Efficiency, robustness and accuracy in picky chart parsing
高精度追踪与活动平台-摄像头W Fu, L Gao说明书
High Accuracy Tracking with an Active Pan-Tilt-Zoom CameraW. Fu, L. GaoSchool of Electronics and Information EngineeringXi’an Technological UniversityXi’an, ChinaAbstract—Traditional PTZ tracking system focus on tracking algorithm, but PTZ camera control is not taken seriously and the control method has the large deviations. The algorithm for PTZ camera control is defined according to the target position which is achieved by the tracking algorithm in the image, calculates the rotation angles of the two motors to control the PTZ camera. As a result, the target appears in the centre of the image. This paper proposes a model-based algorithm for PTZ camera control, taking into account the camera distortion and the deviation is within 5 pixels. Experimental results show that the algorithm can be effectively applied to PTZ tracking system.Keywords-PTZ camera control; tracking algorithm; camera calibrationI.I NTR ODUC TIONPTZ(Pan, Tilt and Zoom) tracking mainly means that the PTZ cameras is under the automatic control for tracking of moving objects of visual guide to ensure tracking target always appearing in the center of the lens in the range of the camera monitoring scene. In the video surveillance, the distance of the target from the camera is generally 10 meters or more. If the object is a pedestrian, the characteristics of the pedestrian cannot be seen clearly since the image is low quality. In this case, we can manually zoom in the camera to obtain the high-resolution images of pedestrians. However, the method is only effective when the pedestrian is not moving. If the pedestrian is moving, it is easy to get out of the view of camera's field. It is very difficult to manual control the PTZ camera to make the pedestrians appear in the center of view. But PTZ tracking can make ensure that the moving target appears in the center of view. After suitable optical zoom in, the target will not be out of the scope of view and then we can obtain the high-resolution images of pedestrians.In the PTZ tracking system, the most important is the tracking algorithm and the control algorithm. Tracking algorithm detects and tracks the target in the image. The control algorithm controls the rotation angle of the motor according to the target position in the image, and then to make the target appear in the center of the image. Chang et al [1]. Suggested that the center of the image can be divided into eight directions, namely east, west, south, north, southeast, southwest, northeast and northwest. If the target is detected at the center of the southeast, then make the control to let motor go to southwest. Xiang et al [2]. proposed closed-loop control method basing on the concept of feedback.Owing to the larger deviation of the traditional algorithm, this paper directly establishes the relationship between the target position and the motor control angle for achieving a high accuracy based on pinhole camera model and camera distortion model.II.PTZ T RACKING S YSTEM O VER VIEWFIGURE I. THE STRUC TURE OF PTZ TR ACKING SYSTEM.As is shown in Figure 1, the whole PTZ tracking system consists of three parts, the image acquisition part, the tracking algorithm part and the PTZ control algorithm part. The system firstly obtains video streaming from the PTZ camera, and then uses the tracking algorithm to obtain the position of the targets. Then the PTZ camera control algorithm calculates the rotation angle and gives the orders to the PTZ camera. Drive the PTZ camera moving and make the target appear in the center of the image.Image acquisition part mainly acquires images from the video stream as an input of tracking algorithm. In order to meet the requirements of the tracking algorithm, this part also consists of image feature enhancement and pre-image processing algorithm. It is noted that the poor image quality and the image blurring problems because of the moving of camera. So during the video capture, we capture the image interval, the image is captured at a frame 1, frame 7.There are many different kinds of tracking algorithm. There is not a tracking algorithm which is popular now can perfectly solve the pose variation, illumination, occlusion and blur in tracking. And specific applications, the tracking algorithm is strict with real-time capability. This paper used compressive tracking algorithm proposed by Zhang Kaihua et al [3]. The compressive tracking algorithm runs in real-time and performs favorably against state-of-the-art algorithms on challenging sequences in terms of efficiency, accuracy and robustness. Our experiments show that the speed can reach about 50 frames per second under the 720P resolution and the algorithm can meet requirements of the PTZ tracking system. So this paper focuson PTZ control algorithm.International Conference on Computer Information Systems and Industrial Applications (CISIA 2015)III. T HEOR YPTZ control algorithm means