracted by thresholding and morphological operations. The regional features will be matched to the predicted scattering center features of different targets to arrive at a decision. This region-to-point matching is much easier to implement and is less sensitive to nonideal factors such as noise and pose estimation error than point-to-point matching. A matching scheme going through from coarse to fine regional features in the inner cycle and going through different pose hypotheses in the outer cycle is designed to improve the efficiency and robustness of the classifier. Experiments using both data predicted by a high-frequency electromagnetic (EM) code and data measured in the MSTAR program verify the validity of the method.

The Bayes principle is applied to the joint model selection and parameter estimation of GTD model to explore the prior information. An algorithm using RJ-MCMC is designed. It not only has better model selection and parameter estimation performance than the non-Bayes algorithms, but also solves the mixed parameter estimation problem in GTD model effectively. The advantage of this algorithm is especially evident at low SNR, for short data and with closely-spaced components. Simulations verify the effectiveness of this algorithm.

Performance bounds on the estimates of position, intensity and geometry parameter of scattering centers based on wideband radar measurements are presented in analytic forms. The resolution limit for wideband radar and the SNR threshold for identifying scatterer's geometry are further deduced. Though the results are obtained from the Cramér-Rao Bound (CRB) matrix for damped exponentials (DE) after simplification, their validity and adaptability for geometric theory of diffraction (GTD) based scattering data have been verified by simulations.