As the deep learning exhibits strong advantages in feature extraction, it has been widely used field of computer vision and among others, gradually replaced traditional machine algorithms. This paper first reviews main ideas learning, displays several related frequently-used algorithms for vision. Afterwards, current research status is demonstrated this paper, particularly applications field.

The nucleated-cell differential count on the bone marrow aspirate smears is required for clinical diagnosis of hematological malignancy. Manual time consuming and lacks consistency. In this study, a novel artificial intelligence (AI)-based system was developed to perform cell automatic classification cells determine its potential applications.Bone were collected from Xinqiao Hospital Army Medical University. First, an automated analysis (Morphogo) scanned generated whole digital images...

Recently, with the increasing difficulty of parking in cities, wireless vehicle detectors (WVDs) based on Internet-of-Things (IoT) technology and magnetic sensor have been widely used to obtain space status data, which lay foundation for smart management system (SPMS). However, there is a weak region under every where variation even smaller than that distorted by adjacent vehicles, leads accuracy decrease magnetometer-based detection algorithm (MB-VDA). To improve performance WVDs, an...

In environmental monitoring, such as air quality accurate and continuous data collection through a wireless sensor network is of significance. this paper, we describe deployed participatory sensing system which monitors the parameters nodes or crowd-owned sensors, e.g., smartphones with affordable additional sensors. The noisy sensory information processed using Kalman filter an metric computed used in incentive scheme. demonstration, show that proposed effectively improved accuracy coverage.

In the future sixth-generation wireless communications, sensor networks (WSNs) are expected to support target sensing, information communication, and computational tasks concurrently over same spectrum. Integrated sensing communication (ISAC) over-the-air computation (AirComp) arise as two promising techniques. this article, we design an integrated computing framework improve spectrum efficiency quality of service in WSNs. We investigate omnidirectional directional beampattern designs...

Visual object tracking aims to estimate the location of an arbitrary target in a video sequence given its initial bounding box. By utilizing offline feature learning, siamese paradigm has recently been leading framework for high performance tracking. However, current existing trackers either heavily rely on complicated anchor-based detection networks or lack ability resist distractors. In this paper, we propose Siamese keypoint prediction network (SiamKPN) address these challenges. Upon...

Performance optimization of the inter-satellite link is key to improve performance satellite networks. This article proposes a dynamic power allocation method for Low Earth Orbit (LEO)/Medium (MEO) satellites based on computing LEO satellites. A utility function designed according transmission characteristic link. The existence Nash Equilibrium (NE) proved super-modular game theory. obtains its balanced solution through Newton iteration method. In this process, initial orbit parameters and...

It's very dangerous for the elderly to fall, so fall detection is important in nowadays society. This paper addresses detect activities by combing Gaussian mixture model (GMM) and special-temporal analysis of aspect ratio. First, we use GMM get background part foreground from an image. After morphological operations, some small gaps are removed empirical knowledge part. Second, calculate ratio feature minimum external rectangle a human body. Through spatial-temporal ratio, output behaviour...

Feature-based pedestrian detection method is currently the mainstream direction to solve problem of detection. In this kind method, whether appropriate feature can be extracted key comprehensive performance whole system. It believed that appearance a better captured by combination edge/local shape and texture feature. field, current simply concatenate HOG (histogram oriented gradient) features LBP (local binary pattern) from an image produce new with large dimension. This achieves at cost...

A Tracking-Learning-Detection (TLD) is a tracking framework that decomposes the long-term task into tracking, learning and detection. To improve accuracy precision with TLD, constraint target cannot move significant distance within short time period limited velocity added, displacement in previous frame analyzed used to reduce detection range of TLD detector. Instead using Median Flow Tracker, K-Nearest Neighbor (Knn) matching Fused Lukas-Kanade method ORB feature points track targets....

For online multi-object tracking, the appearance model of a target is essential. It has to be consistent within track and discriminative between tracks different targets. To satisfy these requirements, new two-stage frame-by-frame method that takes advantage both global local features proposed. In this paper, first, targets are tracked with their feature more than features. Then, feature-MSER (maximally stable extremal regions)-based color histogram proposed used for tracking. Experiments on...

In practical applications such as cancer diagnosis and industrial detection, there is a critical demand for fast fluorescence lifetime imaging (Fast-FLIM). The Fast-FLIM systems suitable complex environments are typically achieved by enhancing the hardware performance of time-correlated single-photon counting (TCSPC), with an acquisition speed about few frames per second (fps). However, due to limitation acquisition, still far from application. synchroscan streak camera (SC) maps signals...

Given the central role of webcams in monitoring physical surroundings, it behooves research community to understand characteristics webcams' distribution and their privacy/security implications. In this paper, we conduct first systematic study on live from both aggregation sites individual (webpages/IP hosts). We propose a series efficient, automated techniques for detecting fingerprinting webcams. particular, leverage distributed algorithms detect generate webcam fingerprints by utilizing...

This paper focuses on probe-based queue length estimation for congested arterial roads. Queue and the portion of non-stopped vehicles have a strong relationship since latter will decrease as former increases. The proposed model is based this has two steps. first step using link travel time data from few probe vehicles’ data. second vehicles. evaluated by simulation errors are shown at different market penetration levels

Computer generated holograms (CGHs) are state-of-the-art components in optical systems, and widely used combination with standard Fizeau interferometers. The primary role of the CGHs is to generate reference wavefront any desired shape. A method interferometrically measuring large convex lens adopted, results from a set experiments that demonstrate accuracy simplicity performing holographic test presented. direct comparison CGH measurement compensation shows excellent agreement. Finally,...

Outline extraction is the key technology of automatically generating roll trajectory in flexible forming process. The main purpose this article to extract outline from grid lines 3D surface STL (Stereo lithography) files. Using functions file stream and character manipulation provided by C++ Standard Library read file, using GLUT(OpenGL Utility Toolkit) build models, investigates algorithms for based on MFC Single Document Interface (SDI) Application.

With the increasing number of private vehicles, intelligent parking space management systems (IPSMS) are widely used to analyze spaces status information and improve efficiency. The wireless vehicle detectors (WVD) based on magnetic sensor have been in IPSMS collect spaces. However, due existence signal blind zone between front rear wheels high-chassis accuracy single detector decreases markedly detecting vehicles such as SUVs, MPVs, Pickups, etc. In order performance WVDs, this paper...

This paper presents a sampling speed-enhancement technique for SAR ADCs under near-threshold supply voltages. The proposed level-shifted boosting circuit generates sharp falling edges the clock, which is found key factor limiting sample speed ultra-low A 0.35V 8b 12MS/s ADC designed in 65nm CMOS technology to prove this technique. post-layout simulated consumes only 6.71/iW and achieves SNDR of 48.8dB at Nyquist input, resulting figure-of-merit (FoM) 2.47 fj/convertion-step. Simulation...