that how to control the camera pan and tilt to make the tracking target always has been the central position of the image. There are generally two kinds of thoughts to solve the problem of control. First ，we can stresses in precise mathematical mode .In other words, we can get accurate mathematical model of system through inputs and outputs based on a certain theory. Along with the simple controller, we can achieve very good effect. Second, we can design the controller of excellent performance. However, generally it is difficult to get the precise mathematical model, so we always choose the second solution to solve the control problem, but is not better than the first solution in control precision.Specific to the PTZ control, also have the above two kinds of thought. We can use classical control algorithms such as the PID. We can detect error (the distance between detected positon and the center of the image) at every frame then put it into PID algorithm and get outputs to control camera. The PID has three parameters, proportional coefficient, the integral coefficient and the differential coefficient and this three parameters can be adjusted based on experiments. When selecting the appropriate parameters, the system will be in the steady-state after n frames within the permissible range of error. But the best control algorithm cannot reach steady state in the k+1 time based on the position of k time. If we want to do this, we must select the first method and know the exact mathematical model of the system. At the same time, the system has no integral parts and great noi se. The tracking system of this paper also happens to meet these conditions, so it can be done in one step and make target appears in the center position of the image.In the 3.1 section, the pinhole camera model and lens distortion model will be given a brief description. This part is the theoretical basis to get accurate mathematical model. In 3.2 sections, a model-based algorithm for PTZ camera control was described in detail.A. The Pinhole Camera Model and the Lens Distortion Model o-xyz is defined as the camera coordinate where o is the optical center of the camera. o-uv is defined as camera image plane coordinates in pixels units, where origin point is the upper left corner. o-xy is defined as the physical image plane coordinate in millimeters unis where origin point is defined the intersection of the optical axis and the image plane. It is also called the principal point of the image in the o-uv and coordinates is [u0, y0]T . As is show in figure 2, a point P=[X, Y, Z]T in the camera coordinate is projected to a pointP =[xc,yc,1]Tin the camera image plane coordinates. According to the pinhole camera model, the relationship of projection is0000101c xc y x f u X y f v Y Z λ⎡⎤⎡⎤⎡⎤⎢⎥⎢⎥⎢⎥=⎢⎥⎢⎥⎢⎥⎢⎥⎢⎥⎢⎥⎣⎦⎣⎦⎣⎦(1) where fx and fy denote focal length (pixels) of the camera. Since the optical system of the camera has some flaws in the process of machining and assembly, when a point is projectedonto the image plane, there is an offset between the actual pointand the ideal point. In this paper, we only consider the radial distortion and eccentric distortion [4]. The normalized coordinates of P is Pn=[X/Y, Y/Z]T =[x, y]T . Let r represents the distance from point P to the principle point. We can get offset24622123122462212312()2(2)()(2)2x y x k r k r k r p xy p r x y k r k r k r p r y p xyεε=+++++=+++++ (2)In the eqn(2), k1, k2 and k3 is the radial distortion coefficients, p1and p2 is the eccentric aberration coefficients. Let P d =[x d , y d ]T denotes the idea positon. It can be expressed asd x d y x x y y εε=+=+， (3)There are 4 intrinsic parameters and 5 distortion coefficients above the formula.we can get 9 parameters based on zhang[5].B. A Model-Based Algorithm for PTZ Camera ControlFIGURE II. THE DIAGRAM OF PTZ CONTROL ALGORITHM BASEDON THE MODEL.It is necessary to find the mathematical relationship between the control amount and the amount of error and to build mathematical model to achieve the steady state at time k based on time k-1 the position of the target. A model-based algorithm for PTZ camera control proposed in Figure 2. [x 0, y 0]T is the point that expected position of the target appearing. We can make o-xyx passes through expected point by transform. If we rotate o-xyx by specific angles around X-axes and Y-axes respectively and let the Z-axis passes through the point P, the point P can be imaged the center position of the CCD sensor. According to the basic principles of geometry, if the rotation angle follows the eqn(4), the point P can appear at the center position of the image.00()()arctan arctan c c x y x x y y p t f f ⎛⎫⎛⎫--∆=∆= ⎪ ⎪ ⎪⎝⎭⎝⎭， (4) In the eqn(4), in order to calculation of the correct point (x c, y c ), We need five distortion parameters. Among them, f x and f y are camera intrinsic parameters. We can get them from 3.1 sections. Since the PTZ camera is the zoom camera, we need to find a function of Z and focal length. We