RESUMEN
Person re-identification (re-ID) is one of the essential tasks for modern visual intelligent systems to identify a person from images or videos captured at different times, viewpoints, and spatial positions. In fact, it is easy to make an incorrect estimate for person re-ID in the presence of illumination change, low resolution, and pose differences. To provide a robust and accurate prediction, machine learning techniques are extensively used nowadays. However, learning-based approaches often face difficulties in data imbalance and distinguishing a person from others having strong appearance similarity. To improve the overall re-ID performance, false positives and false negatives should be part of the integral factors in the design of the loss function. In this work, we refine the well-known AGW baseline by incorporating a focal Tversky loss to address the data imbalance issue and facilitate the model to learn effectively from the hard examples. Experimental results show that the proposed re-ID method reaches rank-1 accuracy of 96.2% (with mAP: 94.5) and rank-1 accuracy of 93% (with mAP: 91.4) on Market1501 and DukeMTMC datasets, respectively, outperforming the state-of-the-art approaches.
Asunto(s)
Inteligencia , Humanos , Iluminación , Aprendizaje Automático , Grabación de Cinta de VideoRESUMEN
Robotic arms have been widely used in various industries and have the advantages of cost savings, high productivity, and efficiency. Although robotic arms are good at increasing efficiency in repetitive tasks, they still need to be re-programmed and optimized when new tasks are to be deployed, resulting in detrimental downtime and high cost. It is therefore the objective of this paper to present a learning from demonstration (LfD) robotic system to provide a more intuitive way for robots to efficiently perform tasks through learning from human demonstration on the basis of two major components: understanding through human demonstration and reproduction by robot arm. To understand human demonstration, we propose a vision-based spatial-temporal action detection method to detect human actions that focuses on meticulous hand movement in real time to establish an action base. An object trajectory inductive method is then proposed to obtain a key path for objects manipulated by the human through multiple demonstrations. In robot reproduction, we integrate the sequence of actions in the action base and the key path derived by the object trajectory inductive method for motion planning to reproduce the task demonstrated by the human user. Because of the capability of learning from demonstration, the robot can reproduce the tasks that the human demonstrated with the help of vision sensors in unseen contexts.
Asunto(s)
Robótica , Humanos , Movimiento (Física) , Movimiento , Extremidad Superior , Visión OcularRESUMEN
Survey-grade Lidar brands have commercialized Lidar-based mobile mapping systems (MMSs) for several years now. With this high-end equipment, the high-level accuracy quality of point clouds can be ensured, but unfortunately, their high cost has prevented practical implementation in autonomous driving from being affordable. As an attempt to solve this problem, we present a cost-effective MMS to generate an accurate 3D color point cloud for autonomous vehicles. Among the major processes for color point cloud reconstruction, we first synchronize the timestamps of each sensor. The calibration process between camera and Lidar is developed to obtain the translation and rotation matrices, based on which color attributes can be composed into the corresponding Lidar points. We also employ control points to adjust the point cloud for fine tuning the absolute position. To overcome the limitation of Global Navigation Satellite System/Inertial Measurement Unit (GNSS/IMU) positioning system, we utilize Normal Distribution Transform (NDT) localization to refine the trajectory to solve the multi-scan dispersion issue. Experimental results show that the color point cloud reconstructed by the proposed MMS has a position error in centimeter-level accuracy, meeting the requirement of high definition (HD) maps for autonomous driving usage.
RESUMEN
Action recognition has gained great attention in automatic video analysis, greatly reducing the cost of human resources for smart surveillance. Most methods, however, focus on the detection of only one action event for a single person in a well-segmented video, rather than the recognition of multiple actions performed by more than one person at the same time for an untrimmed video. In this paper, we propose a deep learning-based multiple-person action recognition system for use in various real-time smart surveillance applications. By capturing a video stream of the scene, the proposed system can detect and track multiple people appearing in the scene and subsequently recognize their actions. Thanks to high resolution of the video frames, we establish a zoom-in function to obtain more satisfactory action recognition results when people in the scene become too far from the camera. To further improve the accuracy, recognition results from inflated 3D ConvNet (I3D) with multiple sliding windows are processed by a nonmaximum suppression (NMS) approach to obtain a more robust decision. Experimental results show that the proposed method can perform multiple-person action recognition in real time suitable for applications such as long-term care environments.
