Internet of Things (IoT) and artificial intelligence (AI) can realize the concept "smart city." Video surveillance in smart cities is, usually, based on a centralized framework which large amounts real-time media data are transmitted to processed cloud. However, cloud relies network connectivity that is sometimes limited or unavailable; thus, not sufficient for processing needed video surveillance. To tackle this problem, edge computing - technique accelerating development AIoT (AI across...

Multispectral object detection entails integrating visible and thermal imaging data is critical for reliable performance of intelligent vehicle systems under adverse weather conditions. However, effectively fusing these heterogeneous modalities remains a major challenge. Existing fusion architectures illumination-aware networks often fail to utilize distinguish between modality-specific information, leading suboptimal in severe weather. To address limitations, we propose...

Recently, automated biometric identification system (ABIS) has wide applications involving automatic and data capture (AIDC), which includes security checking, verifying personal identity to prevent information disclosure or fraud, so on. With the advancement of biotechnology, systems based on biometrics have emerged in market. These require high accuracy ease use. Palm vein is a type that identifies palm features. Compared with other features, recognition provides accurate results received...

Three-dimensional object detection plays a key role in autonomous driving systems. The performance of light and ranging (LiDAR)-based models is limited because the sparsity inhomogeneity point clouds. Researchers have developed numerous camera-LiDAR-based fusion methods to address limitations LiDAR-based models. Still, major differences representations between pixel cloud, these cannot appropriately leverage benefits both types sensors. In this study, we 3-D model based on LiDAR-camera...

In this paper, we present an infrared thermal-based pedestrian detection method that can be applied in nighttime intelligent surveillance systems. Pedestrian plays important role computer vision and automation industry applications, which include video surveillance, automotive robot, smart vehicles. Recently, the improvement deep learning techniques, such as convolutional neural networks (CNNs), have significantly increased accuracy of detection. Normally, optical cameras, e.g....

Recently, as one of the most promising biometric traits, vein has attracted attention both academia and industry because its living body identification convenience acquisition process. State-of-the-art techniques can provide relatively good performance, yet they are limited to specific light sources. Besides, it still poor adaptability multispectral images. Despite great success achieved by convolutional neural networks (CNNs) in various image understanding tasks, often require large...

Pedestrian detection is a high-profile topic in computer vision, part because it has great relevance to autonomous driving and intelligent surveillance applications.However, most pedestrian algorithms perform stably only during the daytime with sufficient illumination.At night, there still room for improvement many challenges exist.These include occlusion caused by objects or crowds, problem of image background segmentation environments varying illumination.In this paper, we propose...

With the advent of "beauty economic era", in which people are paying more attention to beauty and health, health scalp is being increasingly valued.However, current care services limited by problems such as they not automatic objective, results significant, make them unacceptable public.Because these reasons, this study, we focus on obstacles that hairdressers face propose an expert inspection system suitable for determining utilizing deep learning, cloud computing techniques, embedded...

As the global demand for coffee rises, has become a part of daily lives many. The taste brewed is closely related to quality beans, which led many researchers developing automated methods accurately distinguish good beans from bad ones. research often used supervised learning technology by utilizing large sets labeled data training, but labeling requires substantial amount manpower that impractical real production line usage. To solve this problem, we proposed method combines semi-supervised...

In recent years, people's need for coffee beans has skyrocketed, which results in automatic quality recognition being highlighted by researchers. this study, a Convolutional Neural Networks-based method the advantage of having high local feature extraction performance was presented. The experimental show that proposed could be effectively utilized to perform green it achieved an F1-score about 97%.

Skin care products should be tailored to suit different skin types. However, testing can expensive and time-consuming, particularly for students or office workers who may need access specialized equipment. In the present study, we developed a type detection system by using computer-vision deep-learning techniques that easily accessed through mobile phone application. Our integrates with TensorFlow Lite framework on Android platform therefore supports various hardware accelerations easy model...

Multispectral object detection using visible–thermal vision sensors is a crucial method for enabling intelligent vehicles to perceive their environment accurately. However, fusing and complementing multispectral information challenging because of problems such as miscalibration resulting from disparate camera fields view (FOV) pattern discrepancies in images captured at different wavelengths. To address these challenges, we propose novel system named the reinforcement- alignment-based fusion...

Lane detection is an important topic in the self-driving system. Having a stable lane system will assist cars to make decisions order bring more comfortable and safe driving environment driver. In this paper, we use network architecture composed of Encoder-Decoder with Feature Shift Aggregator between them prediction comprehensive; through our dataset, found that some problems such as glitches occur when changing lanes. regard, Data Augmentation Filter respectively solve problem. Finally,...

For Internet of Vehicles applications, reliable autonomous driving systems usually perform the majority their computations on cloud due to limited computing power edge devices. The communication delay between platforms and devices, however, can cause dangerous consequences, particularly for latency-sensitive object detection tasks. Object tasks are also vulnerable significantly degraded model performance caused by unknown objects, which creates unsafe conditions. To address these problems,...

The use of intelligent vehicle technology is increasing; however, this requires further improvement. Semantic segmentation enables vehicles to understand the environment. Although advances have been achieved in a deep learning model for semantic segmentation, large-scale real-world datasets with manual pixel-level annotations, which are expensive, required adequately training these models. Unsupervised domain adaptation (UDA) on synthetic source and allows be adapted target without...

With the rapid progress in autonomous driving technology, integration of multiple sensors into systems has become crucial. Existing methods often use point-level fusion, where LiDAR point clouds are projected onto a plane and fused with RGB features. However, fusion approach leads to loss semantic density from features during transformation process. To overcome this limitation, recent have transformed pixels 3D space using depth prediction techniques, generating virtual clouds. While...

This research addresses the enhancement of face anti-spoofing (FAS) in facial recognition systems (FRS) against sophisticated fraudulent activities. Prior methods primarily focus on extracting features like color, texture, and dynamic variations, yet these struggle to accurately identify common characteristics forged faces, thereby limiting generalization practical scenarios. study aims propose a novel representation learning framework incorporating adversarial algorithms, segregate into...

Palm-vein authentication is a secure and highly accurate vein feature technology that has recently gained lot of attention. Convolutional neural networks (CNNs) provide relatively high performance in the field image processing, computer vision, have been adapted for learning palm-vein images. However, they often require computation not only are infeasible real-time verification but also challenge to apply on mobile devices. To address this limitation, we proposed lightweight MobileNet based...