复杂线素图像多因素预测矢量化跟踪

针对具有稠密坐标网格和交叉线素的工程曲线图的矢量化,提出了坐标网格的剔除方法、多义线模拟样条曲线算法和缺口曲线修补策略,以及基于多因素预测跟踪算法.坐标网格剔除方法采用基于线检测算法来剔除坐标网格,再依据过滤后的图像灰度自动地修复被连带除去的趋于水平和垂直的曲线像素;多义线模拟样条曲线算法则是依据已知曲线节点生成若干线段来逼近曲线且它具有与样条曲线一致的局部几何特性;而多因素预测跟踪算法则是利用前面数个边界点的灰度、当前边界点切矢和它的24邻域内非背景像素之间的相对位置来预测曲线下一个边界点,监控曲线曲率角的变化,以避免"跑偏"和按层搜寻同曲线像素来跨越曲线断点.通过与R2V5.5和ScanIn5.1的矢量化对比,证明它具有更显著的智能化和自动化的优点.

Multi-factor forecasting vectorization tracking for a complex line-element image

To present an method of Elimination Coordinate Grid(ECG),an algorithm of Polyline Simulation Spline(PSS) curve,a repaired method for the curve gaps and a algorithm of Multi-factor Forecasting Tracking(MFT) for the vectorization of an engineering curve image with dense coordinate grid and cross line-elements were proposed.ECG used line detecting method to eliminate the coordinate grid and automatically restore the near level and vertical curve pixels based on the gray scale in the filtered image.PSS formed a number of the line-segments to approach the curve by the given curve nodes,whose local geometrical feature was the same as spline curve.MFT forecasted the next boundary point of the curve based on the gray-scale of the front boundary points,the tangential vector of the current boundary point and the relative position among non-background pixels in 24 neighborhoods.The change of the curve curvature angle was monitored to avoid "off-tracking" and the curve break-points were stridden by search the same curve pixels layer by layer.Compared to R2V5.5 and ScanIn5.1,it was the more intelligent and automatic in vectorization.