Asunto(s)
Identificación Biométrica/instrumentación , Aprendizaje Profundo , Actividades Humanas , Sistemas de Computación , HumanosRESUMEN
Accurate estimation of 3D object pose is highly desirable in a wide range of applications, such as robotics and augmented reality. Although significant advancement has been made for pose estimation, there is room for further improvement. Recent pose estimation systems utilize an iterative refinement process to revise the predicted pose to obtain a better final output. However, such refinement process only takes account of geometric features for pose revision during the iteration. Motivated by this approach, this paper designs a novel iterative refinement process that deals with both color and geometric features for object pose refinement. Experiments show that the proposed method is able to reach 94.74% and 93.2% in ADD(-S) metric with only 2 iterations, outperforming the state-of-the-art methods on the LINEMOD and YCB-Video datasets, respectively.
RESUMEN
Amino-terminated silane compound modification was wet-processed on a silicon wafer using four different solvents to investigate the property of the self-assembled monolayer (SAM) and its influence on the adhesion of electroless deposited nickel-phosphorus (Ni-P) films. Analyzed by various tools including dynamic light scattering, the atomic force microscope, X-ray photoelectron spectroscopy, inductively coupled plasma with mass spectroscopy, a proper link between the processing solvent and SAM quality is established. It is found that at least the chemical compatibility, the polarity, and the acidity of solvents can affect the final morphology of the resultant SAM. Unlike toluene and ethanol that are most frequently chosen in literature, we conclude that isopropyl alcohol (IPA) is a superior solvent for amino-terminated silane compounds. Owing to the good SAM quality formed in IPA, the adhesion of electroless deposited Ni-P films is largely strengthened, even as high as the bulk strength of silicon wafers.
RESUMEN
A single-webcam distance measurement technique for indoor robot localization is proposed in this paper. The proposed localization technique uses webcams that are available in an existing surveillance environment. The developed image-based distance measurement system (IBDMS) and parallel lines distance measurement system (PLDMS) have two merits. Firstly, only one webcam is required for estimating the distance. Secondly, the set-up of IBDMS and PLDMS is easy, which only one known-dimension rectangle pattern is needed, i.e., a ground tile. Some common and simple image processing techniques, i.e., background subtraction are used to capture the robot in real time. Thus, for the purposes of indoor robot localization, the proposed method does not need to use expensive high-resolution webcams and complicated pattern recognition methods but just few simple estimating formulas. From the experimental results, the proposed robot localization method is reliable and effective in an indoor environment.
RESUMEN
This paper reports a new approach to realize direct selective electroless deposition (ELD) without the requirement of photolithography. This method involves sequential silane-compound modifications in which the first modification creates a hydrophobic surface on the TiO2-coated glass using a fluorine-rich alkoxysilane compound, followed by a laser ablation to create the pattern. Then, the entire substrate is immersed into an aqueous solution containing amino-silane equipped Pd nanoparticles for the second modification. Because most substrate surface is hydrophobic, the amino-silane-equipped Pd catalysts can only graft on the laser-ablated zone to accomplish selective ELD.
RESUMEN
In this study, the effect of 3-2-(2-aminoethylamino) ethylamino propyl trimethoxysilane (ETAS) modification and post rapid thermal annealing (RTA) treatment on the adhesion of electroless plated nickel-phosphorus (ELP Ni-P) film on polyvinyl alcohol-capped palladium nanoclusters (PVA-Pd) catalyzed silicon wafers is systematically investigated. Characterized by pull-off adhesion, atomic force microscopy, X-ray spectroscopy and water contact angle, a time-dependent, three-staged ETAS grafting mechanism including islandish grafting, a self-assembly monolayer (SAM) and multi-layer grafting is proposed and this mechanism is well correlated to the pull-off adhesion of ELP Ni-P film. In the absence of RTA, the highest ELP Ni-P film adhesion occurs when ETAS modification approaches SAM, where insufficient or multi-layer ETAS grafting fails to provide satisfactory results. On the other hand, if RTA is applied, the best ELP Ni-P film adhesion happens when ETAS modification is islandish owing to the formation of nickel silicide, where SAM or multi-layer ETAS modification cannot provide satisfactory adhesion because the interaction between ETAS and PVA-Pd has been sabotaged during RTA. Evidenced by microstructural images, we also confirmed that ETAS can act as an efficient barrier layer for nickel diffusion to bulk silicon.
