Search

Anomaly detection in videos is a challenging problem in computer vision. Most existing methods need supervised information to train their models, which limits their applications in real world scenario. Therefore, self-trained methods which do not need manually labels receive increasing attentions recently. In this paper, we propose a novel self-trained video anomaly detection method based on teacher-student model. The teacher-student architecture can significantly improve the performance of self-trained video anomaly detection by utilizing the unlabeled samples. We test our method on two surveillance datasets. Experiment results show that our method achieves better performance than state-of-the-art unsupervised methods on both datasets and achieves comparable performance as semi-supervised methods, which experimentally proves the effectiveness of our method.

Steel strip is an important part of steel production. The defect detection for steel strip is a thriving need for steel factories. Workers must monitor the assembly line and detect the defects artificially. Automation of defect detection can save human resource and create values for industry.Nowadays, as the development of computation resources and detection algorithms, there are more and more effective defect detection solutions. In this paper, a new neural network called Inception Dual Network (IDN) is proposed to solve steel defect detection problem. IDN maximize the utilization of residual shortcut connection and concatenation connection through a dual channel structure with Google’s Inception concept. IDN combines classical feature maps to optimize the classification performance. At the same time, a steel strip dataset of 4000 kinds is collected and preprocessed in this paper for other researchers. On the dataset of this paper, IDN has achieved the best performance among some state-of-the-art structure with similar number of parameters while keeping real-time. It can not only be applied on the particular problem in this paper, but also other classical problems.

In order to realize the flexibility and positioning reliability of the positioning manipulator of the minimally invasive vascular interventional surgery robot system, and to avoid the problem of fixed programming of the active positioning manipulator, a new passive positioning manipulator was designed. First, based on the requirements of the system and the self-balancing conditions of the passive positioning manipulator, a configuration of the manipulator is proposed, and the degree of freedom and working space are verified to meet the requirements of the working conditions. Then, according to the characteristics of simultaneously locking and release of the joints when the manipulator works, a locking scheme of multi-joint is proposed, and the design of the mechanical structure is carried out in detail. Finally, the positioning manipulator can achieve the function of being dragged under artificial operation and locating whenever and wherever necessary.

(1−x)(Na 0.475 K 0.475 Li 0.05 )NbO 3−x (Er 0.5 Na 0.5 )TiO 3 (0≤x≤0.025) lead-free multifunctional ceramics were prepared by a conventional solid-state reaction method. Their phase, photoluminescence, dielectric, ferroelectric and piezoelectric properties were investigated systematically by the X-ray Diffraction (XRD), scanning electron microscopy (SEM) spectrophotometer and electrical property measurements. The relationship of composition, microstructure and electrical properties and the mechanisms of up-conversion and down conversion emissions were analyzed detailed. The results show that the ceramics exhibit the optimum piezoelectric and ferroelectric properties when their compositions lie at the nearby of the morphotropic phase boundary (MPB). Moreover, these ceramics can emit strong 550 nm green light when they were excited by a nearinfrared (980 nm) light and a 486 nm blue light, respectively. Together with the excellent ferroelectric properties and good dielectric properties, these ceramics should be one of the promising candidates for multifunctional optoelectronic applications.

Pure phase orthorhombic CaTiO 3 :Pr3+ was synthesized by a conventional solid-state method and its elastico-mechanoluminescence (EML) was characterized. The results revealed that the EML emission peak was centered at 612 nm, similar to its photoluminescence (PL) (Fig.1), indicating both EML and PL originated from 1D 2 -3H 4 transitions of Pr3+. Thermoluminescence and afterglow properties showed tunneling effect in its anomalous fading. The EML mechanism of CaTiO 3 :Pr3+ was discussed and thought to be attribute to tunneling effect instead of piezoelectricity. Lattice distortion caused by strain diminished the distance between electron captured by trap and Pr4+, and thus increased the probability of immigration of electron to Pr4+, eventually leading to EML.

In this paper we report a spiking-resistant all-Al rear junction n-type monocrystalline Si solar cell with an efficiency of 18%. This 1×1 inch2 cell is fabricated by ionic liquid based electroplating of Al as the front electrode on an n-type Si solar cell with screen-printed Al as the rear junction. Partially-processed cells are prepared in an industrial R&D line of Canadian Solar. The cells are patterned by photolithography to open the front contact. A seed layer of Ni is deposited by sputtering to assist Al electroplating. A finished cell is annealed at different temperatures from 150–450°C to evaluate the effect of temperature on its performance. Compared to a p-type all-Al cell we reported before, the n-type all-Al cell is more resistant to high temperature Al spiking. Further comparison between this cell which has an electroplated Al front electrode and control cells with a screen-printed Ag front electrode demonstrates the feasibility of replacing Ag with electroplated Al as the front electrode.