can use the least-squares fitting method to obtain the relationship between Z and the focal length20122012......n x x x x xn ny y y y yn f a a Z a Z a Z f a a Z a Z a Z=++++=++++(5)We can obtain the corresponding f x, f y and Z after camera calibration at different Z values, then use a least squares fit the function of Z and focal length. Five distortion parameters not only can be treated in a similar method but also can use piecewise linear fit method. In this paper, the focal length is fitted with six-order least-squares polynomial. Five distortion parameters are fitted with piecewise linear.From 3.1, we know that the position of the target that is detected by tracking algorithm cannot be directly substituted into the eqn(4) due to the presence of distortion. If you want to get high accuracy, the distortion correction is needed. Let [u,v]T is the position of the target in o-uv , provided by tracking algorithm. P=[X,Y,Z]T is defined as the position of target in o-xyz , the normalized coordinates is defined as P n =[X/Y,Y/Z]T =[x,y]T . We can get00(u )(v )x yu v x y f f --==， (6)If the eqn(6) is substituted into eqn(3), we can get the ideal projection point in o-xy . Let P d =[x d ,y d ]T is the ideal projection point, so the correct point in o-uv is00c d x c d y x x f u y y f v =+=+， (7)If the eqn(7) is substituted into eqn(4), the rotation angle after distortion correction can be got.IV. E XPERIMENT SAs is shown in Figure 3, the experimental platform is based on Hikvision DS-2DF5286 Camera. Pan part can rotate for degree 360.Tilt part can rotate from degree -5 to degree 90. Its control interface uses the RS485 interface and video capture uses RJ45 cable interface.FIGURE III. THE LEFT IMAGE IS THE HIKVISION DS-2DF5286 PTZCAMERA.Other two images are the screenshot of the running test application. When you click any point on the screen, the point will be able to move to the canter of the screen. The middle image is the result of clicking the keyhole of the left cabinet. The right image is the result of clicking the keyhole of the right cabinet.The entire program is implemented in the Windows operating system, using the C ++ programming language. The software environment is the Visual Studio 2013 and Hikvision development kits. Experiments on a PC equipped with Windows 8.1, where the processor is Intel Xeon E5-1603, clocks at 2.8GHz and memory is 8GB.In the figure 3, there is a dot of radius of 5 pixels and a circle of radius of 25 pixels. The two marks can be used to estimates approximately error of PTZ control. The parameters of current PTZ values are displayed on the lower half of the screen. Figure 3 selects two cases from the experimental results. the left PTZ parameters are P275 and T04 and the others are P264 and T05.TABLE I. THE RESUL TS OF PTZ CONTROL ALGORITHM TES Tshows the results. Experimental results show that the algorithm proposed can be effectively applied to PTZ tracking system and the error is within 5 pixels in most cases.FIGURE IV . THE R ESUL TS OF PTZ TRACKING TEST. THE TIMESTAMP AND PTZ PAR AMETERS CAN BE SEEN ON THE TOPAND BOTTOM OF THE APPLICA TIONAs is shown in figure 4, this paper implements a simple PTZ tracking system based on the proposed algorithm and compressive tracking algorithm. The experiment shows the final results of the human face PTZ tracking system. From the change of PTZ parameters and the positon of face, it can be seen that the calculated rotation angle according to the proposed PTZ control algorithm can make the face keep on the center of the screen.V. C ONCLUSIONThe most important is the control algorithm and the tracking algorithm in the PTZ tracking system. The tracking algorithm finds the target position in the image and the control algorithm controls the rotation angle of the PTZ camera based on the target position and finally let the target appear in the center of the image. This paper focuses on the PTZ control algorithms. We propose a model-based algorithm for PTZ camera control using a pinhole camera model and camera distortion model. The experiments have shown that the error of the control algorithm is within 5 pixels. But because the image acquisition is not continuous, resulting in the loss of the many information of the relationship between frame and frame, so the compressive tracking algorithm is not as better as continuous acquisition, and then resulting in poor PTZ tracking effect(such as fast-moving targets). Therefore, the optimal PTZ tracking system should bea staticcamera fordetecting anda PTZ camera for tracking targets that use the target positon from static camera and solve the problem perfectly.R EFERENCES[1] Chang, Faliang, et al. "PTZ camera target tracking in large complex scenes." Intelligent C ontrol and Automation (WCICA), 2010 8th W orld Congress on. IEEE, 2010.[2] Xiang, Guishan. "R eal-time follow-up tracking fast moving object with an active camera." Image and Signal Processing, 2009. CISP'09. 2nd International Congress on. IEEE, 2009.[3] Zhang Kaihua, Lei Zhang,and Ming-Hsuan Yang."Real-time compressive tracking." Compu ter V ision –ECCV 2012. Springer B erlin Heidelberg, 2012. 864-877.[4] Fryer, John G., and Duane C. Brown. "Lens distortion for close-range photogrammetry." Photogrammetric engineering and remote sensing 52.1 (1986): 51-58.[5]Zhang, Zhengyou. "A flexible new technique for camera calibration." Pattern Analysis and Machine Intelligence, IEEE Transactions on 22.11 (2000): 1330-1334.。
