一种基于先验知识的矢量边界更新方法

本文针对GIS中地物矢量边界更新繁琐、低效的问题, 提出一种以GIS数据作为先验知识的简便矢量边界更新方法.该方法针对GIS线性矢量边界只需有限个拐点描述的特性, 将对边界寻找转化为直接对矢量边界上拐点位置的确定, 简化了变化检测与更新过程.以TM遥感影像为数据源,矢量shape文件为先验知识, 分别利用该方法检测、更新扎龙湿地内克钦湖、东升水库以及一块沼泽地边界.并运用简化的缓冲区检测算法计算拐点更新准确率, 同时对本文结果与手动更新结果的相似性进行计算与分析.实验结果表明, 本方法不但简化了变化检测和更新过程, 而且可以得到接近手动更新方法的准确率, 并且更新后矢量多边形与手动更新所得到的矢量多边形保持了较高相似性, 达到了较好的效果.

A Method Based on Prior Knowledge for Vector Edge Update

Data update is an mi portant task forGIS. However, complex and inefficient process has been a principalproblem thatgets in theway ofupdatingGIS data in tmi e during vectordata update inGIS, it isnecessary to substitute anautomaticmethod formanualmethod currently.In order tomake data update lessdependenton arbitrary judgement, auxiliary information should be cited. During thechange detection of remote sensing mi ages, GIS data have played an mi portant role as prior knowledge in increasingaccuracy and smi plifying process. Approaches of change detection and data update has been proposed in recent years.However, those approaches are based on neuralnetwork, knowledge base instead on linearvectordata. Linearvectors thatdescribe the object region edge in GIS, is usually irregular polygon which is represented as connection of the discretepoints on the edge. Hence, itmakes sense that themethod depending on the attributes of linear vectors above is figuredout to update the coordinate values ofpoints on the boundary.A smi ple and convenientmethod, which utilizes data in GIS as prior knowledge, is proposed here for vector edgeupdate to solve the problem of complexity and inefficiency. Thismethod involves three-step processes of searching edge:generally locating points, recognizing the convexity and concavity, and detailedly describing the polygonal shape. Thepointsmoveswithin the lmi ited regions thatthreshold valuesdetermines. Thismethod integrates remote sensing mi ages andpriorGIS data, aswell as replace searching edge curve with locating inflexions to smi plify change detection and updateprocess. In addition, themethod points out an idea of integration of remote sensing and geographical information systemdatawith the carrier of linear vectors.W ith the expermi ental data ofLandsatTM remote sensing mi ages and the prior knowledge of vector data saved asshapefile ofArcGIS(software, themethod is applied to detectand update the edgesofKeqinLake, DongshengReservoirand an area ofmarsh in ZhalongWetland inHeilongjiangProvince, China. And then update accuracies are tested by thesmi plified buffer detection algorithm for computing polygonal error based on original buffer detection algorithm, and thepolygonal comparability between this presentedmethod and themanualmethod is calculated and analyzed. In terms of thecharacteristic of polygon, five factors are adopted in evaluating the polygonal comparability: major direction angle andminor direction angle ofminmi um exterior rectangle, length and width ofminmi um exterior rectangle, and polygonalpermi eter. Relative error of these five factors are computed, based on which mean value is considered as evaluation ofpolygonal comparable percentage. Besides, different from originalbufferdetection algorithm, a smi plified bufferdetectionalgorithm ispointed out. Based on the mi provement, bufferdetection algorithm can be calculated through infinite numbersofpoints on the polygon that stand for object region edge.Estmi ations are obtained according to those two criteria above. The results show that the update accuracy ofproposedmethod reach above 80%, a little lower than the resultderived frommanualmethod, though good polygonal comparability very close to the result ofmanualmethod. In spite of that, the approach presented performs an automatic procedure ofvector edge update in GIS. Apparently, this method well smi plifies change detection and update process, as well asachieves an update accuracy close to themanualmethod, and a relatively satisfactory comparabilitywith themanualupdatemethod. Besides, the results mi plies thatdifferentexpermi entalobjects lead to distinctevaluated results. The smi pler theshape of edge is, the better the method is evaluated. Thus, the result from expermi ent on Keqin Lake is better thanDongshengReservoir andmarsh, as the contour ofKeqin Lake is smi pler than thatof the other two.Nevertheless, there are still several disadvantages to be mi proved. Thismethod loses control of contour shape ofobject so that the movement can not be restricted to be harmonious between adjacent points. Also, the moving rangethreshold ofpoints are supposed to be settled by researcher, whichmeans that the automatic process is affected by humanidea.