Ion-implantation offers numerous advantages (i.e. single side precise control and reproducibility of the dopant, simultaneous SiO2 passivation during annealing, no phosphosilicate glass formation) for solar cell manufacturing. Canadian Solar Inc. (CSI) has developed an average efficiency 19.23% blank emitter solar cell (156mm Cz) process. In order to improve the solar cell efficiency, focus is placed on the well known advanced PERC solar cell architecture with optimized back side passivation. The approach is to combine the surface passivation provided by a thin ALD aluminum oxide layer grown after the post implantation annealing (PIA) process with a deposited capping silicon nitride layer. Laser ablation and proper aluminum paste is also used to locally remove the dielectric layers and to form local contact. Based on this development, implanted emitter and local Al-BSF with Al2O3/SiNx back passivation are integrated in solar cells reaching an average efficiency 19.96%, and champion 20.12%.

In this paper, an automatic gain control amplifier (AGC) is proposed. The proposed AGC aims at a large input and output swing with a rapid tracking capacity. Moreover, the proposed AGC has a quite simple structure, which is very important for High intensity focused ultrasound (HIFU) applications in which an array of identical circuits must be used. The designed circuit was simulated in a CMOS 0.35μm technology. The required performances have been confirmed by simulation results: a large ratio of input signal of 20 has been achieved; a good linearity with a relative error lower than 0.62% is observed; the output peak-to-peak signal can be as high as 2V; the tracking time is less than 6μs, for an input signal of 1MHz.

This paper presents an auto tuning system using a variable synchronous switched capacitor for controlling the resonant frequency of a half bridge voltage mode resonant converter. This original tuning topology does not need the addition of any bulky magnetic and consequently is naturally compatible with MRI (Magnetic Resonance Imaging) equipments.

In this paper, a micro Pirani gauge sensor is studied in the purpose to be exploited for multifunction. Based on the gauge's experimental results, an adequate conditioning circuit and a single configurable CMOS readout have been proposed allowing exploiting the gauge in an original way for multi measurements. Furthermore, the proposed readout is designed in such a way that it offers the possibility to exploit transient response of the sensor, which, unlike the conventional way, extends the measurable pressure range remarkably beyond atmosphere. The proposed functionality has been confirmed by simulation results.

The passivated emitter and rear solar cells (PERC) have been investigated for industrial application in recent years. For PERC cells, the rear side is passivated by dielectric films, whereby the dielectric films are locally opened for contact formation. In screen-printed PERC cells, however, the local contacts suffer from Kirkendall voids and weak BSF layer intensely. In this paper, the influence of different dot contact geometries on the local contact formation of rear passivated solar cells was investigated. Decreasing the contact pitch or increasing the opening size, the presence of voids decreased drastically. Besides, the quality of BSF layer is improved.

Silicon based chalcopyrite (with the compounds composition I(B)-III(A)-VI(A)2) hetero-j unction solar cells have been reported with the potential of high conversion efficiency theoretically. The hetero-j unction solar cell with PNN+ structure is fabricated as: front metal electrodes/ transparent conduction oxide (TCO)/ p-type chalcopyrite layer emitter/ n-type silicon base/ n+ silicon back surface doping/ rear metal electrodes. By inserting artificially one thin order defect compound (ODC) layers with a high bulk density of states, and one thin interface layer to represent the total interface states between the chalcopyrite and silicon, the impacts of the interface defects on PNN+ hetero-j unction solar cells were studied by the means of numerical analysis with AMPS-1D simulation. Our results indicate that band bending, determined by band offset at the ODC/c-Si hetero-interface, could be critical to cell performance. The optimum electron affinity of ODC layer is 4.4eV The total volumetric density of states (NDB) in ODC layer should be controlled lower than 1×1018 cm3, and the density of ODC/c-Si interface state (Dit) should be lower than 1×103 cm-2, to obtain the best cell performance.

Low cost Rear Ag paste is made by silver plated-coated Copper and Tin powder to replace part of pure Ag powder in the paste to reduce the cost of solar cell manufacture and also improve the adhesion of the solar cell busbar. The key issue to use Ag coated copper and tin is the oxidization of cooper and tin during the drying and firing process which will reduce the solderability and adhesion by using this system. We use N2 protection during drying and firing process to prevent the oxidization of cooper and tin. Therefore, the solderability and adhesion has been increased a lot to achieve the industry testing standard. Also Rs reduction is also a big achievement for this paste system.

This paper demonstrates high-efficiency solar cells made on commercial grade cast monocrystalline (cast-mono) silicon using ion implantation. 18.9% efficient solar cells on 156 mm square cast-mono wafers from Canadian Solar are fabricated featuring an oxide passivated emitter, aluminum back-surface field and full industrial processing. Further investigation revealed that the cast-mono cells can benefit from the increased short-circuit current density due to random pyramid texture, compared to isotexture, even down to 60% area of grain on a wafer. This paper highlights the importance of improving the quality and yield of cast-mono wafers and the benefits of ion implant technology to produce high-efficiency cast-mono cells at lower cost. Also highlighted is the importance of increased light trapping from alkaline texture of cast-mono wafers.