the review of scientific instruments
The Review of Scientific InstrumentsIntroductionThe field of scientific research heavily relies on the use of various instruments and tools to conduct experiments, collect data, and analyze results. The review of scientific instruments plays a crucial role in ensuring the accuracy, reliability, and efficiency of scientific studies. This article aims to explore the importance of instrument reviews and provide an in-depth analysis of their significance in scientific research.The Role of Instrument ReviewsInstrument reviews serve as an essential step in the scientific research process, as they enable scientists to select the most suitable tools for their experiments. By reviewing scientific instruments, researchers can assess their performance, quality, and suitability for a specific study. This evaluation process allows scientists to make informed decisions regarding instrument acquisition and usage, ensuring reliable and accurate scientific results.Factors to Consider in Instrument ReviewsWhen conducting instrument reviews, several factors should be taken into account. These factors include:1. Accuracy and PrecisionOne of the primary considerations in instrument reviews is the accuracy and precision of the measurement or analysis it provides. A reliable instrument should yield consistent and precise results, minimizingerrors and uncertainties in scientific experiments.2. Sensitivity and Detection LimitsThe sensitivity of an instrument refers to its ability to detect small changes or variations in the measured parameters. Instrument reviews should focus on the sensitivity of the tools, as this determines their capability to detect subtle changes in experimental conditions. Additionally, the detection limits of instruments should also be considered to ensure they can measure even the lowest concentrations or values accurately.3. Durability and LongevityScientific instruments can be a significant investment, makingdurability and longevity crucial factors in instrument reviews. Researchers need tools that can withstand rigorous usage and maintain their performance over an extended period. Instrument reviews should assess the robustness and reliability of instruments to ensure they can withstand the demands of scientific experiments.4. User-Friendliness and Ease of OperationInstrument reviews should also evaluate the user-friendliness and ease of operation of scientific tools. Researchers need instruments that are intuitive to use, with clear instructions and user interfaces. When instruments are user-friendly, scientists can save time and effort in conducting experiments and data analysis.Importance of Instrument Reviews in Scientific ResearchInstrument reviews play a vital role in scientific research for several reasons:1. Quality AssuranceThrough instrument reviews, researchers can ensure the quality and reliability of the data generated from scientific experiments. These reviews help identify any limitations or shortcomings of instruments and prevent inaccurate or misleading results.2. Cost-EffectivenessSelecting the most appropriate instruments for scientific studies is essential to optimize costs. Instrument reviews allow researchers to compare different options and choose tools that provide the necessary performance at the most cost-effective price.3. ReproducibilityReproducibility is an essential aspect of scientific research, ensuring that experiments and results can be independently verified and validated. By reviewing instruments, scientists can increase the reproducibility of their experiments, as other researchers can acquire the same tools and achieve comparable results.4. Innovation and AdvancementInstrument reviews contribute to the progress and innovation inscientific research. By evaluating existing instruments, scientists can identify areas for improvement and develop advanced tools with enhanced capabilities. This continuous cycle of instrument reviews and innovation drives scientific discoveries and technological advancements.ConclusionThe review of scientific instruments is a critical step in thescientific research process. It helps scientists select the mostsuitable tools, ensures the quality and reliability of data, and contributes to the advancement of scientific knowledge. By considering factors such as accuracy, sensitivity, durability, and user-friendliness, researchers can make informed decisions and conduct experiments with confidence. Additionally, instrument reviews promote cost-effective approaches, reproducibility, and innovation, ultimately leading to groundbreaking discoveries and advancements in various scientific fields.。
