Path loss modeling of millimeter waves is regarded as one the most challenging problems in design fifth-generation (5G) mobile communication networks due to high susceptibility various environmental influences. We present a novel convolutional neural network (CNN)-based path method based on set 28 GHz mmWave field measurements suburban scenarios. Enhanced local area multi-scanning (E-LAMS) algorithm that provides CNN with information proposed. A new structure four subnetworks and...

An accurate and efficient path loss modeling method for millimeter wave communications plays a significant role in the large-scale deployment of fifth-generation (5G) mobile communication system. Conventional methods such as deterministic methods, empirical machine learning-based practice, cannot achieve desired level performance terms accuracy. This paper proposes novel model 5G suburban scenarios using deep learning with dilated convolution attention. Dilated is used to alleviate locality...

Diverse sensing techniques have been developed and combined with machine learning method for forest fire detection, but none of them referred to identifying smoldering flaming combustion phases. This study attempts real-time identify different phases using a wireless sensor network (WSN)-based multi-sensor system artificial neural (ANN). Sensors (CO, CO₂, smoke, air temperature relative humidity) were integrated into one node WSN. An experiment was conducted burning materials from residual...

Multi-channel metasurface holography based on polarization multiplexing has received close attention for its application in increasing information capacity. However, most of the existing spin-selective metasurfaces focus utilization cross-polarized channels while neglecting co-polarized channels. Moreover, real-time dynamic switching incident remains a great challenge. This work presents dielectric that combines geometric and propagation phase modulations. It allows control pair orthogonal...

Considering the non-periodic and non-stationary signatures of series arc fault current signals, a time-frequency analyzing method for faults detection is presented based on short-time Fourier transform. The algorithm STFT using Hanning window function described in detail, which length sliding step parameters are discussed to obtain 20ms time-resolution 25Hz frequency-resolution. relationship between variations even order harmonics, odd harmonics 50Hz fundamental component under arcing...

Convolutional Neural Networks (CNNs) have shown surprisingly good performance in both classification and regression problems. Given a floor plan of building indoor measurement data Wi-Fi received signal strength (RSS) at discrete locations, CNN can be trained to approximate the underlying functions path loss. In this article, we propose novel CNN-based loss modeling approach. Based on data, input images are generated for training CNN, which make predictions RSS 5 GHz an usage scenario....

Text-to-image (T2I) generation models have significantly advanced in recent years. However, effective interaction with these is challenging for average users due to the need specialized prompt engineering knowledge and inability perform multi-turn image generation, hindering a dynamic iterative creation process. Recent attempts tried equip Multi-modal Large Language Models (MLLMs) T2I bring user's natural language instructions into reality. Hence, output modality of MLLMs extended, quality...

In this paper, LTPO structure was used to develop 1500PPI LCD VR display device, and LTPS TFT process for peripheral driving ensure excellent performance. The area adopts oxide achieve high transmittance of 1.45%, which is more than 23.9% higher structure. addition, in order pixel density, narrow channel IGZO‐TFT with W/L=3/3.5 applied as TFT. showed PBTS NBTIS stability, the Vth shift (|ΔVth|) PBTS和NBTIS stress were 0.36 V 0.06V.

The development of Multi-modal Large Language Models (MLLMs) enhances (LLMs) with the ability to perceive data formats beyond text, significantly advancing a range downstream applications, such as visual question answering and image captioning. However, substantial computational costs associated processing high-resolution images videos pose barrier their broader adoption. To address this challenge, compressing vision tokens in MLLMs has emerged promising approach reduce inference costs....

Traditional low-cost GPS installed on vehicles and other equipment has a tolerance of tens meters. With the help auxiliary devices and/or methodologies such as Differential GPSs (DGPSs), Assisted (AG-PSs), Real-Time Kinematic (RTK), computer vision etc., positioning accuracy receivers increased lot, but certain amount cost at same time. In this paper, we propose new cooperative vehicular localization scheme for improving fixes based Vehicle-to-vehicle (V2V) communication. The proposed first...

DWI of the pancreas is challenging due to artifacts from physiologic motion, image distortion, and blurring, but has promising applications in pancreatic cancer detection. We conducted a pilot study comparing single-shot multi-shot EPI protocols as well with without new commercially available deep learning (DL) based denoising reconstruction method. Image quality was subjectively scored key metrics. Multi-shot reduced perceived distortion within bed, while combination DL noise.

Traditional image stitching focuses on a single panorama frame without considering the spatial-temporal consistency in videos. The straightforward approach will cause temporal flicking and color inconstancy when it is applied to video task. Besides, inaccurate camera parameters artifacts warping. In this paper, we propose real-time system stitch multiple sequences into panoramic video, which based GPU accelerated correction warping accurate parameters. We extend traditional 2D-Matrix (2D-M)...

The virtual viewpoint is perceived as a new technique in navigation, yet not supported due to the lack of depth information and obscure camera parameters. In this paper, method for achieving close-up view proposed it only uses optical flow build parallax effects realize pseudo 3D projection without using sensor. We develop bidirectional obtain any by proportional interpolation flow. Moreover, with ingenious application optical-flow-value, we achieve clear visual-fidelity magnified results...

Compared with monochromatic holographic displays, color displays have rich depth and information, which can give observers a more pleasant visual effect. Inspiring by the adaptive weighted GS method, we introduced an optimized iterative method based on exponential constraint to calculate phase-only holograms of images. After channel separation target image obtain each component image, input complex amplitude is formed quadratic phase iterated. During iteration, subtract reconstructed from...

Ho and Tw are the geometrical measures of a video head for paring. A vision system was developed which inspects detecting automatically any defects visually noticeable as well measuring values. Efficient image processing hardware has been using two transputers run in parallel. Several new techniques introduced, such an algorithm autofocusing within second, dynamic autothresholding technique produces many threshold values wanted, several defect detection algorithms heads. Experimental results...

Visible images are always slantwise, simultaneity ground control points layout not easy, so we must choose one method which can correct the slantwise image based on few reference points. When exterior orientation elements known, paper build a correction plane founded collinearity equations, and then using gray interpolation. The corrective effect is very visible, speed rapid.