Contour detection serves as the basis of a variety computer vision tasks such image segmentation and object recognition. The mainstream works to address this problem focus on designing engineered gradient features. In work, we show that contour accuracy can be improved by instead making use deep features learned from convolutional neural networks (CNNs). While rather than using blackbox feature extractor, customize training strategy partitioning (positive) data into subclasses fitting each...

Abstract As a crucial component of product quality surveillance in the industrial manufacturing field, surface defect detection plays pivotal role achieving automation. Although numerous methods have been widely used steel production, complicated characteristics still severely affect accuracy and efficiency methods. To solve this problem, we propose multi-path feature aggregation network that can significantly improve performance. Firstly, design an efficient extraction module combining...

Compared with daytime flights optimal visibility, images captured during the aircraft's landing phase under hazy conditions exhibit significant degradation in quality, thereby increasing risk of runway incursions, whereas current dehazing algorithm performs poorly on airport dataset and fails to meet specific requirement clearly capturing foreground information. To address above issue, considering changing perspective aircraft landing, we propose an edge-guided attention network for image...

With the continuous increase of surface quality requirements in steel industry manufacturing, defect detection has received extensive attention. Correct and rapid defects can significantly improve product productivity. The existing methods accuracy by expanding depth networks or using various feature fusion technologies, but reduce computational efficiency. To achieve balance precision speed, we propose an improved network based on Faster R-CNN for surface. Firstly, emerging ConvNeXt...

High-precision and high-density three-dimensional point cloud models usually contain redundant data, which implies extra time hardware costs in the subsequent data processing stage. To analyze extract more effectively, must be simplified before processing. Given that simplification sensitive to features ensure valid information can saved, this paper, a new algorithm for scattered clouds with feature preservation, reduce amount of while retaining is proposed. First, Delaunay neighborhood...

Feature representation is the key to hyperspectral images (HSI) inpainting. Existing works mainly focus on using spectral and temporal auxiliary inpainting corrupted region, which were proved be low robust for all bands missing high requirements image acquisition. In this work, we propose an end-to-end framework HSI based convolutional neural networks, does not require makes full use of both characteristics spatial information. For characteristics, a channel attention mechanism proposed...

Coded structured light can rapidly acquire the shape of unknown surfaces by projecting suitable patterns onto a measuring surface and grabbing distorted with camera. By analyzing deformation appearing in images, depth information be calculated. This paper presents new concise efficient mathematical model for coded measurement system to obtain information. The interrelations among parameters errors are investigated. Based on geometric structure, results affecting object imaging obtained....

Linear Discriminant Analysis (LDA) has been widely used in appearance-based face recognition. However, it requires lots of training samples for each person with respect to the large dimensionality image space, which is difficult collect reality. To overcome severe constraint sample deficiency, approaches based on single per (SSPP) arise past decades. Though making great improvements years, these methods still suffer from low accuracy when dealing high dimensional features. In this paper, we...

Insulators play a crucial role in ensuring the normal operation of power system, and creepage distance is an important electrical parameter insulators. Most available solutions focus mainly on offline measurement methods, online for insulator transmission lines remains challenging task. To address this issue to further improve corresponding work efficiency, method presented paper. Considering glass material long measuring distance, recognizes type indirectly by calculating structural...

With the recent development of deep convolutional neural networks and large-scale datasets, face recognition has made remarkable progress been widely used in various applications. However, unlike existing public many real-world scenarios recognition, depth training dataset is shallow, which means that only two images are available for each ID. non-uniform increase samples, such issue converted to a more general case, known as long-tail learning, suffers from data imbalance intra-class...

Ellipse detection has a very wide range of applications in the field industrial production, especially geometric metallurgical hinge pins. However, factors images, such as small object size and incomplete ellipse image boundary, bring challenges to detection, which cannot be solved by existing methods. This paper proposes method for utilizes extended proposal operation prevent loss rotation angle features during regression. Moreover, Gaussian distance conforming axioms is adopted combined...

Due to the limited varieties and sizes of existing public hyperspectral image (HSI) datasets, classification accuracies are higher than 99% with convolutional neural networks (CNNs). In this paper, we presented a new HSI dataset named Shandong Feicheng, whose size pixel quantity much larger. It also has larger intra-class variance smaller inter-class variance. State-of-the-art methods were compared on it verify its diversity. Otherwise, reduce overfitting caused by imbalance between high...

Current saliency-based defect detection methods show promise in industrial settings, but the unpredictability of defects steel production environments complicates dataset creation, hampering model performance. Existing data augmentation approaches using generative models often require pixel-level annotations, which are time-consuming and resource-intensive. To address this, we introduce DefFiller, a mask-conditioned generation method that leverages layout-to-image diffusion model. DefFiller...

This paper establishes a multi-view stereoscopic angle model. Through the principle of optical imaging and binocular parallax, factors which impact on stereo can be found out. Using model, we quantitatively simulate display system find relationship between shooting distance angle. Depending formula for maximum fusion limit human eyes, theoretical minimum calculated. Multi-view model better guide to system.

Per-pixel hand detection plays an important role in many human–computer interaction applications while accurate and robust remains a challenging task due to the large appearance variance of hands images. We introduce per-pixel system using one single depth image. propose circle sampling depth-context feature for regions representation, multilayered model is built detection. Finally, postprocessing step based on spatial constraints applied refine results further improve accuracy. evaluate...

多视点立体视角是FTV(Free-viewpoint TV)多视点采集/立体显示的关键因素，该文建立了多视点立体视角的数学模型。根据此模型获得了制约立体角度的相关因素，并定量模拟出用水平摄像机阵列拍摄物体时，拍摄距离与立体视角的关系；进一步根据人眼最大融合极限求出可防止产生重影、双轮廓的最大立体角与拍摄距离和观看距离之间的关系。最后导出立体角度与视点数之间的制约关系。仿真实验验证了所得结果的正确性。该文结果可为优化自由视点多视采集/立体显示系统提供理论依据。

Bag of Contour Fragments (BoCF), derived from the well-known Bag-of-Features (BoF), is an effective framework for shape representation. The feature pooling in this a critical step, while either max or average not learnable process. In paper, we aim at learning function which adaptive to input contour fragment features instead. Towards end, formulate our as weighted sum and pooling, where weight expressed by activation features. To automatically learn weight, output fed into SVM classifier...