RESUMEN
In this paper, an observer-based direct adaptive fuzzy-neural control scheme is presented for nonaffine nonlinear systems in the presence of unknown structure of nonlinearities. A direct adaptive fuzzy-neural controller and a class of generalized nonlinear systems, which are called nonaffine nonlinear systems, are instead of the indirect one and affine nonlinear systems given by Leu et al. By using implicit function theorem and Taylor series expansion, the observer-based control law and the weight update law of the fuzzy-neural controller are derived for the nonaffine nonlinear systems. Based on strictly-positive-real (SPR) Lyapunov theory, the stability of the closed-loop system can be verified. Moreover, the overall adaptive scheme guarantees that all signals involved are bounded and the output of the closed-loop system will asymptotically track the desired output trajectory. To demonstrate the effectiveness of the proposed method, simulation results are illustrated in this paper.
Asunto(s)
Algoritmos , Lógica Difusa , Modelos Biológicos , Modelos Estadísticos , Redes Neurales de la Computación , Dinámicas no Lineales , Reconocimiento de Normas Patrones Automatizadas/métodos , Simulación por Computador , Metodologías Computacionales , Técnicas de Apoyo para la Decisión , Retroalimentación , Análisis Numérico Asistido por Computador , Procesos EstocásticosRESUMEN
In this paper, a novel approach to adjust both the control points of B-spline membership functions (BMFs) and the weightings of fuzzy-neural networks using a reduced-form genetic algorithm (RGA) is proposed. Fuzzy-neural networks are traditionally trained by using gradient-based methods, which may fall into local minimum during the learning process. To overcome the problems encountered by the conventional learning methods, genetic algorithms are adopted because of their capabilities of directed random search for global optimization. It is well known, however, that the searching speed of the conventional genetic algorithms is not desirable. Such conventional genetic algorithms are inherently incapable of dealing with a vast number (over 100) of adjustable parameters in the fuzzy-neural networks. In this paper, the RGA is proposed by using a sequential-search-based crossover point (SSCP) method in which a better crossover point is determined and only the gene at the specified crossover point is crossed, serving as a single gene crossover operation. Chromosomes consisting of both, the control points of BMFs and the weightings of the fuzzy-neural network are coded as an adjustable vector with real number components that are searched by the RGA. Simulation results have shown that faster convergence of the evolution process searching for an optimal fuzzy-neural network can be achieved. Examples of nonlinear functions approximated by using the fuzzy-neural network via the RGA are demonstrated to illustrate the effectiveness of the proposed method.
RESUMEN
In this paper, we propose a novel design of a GA-based output-feedback direct adaptive fuzzy-neural controller (GODAF controller) for uncertain nonlinear dynamical systems. The weighting factors of the direct adaptive fuzzy-neural controller can successfully be tuned online via a GA approach. Because of the capability of genetic algorithms (GAs) in directed random search for global optimization, one is used to evolutionarily obtain the optimal weighting factors for the fuzzy-neural network. Specifically, we use a reduced-form genetic algorithm (RGA) to adjust the weightings of the fuzzy-neural network. In RGA, a sequential-search -based crossover point (SSCP) method determines a suitable crossover point before a single gene crossover actually takes place so that the speed of searching for an optimal weighting vector of the fuzzy-neural network can be improved. A new fitness function for online tuning the weighting vector of the fuzzy-neural controller is established by the Lyapunov design approach. A supervisory controller is incorporated into the GODAF controller to guarantee the stability of the closed-loop nonlinear system. Examples of nonlinear systems controlled by the GODAF controller are demonstrated to illustrate the effectiveness of the proposed method.
