Robotic grasping plays a critical role in robotics, with widespread applications across various domains. The stability of a grasp is crucial for subsequent operations, making accurate and robust stability assessments essential. While existing methods predominantly rely on visual data, the absence of tactile signals and quantitative stability metrics often leads to unreliable grasp execution. This paper makes three fundamental contributions to address these limitations. First, a visual-tactile grasp dataset generation framework is proposed using NVIDIA Isaac Sim, which synthesizes large-scale multimodal grasping scenarios with stability degree labels through a physics-informed three-stage pipeline. Second, the paper introduces the grasp margin matrix, a novel computational model that quantifies directional force margins and rotational moment margins to evaluate grasp robustness. This matrix simplifies traditional grasp wrench space analysis by decoupling complex friction cone calculations into interpretable mechanical metrics, achieving 87.72% stability classification accuracy via our Stability Assessment Network. Third, a vision-based grasp perception network that predicts contact force distributions and object centroids without physical tactile sensors is developed, enabling real-time stability inference through the grasp margin matrix. This perception–evaluation–decision integration links grasp pose generation with stability assessment to inform grasp selection, achieving an 88.15% success rates in real robotic grasping scenarios. Experimental validations demonstrate significant improvements compared to force closure methods, with accuracy improving from 56.52% to 87.72%, offering both theoretical rigor and practical feasibility for industrial robotic systems.

Gradient descent strategy, viewed as an important model optimization method, has been widely used for various tasks (such as model-based image denoising) of computer vision. In the gradient descent denoising model, the learning rate (LR) and residual component are two important parts to be adaptively estimated for its stable point. This letter proposes a deep gradient descent network (DGDNet), including two key points: one is that the LR is designed with eigenvalues of Hessian matrix of remotely sensed images (RSIs) and their local weighted factor (LWF), which can recognize structures from RSIs degraded by additive white Gaussian noise (AWGN). The other is that the residual part is calculated by an U-shaped network (USNet) to speed up the DGDNet convergent to a fixed point. Finally, the two components are plugged into the gradient descent scheme and contribute to an enjoyable result with a few iterations. Quantitatively and qualitatively experimental results demonstrate that the proposed DGDNet can obtain a stable solution efficiently, and produce competitive denoising performance which is even better than that yielded by the state-of-the-art noise reduction methods.

Remotely sensed images degraded by additive white Gaussian noise (AWGN) are not beneficial for the analysis of their contents. Such a phenomenon is usually modeled as an inverse problem which can be solved by model-based optimization methods or discriminative learning approaches. The former pursue their pleasing performance at the cost of a highly computational burden while the latter are impressive for their fast testing speed but are limited by their application range. To join their merits, this letter proposes a nonlocal self-similar (NSS) block-based deep image denoising scheme, namely deep low-rank prior (DLRP), which includes the following key points: First, the low-rank property of the neighboring NSS patches ordered lexicographically is utilized to model a global objective function (GOF). Second, with the aid of an alternative iteration strategy, the GOF can be easily decomposed into two independent subproblems. One is a quadratic optimization problem, and has a closed-form solution. While the other is a low-rank minimization denoising problem and is learned by deep convolutional neural network (DCNN). Then, the deep denoiser, acted as a modular part, is plugged into the model-based optimization method with adaptive noise level estimation to solve the inverse problem. In the experiments, we first discuss parameter setting and the convergence. Then, quantitative/qualitative comparisons of experimental results validate that the DLRP is a flexible and powerful denoising method to achieve competitive performance which even outperforms those produced by state-of-the-arts.

Low/high or uneven luminance results in low contrast of remotely sensed images (RSIs), which makes it challenging to analyze their contents. In order to improve the contrast and preserving fine weak details of RSIs, this letter proposes a novel enhancement framework to correct luminance guided by weighted least squares (WLS), including the following key parts. First, an image is separated into a base layer and a detail layer by employing the WLS. Then, a learning network is proposed to correct luminance for the base layer enhancement. Next, an enhancement operator for improving the detail layer is computed by using the original image and the enhanced base layer. Finally, the output image is obtained with a fusion of the enhanced base and detail components. Both quantitatively and qualitatively experimental results verify that the proposed method performs better than the state of the arts in contrast improvement and detail preservation.

In the above article[1], the model in(1)should be revised as\begin{equation*}\min _{\mathcal{I}^{\mathcal{B}}}\left\{\left(\mathcal{I}-\mathcal{I}^{\mathcal{B}}\right)^{2}+\lambda\left(a_{x}(\mathcal{I})\left(\frac{\partial \mathcal{I}^{\mathcal{B}}}{\partial x}\right)^{2}+a_{y}(\mathcal{I})\left(\frac{\partial \mathcal{I}^{\mathcal{B}}}{\partial y}\right)^{2}\right)\right\} \end{equation*}to be minimized for${\mathcal {I}}^{\mathcal {B}}$.