一种检测脑膜炎球菌多糖-白喉类毒素(CRM197)结合物中游离多糖含量的方法
一种检测脑膜炎球菌多糖-白喉类毒素（CRM197）结合物中游离多糖含量的方法朱涛1，朱婉玉2，宇学峰2，毛慧华2，邵忠琦2*（1. 天津科技大学；2. 天津康希诺生物技术有限公司，天津 300457）摘要：目的建立一种脑膜炎球菌多糖-白喉类毒素（A群、C群、Y群、W135群）结合物中游离多糖含量的检测方法。
方法根据脱氧胆酸钠（NaDC）在酸性条件下可沉淀蛋白的原理，对标准蛋白溶液进行酸沉淀处理，考察其蛋白沉淀效果；分别对脑膜炎球菌多糖-白喉类毒素（A群、C群、Y群、W135群）结合物进行脱氧胆酸钠酸沉淀，检测其游离多糖含量；向脑膜炎球菌多糖-白喉类毒素结合物中标加多糖衍生物，考察其准确度；向标准蛋白溶液中标加不同浓度的脱氧胆酸钠溶液，验证脱氧胆酸钠对于蛋白检测的干扰，考察其耐用性。
结果不同浓度的标准蛋白溶液经脱氧胆酸钠酸沉淀法处理后，蛋白浓度在100~300μg/ml范围内，沉淀中蛋白回收率为90%~110%，蛋白浓度大于300 μg/ml的样品时，回收率显著降低；分别对脑膜炎球菌多糖-白喉类毒素结合物中标加多糖衍生物，上清中多糖回收率为80%~110%。
重复6次对脑膜炎球菌多糖-白喉类毒素结合物进行游离多糖含量检测，其相对标准偏差（RSD）均低于15%；当溶液总中NaDC的浓度不高于2%时，对蛋白检测无干扰。
结论 NaDC沉淀法能专一的沉淀蛋白，不沉淀多糖，该方法有较高的准确度和精密度，重复性好，可用于脑膜炎球菌多糖-白喉类毒素结合物中游离多糖含量的测定。
关键词：脑膜炎球菌多糖；白喉类毒素；游离多糖；脑膜炎球菌多糖-CRM197 结合物；脱氧胆酸钠A Method for Measuring free Polysaccharide in Meningococcal Polysaccharide - Diphtheria Toxoid Conjugate PreparationsZHU Tao1，ZHU Wan-yu 2，YU Xue-feng 2，MAO Hui-hua2，SHAO Zhong-qi 2*（1. Tianjin University of Science & Technology；2. Tianjin CanSino Biotechnology Inc，Tianjin 300457）ABSTRACT：OBJECTIVE To develop a method for quantitative determination of free polysaccharide in the preparations of meningococcal polysaccharide - diphtheria toxoid (CRM197) conjugates. METHODS A series dilution of a standard protein solution were prepared and treated with sodium deoxycholate (NaDC) in an acidic condition to determine the concentration range in which proteins can be precipitated efficiently；Free polysaccharide in four meningococcal1polysaccharide – CRM197（type A、C、Y、W135）conjugate preparations were measured by separating the conjugates and free polysaccharide through NaDC precipitation,. The accuracy, repeatability and robustness of the NaDC precipitation method were investigated in the study. RESULTS The protein recovery was over90% by NaDC precipitation when the protein concentration was in the range of 100~300 μg/ml. The precipitation efficiency would be remarkably reduced when the protein concentration was higher than 300 μg/ml. The recovery of free polysaccharide in the meningococcal polysaccharide – CRM197 conjugates, determined using the spiked samples, was 80%~110%. The relative standard deviation (RSD) of the method was <15%. NaDC did not interfere with the protein assay at the concentration lower than 2%. CONCLUSION The NaDC precipitation method is accurate, fast and highly reproducible, and can be used for quantitative determination of free polysaccharide in meningococcal polysaccharide – CRM197 conjugate preparations.KEY WORDS: Meningoccocal polysaccharides; Diphtheria toxoid (CRM197); Free polysaccharides ; Polysaccharide –CRM197 conjugates; Sodium deoxycholate流行性脑脊髓膜炎（简称流脑）是危害儿童健康的主要传染病之一，具有非常高的病死率和致残率。
英语作文-集成电路设计行业中的人工智能与机器学习应用
英语作文-集成电路设计行业中的人工智能与机器学习应用In the realm of integrated circuit (IC) design, the application of artificial intelligence (AI) and machine learning (ML) has emerged as a transformative force, revolutionizing the way circuits are conceived, optimized, and validated. This synergy between AI/ML techniques and IC design has not only accelerated the development process but has also significantly enhanced the performance, efficiency, and reliability of modern semiconductor devices.AI and ML technologies are particularly advantageous in IC design due to their ability to process vast amounts of data and derive complex insights that are difficult to ascertain through traditional methods. One of the primary areas where AI excels is in the optimization of circuit layouts and architectures. Designers can leverage AI algorithms to explore a myriad of design possibilities, considering parameters that range from performance metrics to power consumption and manufacturing costs.Moreover, AI enables predictive modeling with a level of accuracy that was previously unattainable. By analyzing historical design data and outcomes, machine learning algorithms can predict potential issues early in the design phase, thereby minimizing costly redesigns and ensuring faster time-to-market for new IC products. This predictive capability is crucial in an industry where even minor design flaws can lead to significant setbacks in product development cycles.Another compelling application of AI in IC design is in the realm of automated design synthesis. Traditionally, creating a new IC design involved a labor-intensive process of manually crafting and refining circuit layouts. However, with AI-driven synthesis tools, designers can input high-level design goals and constraints, allowing the algorithms to autonomously generate and optimize circuit architectures that meet specified criteria. This not only reduces the burden on human designers but also opens uppossibilities for exploring innovative design concepts that might have been overlooked using conventional methods.Furthermore, AI plays a pivotal role in enhancing the robustness and reliability of IC