面向样本扩充的新型风格迁移网络研究

全监督语义分割网络在训练时需要耗费大量的人力与时间成本来标注样本。所以减少人工标注样本的时间,同时提升语义分割效果,对于深度学习网络的快速部署和应用推广具有重要意义。提出一种基于改进图像风格迁移网络（CycleGAN-AD）的样本扩充方法。以CycleGAN为基础,在生成器中引入注意力机制并将深度残差网络改为密集连接卷积网络。利用计算机批量产生自带标签的模拟样本,使用CycleGAN-AD网络将模拟样本风格迁移成为真实样本风格（标签不变）,并用于扩充训练样本。对石墨电极的钢印字符进行语义分割的实验结果表明,采用CycleGAN-AD网络进行样本扩充后,其分割效果得到显著提升,MIoU值最高升至0.826 0。可见,提出的样本扩充方法有希望在显著减少人工标注工作量的同时,获得高质量的训练样本。     

Automatic integration of Web search interfaces with WISE-Integrator

An increasing number of databases are becoming Web accessible through form-based search interfaces, and many of these sources are database-driven e-commerce sites. It is a daunting task for users to access numerous Web sites individually to get the desired information. Hence, providing a unified access to multiple e-commerce search engines selling similar products is of great importance in allowing users to search and compare products from multiple sites with ease. One key task for providing such a capability is to integrate the Web search interfaces of these e-commerce search engines so that user queries can be submitted against the integrated interface. Currently, integrating such search interfaces is carried out either manually or semiautomatically, which is inefficient and difficult to maintain. In this paper, we present WISE-Integrator - a tool that performs automatic integration of Web Interfaces of Search Engines. WISE-Integrator explores a rich set of special metainformation that exists in Web search interfaces and uses the information to identify matching attributes from different search interfaces for integration. It also resolves domain differences of matching attributes. In this paper, we also discuss how to automatically extract information from search interfaces that is needed by WISE-Integrator to perform automatic interface integration. Our experimental results, based on 143 real-world search interfaces in four different domains, indicate that WISE-Integrator can achieve high attribute matching accuracy and can produce high-quality integrated search interfaces without human interactions.Received: 2 January 2004, Accepted: 25 March 2004, Published online: 12 August 2004Edited by: M. Carey     

Towards Deeper Understanding of the Search Interfaces of the Deep Web

Many databases have become Web-accessible through form-based search interfaces (i.e., HTML forms) that allow users to specify complex and precise queries to access the underlying databases. In general, such a Web search interface can be considered as containing an interface schema with multiple attributes and rich semantic/meta-information; however, the schema is not formally defined in HTML. Many Web applications, such as Web database integration and deep Web crawling, require the construction of the schemas. In this paper, we first propose a schema model for representing complex search interfaces, and then present a layout-expression based approach to automatically extract the logical attributes from search interfaces. We also rephrase the identification of different types of semantic information as a classification problem, and design several Bayesian classifiers to help derive semantic information from extracted attributes. A system, WISE-iExtractor, has been implemented to automatically construct the schema from any Web search interfaces. Our experimental results on real search interfaces indicate that this system is highly effective.     

基于FLUENT的后混合磨料水射流喷嘴内流场的数值模拟

以后混合磨料水射流喷嘴为研究对象,基于多相流动的欧拉模型,应用FLUENT流体分析软件对几种典型后混合磨料喷嘴的内流场进行数值模拟和仿真,得到压力分布、速度分布、湍流能量分布.结果表明:喷嘴内由于涡旋的形成损耗了部分能量,使得射流出口速度和动能明显降低;存在一个最优圆柱长度使得磨料出口速度达到最大值,其长度为喷嘴出口直径的23～37倍时,磨料得到最高的出口速度;喷嘴内流体紊动能的分布主要集中在喷嘴近壁区和几何形状变化较大的区域.     

轿车轮胎新型2-1冠带条缠绕方式

为了解决EXTRA FRD88花纹轿车轮胎胎里不平的问题,对冠带条缠绕方式进行优化。新型2-1冠带条缠绕方式在原2-1冠带条缠绕方式的基础上,在胎面中心区域左右两侧各多缠绕一圈冠带条,胎面中心金条区域材料厚度加大,使该区域缺少的材料得到补充,消除因胶料流动而导致的胎里不平问题,提高轮胎外观通过率。新型2-1冠带条缠绕方式实际使用效果较好,且操作方便、简捷。     

A methodology to retrieve text documents from multiple databases

This paper presents a methodology for finding the n most similar documents across multiple text databases for any given query and for any positive integer n. This methodology consists of two steps. First, the contents of databases are indicated approximately by database representatives. Databases are ranked using their representatives with respect to the given query. We provide a necessary and sufficient condition to rank the databases optimally. In order to satisfy this condition, we provide three estimation methods. One estimation method is intended for short queries; the other two are for all queries. Second, we provide an algorithm, OptDocRetrv, to retrieve documents from the databases according to their rank and in a particular way. We show that if the databases containing the n most similar documents for a given query are ranked ahead of other databases, our methodology will guarantee the retrieval of the n most similar documents for the query. When the number of databases is large, we propose to organize database representatives into a hierarchy and employ a best-search algorithm to search the hierarchy. It is shown that the effectiveness of the best-search algorithm is the same as that of evaluating the user query against all database representatives.     

基于机械臂的空间水射流切割系统的设计与仿真

笔者以水射流切割机床机械臂为研究对象,基于SolidWorks软件建立了四轴空间水射流机床机械臂的模型,应用ADAMS软件的运动仿真功能,对4自由度机械臂的运动特性进行了仿真分析。仿真结果表明,应用四轴机械臂的空间水射流切割系统可完成空间复杂曲线运动,从而实现三维加工,如圆锥体、圆锥齿轮及螺纹等。通过增加摆动臂来解决水射流切割过程中的锥度补偿,提高了切割几何精度。     

微磨料水射流对多晶硅切割的试验研究与模型分析

笔者根据材料冲蚀磨损原理,分析探讨了微磨料水射流切割多晶硅硅锭的机理。并运用后混合微磨料射流切割装置,对多晶硅硅锭进行了单因素切割试验,从而获得了切割性能曲线。研究结果表明,微磨料水射流对多晶硅的切割完全可行,射流压力和横移速度是重要的试验数据。通过对高压磨料水射流切割性能的预测和加工参数的优化,最终建立切割表面质量经验模型,根据模型分析得出射流压力、横移速度与切割质量之间的关系。     

ACPR50S试验堆压力容器椭圆封头数控加工工艺研究

通过对椭圆封头的几何特征和数控加工工艺进行分析,得出其制造难点在于控制内、外曲面等壁厚达到要求,通过求解内、外曲面轮廓的参数方程,并对不同数控编程方法进行分析,得到该椭圆封头的数控加工及数控编程方案,解决了该椭圆封头的数控加工问题.     

冰粒水射流表面抛光实验研究

对冰粒水射流抛光进行实验研究。在选取合适冰粒直径和其他参数的条件下,根据后混合磨料水射流原理,用冰粒水射流对硬脆性材料的表面进行抛光。选用的抛光试件为未经过热处理的45钢,抛光喷嘴内径为3.175 mm,冰粒直径为0.5～2 mm,冰粒贮存温度为-20～-50℃。设计正交试验,分析冰粒水射流抛光过程中,主要加工工艺参数对表面粗糙度的影响。结果表明:冰粒水射流完全可以达到抛光功用;靶距是影响抛光效果的主要因素。