RESUMEN
In this paper, we investigate a novel robust control approach for a class of uncertain nonlinear systems with multiple inputs containing sector nonlinearities and deadzones. Sliding mode control (SMC) is suggested to design stabilizing controllers for these uncertain nonlinear systems. The controllers guarantee the global reaching condition of the sliding mode in these systems. They can work effectively for systems either with or without sector nonlinearities and deadzones in the inputs. Moreover, the controllers ensure that the system trajectories globally exponentially converge to the sliding mode. Illustrative examples are demonstrated to verify the effectiveness of the proposed sliding mode controller.
RESUMEN
Sliding mode controllers for the bilinear systems with time varying uncertainties are developed in this paper. The bilinear coefficient matching condition which is similar to the traditional matching condition for linear system is defined for the homogeneous bilinear systems. It can be seen that the bilinear coefficient matching condition is very limited and is not generally applicable to the nonhomogeneous bilinear system. Thus, the sliding coefficient matching condition is also considered for the bilinear systems with time varying uncertainties. Then, the sufficient conditions are provided for the reaching mode of the time varying uncertain bilinear systems to be guaranteed by the designed sliding mode controllers. Moreover, the stability of the uncertain bilinear systems with the sliding mode controller is discussed. Simulation results are included to illustrate the effectiveness of the proposed sliding mode controllers.
RESUMEN
A novel adaptive fuzzy-neural sliding-mode controller with H(infinity) tracking performance for uncertain nonlinear systems is proposed to attenuate the effects caused by unmodeled dynamics, disturbances and approximate errors. Because of the advantages of fuzzy-neural systems, which can uniformly approximate nonlinear continuous functions to arbitrary accuracy, adaptive fuzzy-neural control theory is then employed to derive the update laws for approximating the uncertain nonlinear functions of the dynamical system. Furthermore, the H(infinity) tracking design technique and the sliding-mode control method are incorporated into the adaptive fuzzy-neural control scheme so that the derived controller is robust with respect to unmodeled dynamics, disturbances and approximate errors. Compared with conventional methods, the proposed approach not only assures closed-loop stability, but also guarantees an H(infinity) tracking performance for the overall system based on a much relaxed assumption without prior knowledge on the upper bound of the lumped uncertainties. Simulation results have demonstrated that the effect of the lumped uncertainties on tracking error is efficiently attenuated, and chattering of the control input is significantly reduced by using the proposed approach.
RESUMEN
This paper proposes a novel method of online modeling and control via the Takagi-Sugeno (T-S) fuzzy-neural model for a class of uncertain nonlinear systems with some kinds of outputs. Although studies about adaptive T-S fuzzy-neural controllers have been made on some nonaffine nonlinear systems, little is known about the more complicated uncertain nonlinear systems. Because the nonlinear functions of the systems are uncertain, traditional T-S fuzzy control methods can model and control them only with great difficulty, if at all. Instead of modeling these uncertain functions directly, we propose that a T-S fuzzy-neural model approximates a so-called virtual linearized system (VLS) of the system, which includes modeling errors and external disturbances. We also propose an online identification algorithm for the VLS and put significant emphasis on robust tracking controller design using an adaptive scheme for the uncertain systems. Moreover, the stability of the closed-loop systems is proven by using strictly positive real Lyapunov theory. The proposed overall scheme guarantees that the outputs of the closed-loop systems asymptotically track the desired output trajectories. To illustrate the effectiveness and applicability of the proposed method, simulation results are given in this paper.
Asunto(s)
Algoritmos , Modelos Teóricos , Redes Neurales de la Computación , Dinámicas no Lineales , Simulación por Computador , Retroalimentación , Sistemas en LíneaRESUMEN
This paper presents a distance measurement method based on pixel number variation of CCD images by referencing to two arbitrarily designated points in the image frames. By establishing a relationship between the displacement of the camera movement along the photographing direction and the difference in pixel count between reference points in the images, the distance from an object can be calculated via the proposed method. To integrate the measuring functions into digital cameras, a circuit design implementing the proposed measuring system in selecting reference points, measuring distance, and displaying measurement results on CCD panel of the digital camera is proposed in this paper. In comparison to pattern recognition or image analysis methods, the proposed measuring approach is simple and straightforward for practical implementation into digital cameras. To validate the performance of the proposed method, measurement results using the proposed method and ultrasonic rangefinders are also presented in this paper.