Segmentation of micro-adhesive spots in high-power laser packaging is challenged by morphological variability, complex backgrounds, and blurred edges, causing traditional models to fail from “feature dilution.” Inspired by physical optics knowledge, we propose a scattering neural representation framework guided by Rayleigh scattering theory. We first pretrain a denoising diffusion model, using light scattering properties, including wavelength, scattering angle, and particle number density as an inductive bias to generate high signal-to-noise ratio target features while suppressing background clutter. Subsequently, three synergistic attention modules, including an adaptive dual-attention module, an edge attention module, and a small object enhancement module, refine target features by dynamically expanding the receptive field, sharpening boundaries, and enhancing microtarget responses. Extensive experiments on proprietary and public datasets demonstrate that our model significantly outperforms state-of-the-art methods in precise segmentation and background interference suppression. This work translates physical insights into architectural advantages, establishing an efficient and interpretable paradigm for addressing the persistent challenge of industrial small object segmentation.

China is the largest producer and consumer of cigarettes in the world. However, counterfeit and substandard cigarettes frequently appear in the market, which not only causes financial losses to the country but also disrupts normal market order and poses serious threats to public health. Traditional methods for authenticity identification of cigarettes are plagued by low efficiency, a high reliance on manual labor, and time-consuming procedures. These limitations render them increasingly ineffective against complex and diverse counterfeiting techniques. In this paper, we propose a method for authenticity identification of cigarettes by integrating YOLOv5 and ResNet34, aiming to improve automation and accuracy. The method operates in two stages. In the first stage, a pre-trained YOLOv5 network rapidly detects and crops cigarette pack images, removing distracting backgrounds to produce cleaner data. In the second stage, a ResNet34 network extracts finegrained features from these processed images to classify their authenticity. We construct a dataset comprising images of genuine and counterfeit cigarettes from 12 mainstream brands to train a ResNet34 model. Experiments demonstrate that our method achieves a classification accuracy of 94.464 % on the test set. This result validates the effectiveness and adaptability of our method in different environments. Furthermore, the method significantly reduces the need for human intervention, providing reliable technical support for both regulatory agencies and consumers.

Due to the influence of speckle noise and geometric distortion between images, synthetic aperture radar (SAR) image registration under different imaging conditions is a challenging task in remote sensing. To address the issues of significant differences in scattering and geometric characteristics of SAR images under different viewing angles, this letter proposes a novel SAR image registration method. The existing methods mainly rely on gradient information in the feature point selection process, which leads to uneven distribution of feature points and poor global matching. We design a Harris-based neighborhood difference map (HNDM) detector. This detector uses the degree of difference between neighbor regions and the central region to obtain feature points that are homogeneous and significant. Then, a Gaussian–Gamma-shaped (GGS) feature map is used to construct the feature point characterization, which is more robust to dark region noise. Experimental results of SAR image registration under different conditions show that our method achieves better performance in matching accuracy and the number of correct correspondences, outperforming three existing advanced algorithms.

Sensor errors, platform ephemeris errors, and auxiliary digital elevation model (DEM) errors can have an impact on the positioning accuracy of synthetic aperture radar (SAR) images. Utilizing corner reflectors for geometric calibration is a common way to improve parameter accuracy and therefore positioning accuracy. In this article, we propose an energy-based geometric self-calibration method without relying on corner reflectors. Based on the radiometric and geometric properties of SAR images, the energy of SAR orthophoto over the rugged mountainous areas is taken as objective function. The conjugate gradient method is used to estimate fast time offset and slow time offset by maximizing the objective function, which achieves the equivalent compensation for positioning errors. SAR images from Radarsat-2, COSMO-SkyMed, TerraSAR-X, GaoFen-3, LuTan-1 and ChaoHu-1 satellites were used for the experiments, and the experimental results demonstrate the effectiveness and applicability of the proposed method. The accuracy evaluation results of corner reflectors and field-measured checkpoints show that the proposed method improves the positioning errors of SAR images from different satellites from tens of meters to less than 10 m. Our proposed method not only provides a new perspective on the geometric calibration of SAR but also reduces the maintenance cost and the workload of external calibration during the daily operation of spaceborne SAR, which is of great significance for low-cost commercial satellites.

Nearshore ship detection remains a challenging task due to the similarity in grayscale between harbor areas and ships. In this paper, we propose a new superpixel-based nearshore ship detection method, which effectively reduces land false alarms by the fuzzy C-means (FCM) algorithm. First, the method applies the simple linear iterative clustering (SLIC) algorithm to generate superpixel regions. Superpixels can preserve the boundary of the target and reduce the effects of speckle noise for target detection in synthetic aperture radar (SAR) images. Subsequently, we used the FCM algorithm to quantify the statistical differences between different super-pixels, and the clustering results can serve as an indicative measure of the probability that each superpixel belongs to the sea category. Finally, we incorporate this probability into our proposed saliency detection method, facilitating the efficient identification of the ship regions. The experimental results show that the method can robustly and efficiently detect nearshore ship targets.

