Although convolutional neural networks (CNNs) have become the mainstream for image processing and achieved great success in past decade, due to local characteristics, CNN is difficult obtain global long-range semantical information. Moreover, some scenes, pure RGB image-based model accurately identify pixel classification finely segment edge of objects. This study presents a hierarchical vision Transformer named Swin-RGB-D incorporate exploit depth information images supplement enhance...

In order to obtain global contextual information precisely from videos with heavy camera motions and scene changes, this study proposes an improved spatiotemporal two-stream neural network architecture a novel convolutional fusion layer. The three main improvements of are: 1) the Resnet-101 has been integrated into two streams target independently; 2) kinds feature maps (i.e., optical flow motion RGB-channel information) obtained by corresponding convolution layer respectively are...

Abstract Fringe projection profilometry (FPP) is a widely used non-contact 3D measurement method. Though maturing in the last decade, it remains significant challenge when facing phase unwrapping of measured object surfaces single-shot setting. With rapid development deep learning techniques, adoption data-driven approach gaining popularity field optical metrology. This study proposes new absolute recovery method based on devised single-stage network. The aim to ensure high-quality from...

Centreline extraction is one of the critical processes structured light systems that play an increasingly important role in industry 4.0. This paper proposes a novel algorithm to extract centreline by combining improved skeleton thinning initial and grey centroid establish final centreline. It consists several steps: Firstly, image difference operation thresholding are undertaken, Canny edge detection applied locate approximate position laser stripe reduce matrix calculation cost; Secondly,...

RNA-seq analysis was used to identify differentially expressed genes (DEGs) at the genetic level in longissimus dorsi muscle from two pigs investigate mechanisms underlying difference meat quality between Debao and Landrace pigs. Then, these DEGs underwent functional annotation, Gene Ontology (GO) Kyoto Encyclopedia of Genes Genomes (KEGG) pathway enrichment, protein-protein interaction (PPI) analyses. Finally, expression levels specific were assessed using qRT-PCR. The reference genome...

Phase measurement profilometry (PMP) is primarily employed to analyze the morphology of a functional surface with precision. Historically, one most complex and persistent challenges in PMP has been reducing errors stemming from inconsistent indicators at edges surface. In response this challenge, we propose an optimized error compensation methodology specifically designed handle edge artefacts. This introduces Hilbert transform object albedo as tools detect artefact region that need be...

With the development of intelligent manufacturing, production and assembly accuracy components in factories is increasing line with growing demand. However, traditional manual quality inspection inefficient, inaccurate, costly. To this end, digital optical imaging techniques are used to achieve inspection. during reconstruction process, high reflectivity object materials affects speed results. overcome these problems, study investigated three-dimensional (3D) based on laser scanning. It...

Abstract In order to solve the problem that wrapped phase cannot be unwrapped effectively, this paper proposes an algorithm combining branch-cut method and rhombus unwrapping for unwrap phase. The combines advantages of two methods, at first, Goldstein’s is used find residual points in wrapping phase, then non-residual area finally obtained by using cubic spline interpolation method, so as ensure each pixel can smoothly. simulation results verify feasibility effectiveness algorithm, combined...

Fall detection has always been challenging. There are many current works that have achieved good results in fall detections, but most of the existing only consider single feature dimension (i.e., temporal features or spatial features). This study proposes a method combines LSTM and XGBoost model to detect human body falls through fusing both features. It starts with extracting bond points on each frame video by using AlphaPose. In second step, three vertical height, aspect ratio external...

Abstract This study proposes a novel filter, namely areal RrVMD based on Variational Mode Decomposition (VMD), for decomposing surface texture into the form, waviness and roughness. VMD is one of latest signal decomposition techniques has been introduced field metrology recently. The paper develops residual-restrained method to further improved algorithm. It consists three processing steps: firstly, calculating robust weight function; secondly, corresponding k modes residual by using devised...

Realizing the automated and online detection of crowd anomalies from surveillance CCTVs is a research-intensive application-demanding task. This research proposes novel technique for detecting abnormalities through analyzing spatial temporal features input video signals. integrated solution defines an image descriptor that reflects global motion information over time. A non-linear SVM has then been adopted to classify dominant or large-scale abnormal behaviors. The work reported focused on:...

The powerful parallel computing capability of modern GPU (Graphics Processing Unit) processors has attracted increasing attentions researchers and engineers who had conducted a large number GPU-based acceleration research projects. However, current single based solutions are still incapable fulfilling the real-time computational requirements from latest big data applications. Thus, multi-GPU solution become trend for many application attempts. In those cases, load balancing over multiple...

In order to achieve static hand gesture recognization within complex skin-like background regions in an effective and intelligent manner, this study proposed integrated recognition model based on the improved centroid watershed algorithm (ICWA) a dual-channel convolutional neural network (DCCNN) structure. The effectiveness of approach stemmed from more accurate segmentation gestures original image by using ICWA. segmented corresponding Local Binary Patterns (LBP) features extracted then...

Currently, electro-bicycles are one of the most convenient and affordable traveling tools. However, when an electro-bicycle enters a crowded elevator, it will not only affect service life but also has great potential safety hazards battery charges at home. Thus, this study focuses on detecting various intelligently in elevators online manner through improving YOLO (You Only Look Once) v4 network. The algorithm high detection accuracy speed due to its single stage mode priori frame mechanism....

Abstract In order to solve the problem that existing adaptive window stereo matching algorithms have insufficient feature extraction in low-texture regions, resulting low accuracy. An algorithm based on gradient is proposed. Firstly, Sobel operator used extract value of each pixel image. Then, divided into high, medium and texture regions according value. Next, different arm length thresholds are assigned region pixels, windows generated dynamically color threshold. Finally, pixels closer...

Gaussian filtering has been extensively used in the field of surface metrology. However, computing performance becomes a core bottleneck for algorithm based applications when facing large-scale and/or real-time data processing. Although researchers tried to accelerate by using GPU (Graphics Processing Unit), single still fail meet and requirements texture micro- nano-measurements. Therefore, solve this problem, paper proposes node multi-GPUs framework 2D algorithm. This presents that devised...

Waste sorting and recycling play a significant role in carbon neutrality, the government has promoted waste stations various cities while have limited efficiency due to absence of intelligent surveillance systems monitor analyze scene stations, especially detect irregularly dropped garbage. To take most advantage these this study proposes an improved YOLO (You Only Look Once) v5s detector named YOLOv5s-Garbage real-time. This enhances its ability garbage by introducing CBAM (Convolutional...