designs. Machine learning algorithms can analyze real-time operating conditions and performance data, continuously optimizing parameters to adapt to varying environmental factors or workload demands. This adaptive capability is particularly advantageous in applications such as automotive electronics, where ICs must operate reliably under diverse and often harsh conditions.In addition to design optimization and predictive modeling, AI is also instrumental in the verification and testing phases of IC development. Verification of complex IC designs traditionally required exhaustive simulations and testing scenarios, which are time-consuming and resource-intensive. AI-powered verification tools can automate this process by intelligently generating test cases, detecting potential errors, and even suggesting design improvements based on simulation results.Moreover, the integration of AI in IC design extends beyond the development phase into manufacturing and quality control. AI algorithms can analyze manufacturing data to detect anomalies or deviations from optimal production parameters, thereby improving yield rates and reducing defect rates in semiconductor fabrication.In conclusion, the application of artificial intelligence and machine learning in the integrated circuit design industry represents a paradigm shift, enhancing every stage of the IC design lifecycle from conception to production. By leveraging AI's computational prowess and predictive capabilities, designers can innovate faster, achieve higher performance metrics, and deliver more reliable semiconductor products to meet the demands of today's technology-driven world. As AI continues to evolve, its role in IC design will likely expand, ushering in a new era of efficiency, innovation, and reliability in semiconductor technology.。
科技英语的翻译——02 专业词汇的构成
专业词汇的构成1.词汇的分类专业英语文献中的专业词汇（或科技词汇）一般分为三类：技术词、半技术词和非技术词（或普通词）。
技术词（technical word）指科技术语，用以记录和表达科学技术专门领域的现象、过程、特性、关系、状态、数量等，意义单纯，只有一种专业含义，如bandwidth, flip-flop, microprocessor, rectification。
半技术词（semi-technical words）是指跨学科出现的频率很高的词，在不同的专业领域含有不同的精确含义，如power, carrier等。
非技术词是指在非专业英语中很少使用，却严格属于非专业英语性质的词汇，如application, implementation, to yield 等。
非技术词和普通词（尤其是其中的功能词，如限定词、介词、连词等）词频最高，半技术词出现得最多，词汇覆盖面最大，而技术词词频最低，出现率最小。
２．词汇的构成科技英语词汇中有很大一部分符合构词法。
科技英语构词法主要包括：派生、复合、转化、拼缀、缩略等，下面分别予以介绍。
(1) 派生（derivation）法是通过在原有词或词根的基础上加前缀或（和）后缀而构成新词。
前缀通常用以修饰或改变词意，后缀显示词性。
科技英语中以这一方法构成的新词最多，可以说俯拾皆是，如antiparticle （反粒子），antineutron （反中子），antibody （抗体）；autocorrelation （自相关）等。
需要指出的是以V+er/or 构成的词，有许多是指某一仪器而不指人，如semicorrelator （自相关器），conductor （导体），holder （支架/托）等，这一点译者应留意。
表2.1-2.4分别列出电子与通信专业常用的前后缀及词根。
表2.1 常用前缀表2.2 表示数量关系的常用前缀表2.3 常用后缀表 2.4 常用词根(2) 复合(composition)法是由两个或两个以上的词按照一定的次序排列，以构成新词，其词性可以是形容词、名词、代词、动词、副词等。
Robust Control and Estimation
Robust Control and Estimation Robust control and estimation are vital components in the field of engineering, particularly in the areas of aerospace, automotive, and industrial control systems. These techniques are used to ensure that a system can perform reliably in the presence of uncertainties and disturbances. In this response, we will explore the importance of robust control and estimation, the challenges associated with implementing these techniques, and the potential impact on various industries. One of the key reasons why robust control and estimation are essential is the presence of uncertainties in real-world systems. No system operates in a perfect environment, and there are always factors that can introduce variations and disturbances. Robust control techniques allow engineers to design systems that can accommodate these uncertainties, ensuring that the system remains stable and performs as expected. Similarly, robust estimation techniques enable engineers to accurately estimate the state of a system, even in the presence of noise and disturbances. In the aerospace industry, for example, robust control and estimation are critical for ensuring the stability and performance of aircraft and spacecraft. These systems operate in highly dynamic and uncertain environments, where external factors such as turbulence and sensor noise can significantlyimpact the behavior of the vehicle. By employing robust control and estimation techniques, aerospace engineers can design vehicles that can withstand these uncertainties and safely reach their destinations. In the automotive industry, robust control and estimation play a crucial role in the development of autonomous vehicles. These vehicles rely on a multitude of sensors and actuators to perceive the environment and make decisions in real-time. However, these sensors can be affected by various sources of uncertainty, such as weather conditions and sensor failures. Robust control and estimation techniques are essential for ensuring the safety and reliability of autonomous vehicles, allowing them to operateeffectively in diverse and unpredictable scenarios. In the realm of industrial control systems, robust control and estimation are instrumental in optimizing the