Discontinuous ground-based synthetic aperture radar monitoring (D-GBSAR) has gained increasing attention in the last five years, while the full processing framework has not been significantly reported. In this paper, a new full framework for D-GBSAR is presented, in which we integrate the advanced technique of image registration, permanent scatterer (PS) selection, phase filtering, repositioning error, and atmospheric phase compensation. Particularly, we reveal that the sensor’s spatial baseline is small thus making it relatively simple to registrate the images. Furthermore, we apply complex mean filtering to mitigate the stochastic noise for better performance. Thereafter, we utilize the newly proposed methods to perform the azimuth-based repositioning error and slant distance-based atmospheric phase compensation. Finally, typical experiments verify the effectiveness of the proposed method, which is comparable with the advanced method and showcases an important technical reference for D-GBSAR applications.

Port traffic flow modeling based on big data is an important research direction in the shipping field, having the task of traffic forecasting for ports worldwide. Graph neural networks have a strong ability to capture the spatial topology characteristics and may be combined with recurrent neural networks or dilated convolution methods in time series prediction, producing a large number of spatiotemporal graph convolution models. Such models have been widely and successfully applied in traffic forecasting. Differing from urban traffic flow data, the statistical time span of port vessel flow and throughput data is large, its spatial span is wide, and the data experience significant fluctuations. Consequently, certain spatiotemporal graph convolution traffic prediction models are unsuitable for shipping scenarios. To address this shortcoming, we have created a unique port flow dataset based on automatic identification system (AIS) and port geographic data. Using theoretical analysis and experimental comparison, we have determined the most appropriate model for shipping predictions based on existing spatiotemporal graph models and have proposed model optimization recommendations for the maritime domain. Our experiment based on an open source traffic forecasting framework to compare the results of multiple existing spatiotemporal graph models under fair conditions with the central ports of Rotterdam, Shanghai, Boston, and Singapore. The results show that Graph WaveNet exhibits better performance in shipping scenarios.

Object detection is a challenging task in computer vision. Now many detection networks can get a good detection result when applying large training dataset. However, annotating sufficient amount of data for training is often time-consuming. To address this problem, a semi-supervised learning based method is proposed in this paper. Semi-supervised learning trains detection networks with few annotated data and massive amount of unannotated data. In the proposed method, Generative Adversarial Network is applied to extract data distribution from unannotated data. The extracted information is then applied to improve the performance of detection network. Experiment shows that the method in this paper greatly improves the detection performance compared with supervised learning using only few annotated data. The results prove that it is possible to achieve acceptable detection result when only few target object is annotated in the training dataset.

Compared with current low-Earth-orbit synthetic aperture radar (SAR), geosynchronous SAR (GEO SAR) is featured with its ultrahigh orbit and ultralong integration time. In this letter, we answer the question whether the orbital perturbation items have effects on GEO SAR imaging during the long integration time and, if so, how they produce errors. To achieve these goals, we first develop a perturbing orbital elements errors model based on perturbation analysis. Then, we propose an accurate analytical expression of first to fourth Doppler parameters for GEO SAR and analyze the effects of perturbing orbital elements errors on the Doppler parameters. Furthermore, the relationship between the perturbing Doppler errors and the phase coherence of GEO SAR signal is deduced. Experiment and simulation results demonstrate the image defocusing caused by orbital perturbation. The conclusions are that the orbital perturbation does affect GEO SAR imaging by producing nonnegligible phase errors and that the required accuracy of the orbit determination should be at centimeter level in the radial direction.

Multi-channel peculiarity is one of the most widely accepted human visual system (HVS) models for perceptual image quality assessment (IQA). Otherwise than extensive studies of channel decomposition and intra-channel distortion measure, relatively scant research effort has been devoted to develop efficient multichannel evaluation pooling strategies. In this paper, we review and address the limitations of the conventional pooling models based on HVS sensitivities-weighted average. Instead, we explore the utilization of machine learning for this pooling problem, since machine learning can establish an optimal and generalized mapping that models the highly complex relationship between the multi-channel distortion evaluations and the perceived image quality. Experiments based on available subjective IQA databases demonstrate the rationality, reliability and robustness of our proposed scheme.

In this paper, we propose a novel remote sensing image compression method based on double-sparsity dictionary learning and universal trellis coded quantization (UTCQ). Recent years have seen a growing interest in the study of natural image compression based on sparse representation and dictionary learning. We show that using the double-sparsity model to learn a dictionary gives much better compression results for remote sensing images, the texture of which is much richer than that of natural images. We also show that the compression performance is improved significantly when advanced quantization and entropy coding strategies are used for encoding the sparse representation coefficients. The proposed method outperforms the existing dictionary-based image coding algorithms. Additionally, our method results in better ratedistortion performance and structural similarity results than CCSDS and JPEG2000 standard.

This letter proposes a novel approach for automated recognition of human-human interactions. First, the motion directions are obtained and the spatial relationships between persons (topological and directional relations) are modeled based on the tracking results. Then, we propose a method to extract the spatial semantics between persons, including front, back, face to face, back to back, and left or right. Finally, we adopt context-free grammar (CFG) to recognize the interactions, and thereinto the production rules are established based on the transformation of spatial semantics. Extensive experiments validate the effectiveness of the proposed approach.

The spatial relationships between objects are the important specificities of the images. This letter proposes a histogram to represent the spatial relationships, and use fuzzy k-nearest neighbors (k-NN) classifier to classify the spatial relationships (left, right, above, below, near, far, inside, outside) with soft labels. Then probabilistic latent semantic analysis (pLSA) is extended by taking into account the spatial relationships between topics (SR-pLSA), and SR-pLSA is used to model the image as the input for support vector machine (SVM) to classify the scene. Experiments demonstrate that the proposed method can achieve high classification accuracy.