数字图像处理法确定林带疏透度随机误差研究

Study on random error of shelterbelt porosity estimated by measuring photo with the help of digitized photographic silhouettes.

在分析以数字图像处理为测定法确定林带疏透度误差来源的基础上,对其中的随机误差进行研究,结果表明:林带整体疏透度的机误方差小于其冠部和干部的较大者;所研究的各类型林带各部位疏透度随机误差均遵从正态分布;林带整体疏透度随机误差的分布与树种和带内配置无关;北京杨、双阳快杨和其它类杂交杨的矩形或品字形配置林带各自冠部与干部疏透度机误方差之间无显著差异,而乡土杨林带干部的显著大于冠部的。本文还分别各类型林带的各部位确定了由林带疏透度测定值估计其总体实际值的随机误差限,并讨论了在测定林带疏透度过程中据该误差限计算样本量和划定测定范围的应用意义。最后总结提出:以增加测定同一林带不同样段像片数限定随机误差,通过模型计算订正疏透度测定值中的投影误差和影缩误差确定林带疏透度是建立完善的“数字图像处理法确定林带疏透度”新方法的可行途径。

Error resources of shelterbelt porosity estimated by measuring shelterbelt photo with digitized photographic silhouettes were analyzed, of which, the random error is studied. Results show that the bigger random error variance of crown and trunk porosities is the upper limit of whole variance, that the random errors of crown, trunk, and whole porosities in each typical shelterbelts studied distribute in accordance with normal distribution, that the random error variance of whole porosity is irrelevant to tree species and patterns of the shelterbelts, and that the random error variances between crown and trunk porosities of the shelterbelts, which are composed of Populus×beijingensis or P.×xiaozhuanica or other hybrid popular arranged by rectangle or triangle, have no significant difference, however, for local popular, the random error variance of trunk porosity is signifieantly bigger than that of crown. The random error limits of generalactual porosity estimated from average value of measuring photo with digitized photographic silhouettes were defined, and the idea of applying the limits for calculating the number of observation and delimiting range to be measured were also discussed. Finally, available way for establishing the perfect method such as 'estimation of shelterbelt porosity by measuring photo with the help of digitized photographic silhcuettes' is advanced. The available way is that both random error and error from both projective aberration and contractive aberration in the porosity respectively, is restricted by increasing the number of samples from different parts of the same belt, and is emended by calculating with their model.