performance of complex manufacturing processes. These processes often involve multiple interconnected systems that are subject to disturbances and uncertainties. By implementing robust control and estimation techniques, engineers can improvethe stability, efficiency, and robustness of these systems, ultimately leading to higher productivity and lower operational costs. Despite their significance, implementing robust control and estimation techniques comes with its own set of challenges. Designing robust controllers and estimators requires a deep understanding of system dynamics and uncertainties, as well as advanced mathematical tools for analysis and synthesis. Moreover, the performance of these techniques heavily relies on the accuracy of the system models and the quality of the available data. This means that engineers must invest significant time and resources into system identification and validation to ensure the effectiveness of robust control and estimation techniques. Furthermore, the complexity of real-world systems often makes it difficult to develop robust controllers and estimators that can guarantee performance under all possible scenarios. Engineers must carefully balance the trade-offs between robustness and performance, as overly conservative designs can lead to suboptimal system behavior. Additionally, the computational requirements of robust control and estimation algorithms can be substantial, especially for real-time applications, necessitating efficient implementation and hardware considerations. In conclusion, robust control and estimation are indispensable tools for ensuring the reliability and performance of engineering systems in the face of uncertainties and disturbances. From aerospace to automotive to industrial applications, these techniques enable engineers to design and operate systems that can withstand the challenges of the real world. However, the implementation of robust control and estimation techniques is not without its complexities, requiring a deep understanding of system dynamics, advanced mathematical tools, and careful consideration of trade-offs. As technology continues to advance, the importance of robust control and estimation will only grow, shaping the future of engineering and technology.。
最终最大重投影误差 英文
最终最大重投影误差英文The Importance of Maximizing the Maximum Reprojection Error in Camera CalibrationCamera calibration is an important aspect in computer vision and robotics. It refers to the process of estimating the intrinsic and extrinsic parameters of a camera, which are essential in transforming 2D images into 3D objects. One of the most widely used methods in camera calibration is the Maximum Reprojection Error (MRE) optimization, which aims to minimize the difference between the observed and predicted image points.However, despite its popularity, there has been a longstanding debate on the effectiveness of MRE optimization. Some argue that minimizing the MRE is not enough to ensure robustness and accuracy in camera calibration. They highlight the potential risks of underfitting the calibration model, which could lead to incorrect parameter estimates and poor performance in subsequent applications.To address these concerns, recent research has proposed a new approach to MRE optimization, which instead aims to maximize themaximum reprojection error. This involves finding the optimal parameter values that result in the largest possible MRE, subject to a certain level of constraint. The rationale behind this approach is that by emphasizing the effect of outliers, it can improve the robustness and accuracy of the calibration model, especially in cases where the data is noisy or corrupted.For instance, in a scenario where the camera captures images under varying lighting conditions, there may be significant variations in the observed image points, which could cause the MRE optimization to fail. However, by maximizing the maximum reprojection error, the calibration model can be more tolerant to nonlinearities and deviations, which can improve its overall performance.In conclusion, maximizing the maximum reprojection error is an important consideration in camera calibration. It offers a promising solution to the challenges associated with MRE optimization, by emphasizing the importance of outliers and nonlinearities in the data. As computer vision and robotics continue to evolve, it is likely that this approach will become increasingly relevant in ensuring robust and accurate camera calibration.。
Accuracy, robustness, and efficiency comparison in iterative computation of convection diff
Lixin Gey and Jun Zhangz Department of Computer Science, University of Kentucky, 773 Anderson Hall, Lexington, KY 40506-0046, USA August 9, 1999
convection di usion equation, boundary layer, grid stretching, multilevel preconditioner, multigrid method
Key words:
1 Introduction
