计算机辅助设计与图形学学报
計算機輔助設計與圖形學學報
계산궤보조설계여도형학학보
JOURNAL OF COMPUTER-AIDED DESIGN & COMPUTER GRAPHICS
2010年
1期
66-75
,共10页
苏智勇%唐卫清%李蔚清%何涛
囌智勇%唐衛清%李蔚清%何濤
소지용%당위청%리위청%하도
基于外存的绘制%预取%基于网络的绘制%分层层次细节模型
基于外存的繪製%預取%基于網絡的繪製%分層層次細節模型
기우외존적회제%예취%기우망락적회제%분층층차세절모형
out-of-core rendering%pre-fetching%network-based rendering%hierarchical levels of detail
现有的基于网络的远程绘制系统在绘制过程中对网络带宽和时延具有较强的依赖性,为了获得较高的绘制速度,需要耗费大量的预处理时间和存储空间.针对网络环境下模型的设计校审工作对预处理时间、绘制速度和图像质量的实际需求,提出一种基于外存的大规模流程工厂模型交互绘制算法.绘制前,首先从服务器端获取模型的几何参数和拓扑信息;然后根据流程工厂模型特征,在客户端以设备和管线为基本单位组织外存数据,采用体元合并的方法快速完成模型层次细节的计算和存储.分析了校审内容和校审人员的运动习惯,并将其与基于视点可见性的预取算法相结合,在本地实现外存数据的高效预取,且绘制过程中无需传输模型面片信息.实验结果表明,文中方法在普通PC机上能够将具有21M左右面片模型的预处理时间控制在5min以内,在保证校审所需图像质量的前提下取得平均30帧/s的平稳帧速,且绘制过程不依赖网络带宽和时延.
現有的基于網絡的遠程繪製繫統在繪製過程中對網絡帶寬和時延具有較彊的依賴性,為瞭穫得較高的繪製速度,需要耗費大量的預處理時間和存儲空間.針對網絡環境下模型的設計校審工作對預處理時間、繪製速度和圖像質量的實際需求,提齣一種基于外存的大規模流程工廠模型交互繪製算法.繪製前,首先從服務器耑穫取模型的幾何參數和拓撲信息;然後根據流程工廠模型特徵,在客戶耑以設備和管線為基本單位組織外存數據,採用體元閤併的方法快速完成模型層次細節的計算和存儲.分析瞭校審內容和校審人員的運動習慣,併將其與基于視點可見性的預取算法相結閤,在本地實現外存數據的高效預取,且繪製過程中無需傳輸模型麵片信息.實驗結果錶明,文中方法在普通PC機上能夠將具有21M左右麵片模型的預處理時間控製在5min以內,在保證校審所需圖像質量的前提下取得平均30幀/s的平穩幀速,且繪製過程不依賴網絡帶寬和時延.
현유적기우망락적원정회제계통재회제과정중대망락대관화시연구유교강적의뢰성,위료획득교고적회제속도,수요모비대량적예처리시간화존저공간.침대망락배경하모형적설계교심공작대예처리시간、회제속도화도상질량적실제수구,제출일충기우외존적대규모류정공엄모형교호회제산법.회제전,수선종복무기단획취모형적궤하삼수화탁복신식;연후근거류정공엄모형특정,재객호단이설비화관선위기본단위조직외존수거,채용체원합병적방법쾌속완성모형층차세절적계산화존저.분석료교심내용화교심인원적운동습관,병장기여기우시점가견성적예취산법상결합,재본지실현외존수거적고효예취,차회제과정중무수전수모형면편신식.실험결과표명,문중방법재보통PC궤상능구장구유21M좌우면편모형적예처리시간공제재5min이내,재보증교심소수도상질량적전제하취득평균30정/s적평은정속,차회제과정불의뢰망락대관화시연.
The performance of current network-based remote rendering systems relies on the bandwidth and latency of the network during visualization. In order to gain high frame rates, they need time consuming preprocessing and additional storage. Design review has practical requirements on the preprocessing time, frame rates and visual fidelity. In this paper, a network-based interactive out-of-core rendering system for large-scale process plant models was proposed. First, the model's geometry parameters and topology information are transmitted from the server before visualization. Then the model's out-of-core representation is organized and computed locally according to the model feature. By analyzing the content and motion characteristic of the reviewer, a novel visibility-based pre-fetching technique is presented. It takes into account the reviewer's motion characteristic. No triangles are transmitted during visualization. The algorithm can be implemented on a general PC over the network. It can preprocess a 21-million-triangles model within 5 minutes and achieve 30 fps smooth frame rates on average without sacrificing the image quality. Experimental results show that the rendering performance doesn't depend on the network's bandwidth and latency.