Based on time-domain quality factor (Q-factor) measurement principle, we have proposed several architectures allowing on chip Q-factor monitoring. In our latest architecture, not only the major error sources are removed but also the remaining error source, the offset of OTA, is cancelled automatically thanks to an original joint implementation of both peak detector and comparator by a reconfigurable circuit, leading to a high accuracy measurement of Q-factor. In this paper, a new implementation of the latest architecture is proposed. With the newly proposed implementation, not only the best expectable measurement accuracy in theory can be achieved, which is 0.1% in our case, more advantages can be benefited as well.

This paper presents direct torque control principle for a permanent magnet synchronous motor. Mathematical model of the permanent magnet linear synchronous motor is further described on the basis of this principle. The thrust control and flux linkage control are studied in this paper. In order to implement direct trust control, voltage vectors are suitably selected in the system according to the error between the true values and their reference of the trust and primary flux linkage. Moreover this paper introduces the realization of some key parts for the control system and analyzes the feasibility of direct thrust control of the linear motor. Some simulations with SIMULINK of MATLAB have been given to demonstrate validity of the control system.

Silicon based chalcopyrite (with the compounds composition I(B)-III(A)-VI(A) 2 ) hetero-junction solar cells have been reported with the potential of high conversion efficiency theoretically. The hetero-junction solar cell with PNN+ structure is fabricated as: front metal electrodes/ transparent conductive oxide (TCO)/ p-type chalcopyrite layer emitter/ n-type silicon base/ n+ silicon back surface doping/ rear metal electrodes. By the means of numerical analysis with AMPS-1D simulation, the impacts of chalcopyrite emitter carrier density and Si base doping concentration were studied. Higher chalcopyrite acceptor density N A and TCO work function Ф help to obtain higher efficiency until N A =7×1016/cm3. As for n-type silicon base doping, the concentration N D should be as large as possible for more space charge and electric field in space charge region (SCR). Normally it is hard to get heavily doped n-type feedstock and wafers. Instead of entirely heavy doping silicon, a thin heavily doped layer (N D =1×1020/cm3) were inserted between chalcopyrite layer and silicon base (N D =1×1016/cm3), and a similar conversion efficiency was obtained.

Electric arc furnaces (EAFs) represent one of the most disturbing loads in the subtransmission or transmission electric power systems. Therefore, it is necessary to build a practical model to described the behavior of EAF in the simulation of power system for power quality issues. This EAF operation, the seemingly random arc paper deals with the modeling of EAF based on the combination of extended Kalman filter to identify the parameter of arc current and the power balance equation to obtain the dynamic, multi-valued v - i characteristics of EAF load. The whole EAF systems are simulated by means of power system blockset in MATLAB to validate the proposed EAF model. This model can also be used to assess the impact of the new plant or highly varying nonlinear loads that exhibit chaos in power systems

DC electric arc furnace is an important nonlinear time-varying load in power system. Due to the adverse effects produced by the operation of arc furnace, it is important to build a practical model to described the behavior of electric arc furnace. The electrical fluctuations in the arc furnace voltage have proven to be chaotic in nature. Therefore, this paper deals with the problem of DC electric arc furnace modeling using the chaos theory and neural network. The radial basis function neural network is used to predict the arc voltage of arc furnace with one-step and multi-step ahead based on the embedding dimension in the reconstructed phase space. The results can also be applied to estimate the future state of arc furnace for control purpose.

Recently, we have found a new phenomenon, named elasticoluminescence, in which intensive visible light can be emitted by the application of mechanical stress in the elastic deformation region. This phenomenon, contrary to piezoelectricity, can sense changes in mechanical stress without any electrical contacts or electrodes. Elastico-deformation luminescence has been investigated systematically using precisely controlled pure-phase of various oxides and their mixed phases. This study reveals that only certain ferroelectric phases produce strong elastico-deformation luminescence; other pure-phases of non-piezoelectricity show no deformation luminescence. Correlation of deformation luminescence and crystal structure has been found. This finding can be applied in designing strong elastico-deformation-luminescent materials. Strong elasticoluminescence materials and/or coatings can give real-time stress images, so that it is a prospective candidate for developing a new non-destructive evaluation technology.

Pr/sup 3+/-doped (Ba, Sr)(Ti, Al)O/sub 3/ ceramics were prepared using conventional solid-stale reaction technique. Multiple functions of electrostriction and photoluminescence were realized in the ceramics. Red emission (612 nm) was observed in the samples and the intensity got its largest value in the sample of 40 mol% Sr addition. Although the electrostrictive strain decreased in the Pr and Al doped samples, the temperature dependence is remarkably improved comparable to the undoped sample.