We consider the two dimensional convection di usion equation with the Dirichlet boundary conditions written in the form uxx + uyy + p(x; y)ux + q(x; y)uy = f (x; y); (x; y) 2 ; (1) u(x; y) = g(x; y); (x; y) 2 @ : The convection coe cients p(x; y) and q(x; y) are functions of the independent variables x and y, and are assumed to be su ciently smooth. Here is a convex domain consisting of a union of rectangles, and @ is the boundary of . For convenience, we refer to the magnitude of p(x; y)
- 1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。
ABSTRACT
This paper describes Picky, a probabilistic agenda-based c h a r t p a r s i n g a l g o r i t h m w h i c h u s e s a t e c h n i q u e called p~'ob-
1. I n t r o d u c t i o n
This paper addresses the question: Why should we use probabilistic models in natural language understanding? There are many answers to this question, only a few of which are regularly addressed in the literature. The first and most common answer concerns ambigu~ ity resolution. A probabilistic model provides a clearly defined preference nile for selecting among grammatical alternatives (i.e. the highest probability interpretation is selected). However, this use of probabilistic models assumes that we already have efficient methods for generating the alternatives in the first place. While we have O(n 3) algorithms for determining the grammaticality of a sentence, parsing, as a component of a natural language understanding tool, involves more than simply determining all of the grammatical interpretations of an input. Ill order for a natural language system to process input efficiently and robustly, it must process all intelligible sentences, grammatical or not, while not significantly reducing the system's efficiency. This observ~ttiou suggests two other answers to the central question of this paper. Probabilistic models offer a convenient scoring method for partial interpretations in a well-formed substring table. High probability constituents in the parser's chart call be used to interpret ungrammat.ical sentences. Probabilistic models can also
be used for efficiency by providing a best-first search heuristic to order the parsing agenda. This paper proposes an agenda-based probabilistic chart parsing algorithm which is both robust and efficient. The algorithm, 7)icky 1, is considered robust because it will potentially generate all constituents produced by a pure bottom-up parser and rank these constituents by likelihood. The efficiency of the algorithm is achieved through a technique called probabilistic prediction, which helps the algorithm avoid worst-case behavior. Probabilistic prediction is a trainable technique for modelling where edges are likely to occur in the chart-parsing process. 2 Once the predicted edges are added to the chart using probabilistic prediction, they are processed in a style similar to agenda-based chart parsing algorithms. By limiting the edges in the chart to those which are predicted by this model, the parser can process a sentence while generating only the most likely constituents given the input. In this paper, we will present the "Picky parsing algorithm, describing both the original features of the parser and those adapted from previous work. Then, we will compare the implementation of `picky with existing probabilistic and non-probabilistic parsers. Finally, we will report the results of experiments exploring how `picky's algorithm copes with the tradeoffs of efficiency, robustness, and accuracy. 3
2. P r o b a b i l i s t i c Models in "Picky The probabilistic models used ill the implementation of "Picky are independent of the algorithm. To facilita.te the comparison between the performance of "Picky and its predecessor, "Pearl, the probabilistic model ilnplelnented for "Picky is similar to "Pearl's scoring nlodel, the contextl'pearl =-- probabilistic Earley-style p a r s e r ( ~ - E a r l ) . "Picky =probabilistic CI(Y-like parser ('P-CKY). 2Some familiarity with c h a r t p a r s i n g t e r m i n o l o g y is a s s u m e d in this paper. For terminological definitions, see [9], [t0l, [11], or [17]. 3Sections 2 a n d 3, t h e descriptions of t h e probabilistie m o d e l s used in ",Picky and t h e T'icky algorithn,, are s i m i l a r in c o n t e n t to t h e c o r r e s p o n d i n g sections of M a g e r n m n a n d Weir[13]. T h e e x p e r i m e n t a l r e s u l t s a n d d i s c u s s i o n s which follow in sections .1-6 ~tre original.
Efficiency, Robustness and Accuracy in Picky Chart Parsing*
David M. Magerman
Stanford University S t a n f o r d , C A 94305 magerman@cs.st
Carl Weir
abilistic prediction to predict which grammar rules are likely
to lead to an acceptable parse of the input. Using a suboptimal search method, "Picky significantly reduces the number of edges produced by CKY-like chart parsing algorithms, while maintaining the robustness of pure bottom-up parsers and the accuracy of existing probabilistic parsers. Experiments using Picky demonstrate how probabilistic modelling can impact upon the efficiency, robustness and accuracy of a parser.