【作者】 智长贵；

【导师】 郎奎建；

【作者基本信息】 东北林业大学 ， 森林经理， 2005， 博士

【摘要】 利用航空像片(以后简称航片)进行林业制图和森林测量是森林资源调查的主要任务。本文根据自检校光束法区域网平差原理,利用ERDAS的OrthoBhSE模块对2003年东北林业大学帽儿山试验林场拍摄的176张航片,共13条航带进行了空三加密,并对其进行了数字微分纠正,最后生成了帽儿山林场比例尺为1∶25000影像林相图和传统的林相图以及各个施业区比例尺为1∶10000影像林相图和传统的林相图,影像图点位中误差仅为2.953m,达到了相当高的精度。这也是我国林业部门第一次通过数字微分纠正生成的正射影像林相图和林相图。正射纠正消除或大大降低了因摄影机倾斜和地形起伏两种因素引起的影像位移,消除或大大降低了航片各种误差的影响,主要包括:摄影机物镜畸变、摄影感光材料变形、大气折光和地球曲率对像点坐标的影响。最后,本文以尖砬沟施业区为例,利用制作的正射影像林相图以及光束法空中三角测量得到的航片外方位元素和数字表面模型,进行了森林测量研究,包括面积、郁闭度、林分平均高和单个立木树高测量。在此过程中,本文主要完成了以下工作: 第一、采用GPS的RTK测量技术,按照近似密周布点方案,在野外共测得帽儿山像控点91个(全部为Fixed状态),全部为平高点,加上购买的4个控制点共95个像控点,像控点最大误差为dm级。 第二、对四幅1∶50000比例尺的地形图采用600dpi分辨率进行扫描,采用3次多项式进行纠正,配准精度分别为:第一幅地形图的控制点X坐标误差均方根为0.0036 pixel,Y误差均方根为0.0014 pixel,总误差均方根为0.0038 pixel;第二幅地形图的控制点X误差均方根为0.0042pixel,Y误差均方根为0.0020pixel,总误差均方根为0.0047pixel;第三幅地形图的控制点X误差均方根为0.0043pixel,Y误差均方根为0.0022 pixel,总误差均方根为0.0048 pixel;第四幅地形图的控制点X误差均方根为0.0026 pixel,Y误差均方根为0.0007 pixel,总误差均方根为0.0027 pixel。最后,对拼接后的四幅地形图进行精度检验,结果显示:六个检查点中,X坐标误差(ΔX)最大值为1.741m,最小值为0.037m,平均值为1.119m;Y坐标误差(ΔY)最大值为1.501m,最小值为0.516m,平均值为1.006m。单个检查点的坐标误差均方根(T_i)中最大值为2.026m,最小为0.517m,所有检查点坐标总误差均方根(T)为1.749m。 第三、采用三角形线性插值及三角形非线性插值方法,分别生成了分辨率为5m的DEM,并采用检查点法和等高线回放抽样法对DEM进行了精度检验,其中回放等高线方法中,首次提出了“重合度”检验法。检查点法共选择49个控制点进行检验,结果显示:线性插值总误差均方根为4.1m,非线性插值总误差均方根为3.8m,非线性插值精度偏高。回放等高线法,共选择18个公里网格的等高线与原等高线进行比较,结果显示:线性插值其重合度为96%,非线性插值重合度为96.5%。非线性插值总体回放精度较高。 第四、首次提出了利用绝对与相对位置求航片外方位元素初始值的方法,解决了应用自检校光束法进行区域网平差所需要的外方位元素初始值问题。 第五、运用自检校光束法区域网平差理论,对各航片的外方位元素进了解算,对原来更多还原

【Abstract】 Forestry map-making and forest measurement are main tasks using airscapes in forest resource survey .according to the theory of self-calibrating bundle aerotriangulation .Using OrthoBASE module of ERDAS software, 176 digital airscapes consisted of 13 strips were orthorectified differentially finally , orthophoto - forest map and traditional forest map which only have 2.953 meter RMS was made firstly including total maoershan forest farm map of 1:25000 and every forest region map of 1:10000 in forestry department of China. Image displacement caused by camera tilt and terrain fluctuation is eliminated or reduced greatly. And image errors mainly caused by camera lens distortion, photographic material deformation, atmospheric refraction and earth’s curvature are eliminated or reduced greatly. Finally, using Janlagou forest region as example, forest measurement of area, crown closure, forest stand mean height and individual standing tree height is studied by means of map include digital orthomage and forest map , elements of exterior orientation and Digital Surface Model. Main tasks have been finished in this paper, as follows:Firstly , according to the control point distribution plan , 91 image control points include X,Y and Z coordinates (all are Fixed) were measured adopting RTK measurement technology of GPS. Total 95 image control points are gained adding 4 control points which are purchased from survery and mapping department, the control point maximum error is decimeter.Secondly , four relief maps with the scale of 1:50000 were scanned by 600 dpi and rectified by three order polynomial , coordinate accuracy of every map is as follows: X coordinate error root-mean-square (ab, RMS) is 0.0036 pixel , Y coordinate error RMS is 0.0014 pixel , total error RMS of X and Y is 0.0038pixel in No.l relief map; X coordinate error RMS is 0.0042 pixel , Y coordinate error RMS is 0.0020 pixel , total error RMS of X and Y is 0.0047pixel in No.2 relief map; X coordinate error RMS is 0.0043 pixel , Y coordinate error RMS is 0.0022 pixel , total error RMS of X and Y is 0.0048pixel in No.3 relief map; and X coordinate error RMS is 0.0026 pixel , Y coordinate error RMS is 0.0007 pixel , total error RMS of X and Y is 0.0027pixel in No.4 relief map. Finally , mosaic relief map made of four relief maps accuracy is verified , and the results are showed : in the six check points , X coordinate maximum error (△X ) is 1.741meter , minimum 0.737 meter and average 1.119 meter; Y coordinate maximum error (△Y) is 1.501meter , minimum 0.515 meter and average 1.006 meter ;the maximum error is 2.026, and the minimum error is 0.517meter in the total error RMS (T_i )of a single check point; the six check point total error RMS (T) is 1.749meter.Thirdly , DEM with 5 meter resolution was made respectively by triangle linear interpolation and triangle nonlinear interpolation .and the accuracy of two DEMs was verified using the method of check point and method of contour line recreation and sampling . Total 49 check points were checked in the method of check point , and the results show that the total error RMS of linear interpolation is 4.1meter , and the total error RMS of nonlinear interpolation is 3.8 meter , accuracy of nonlinear interpolation is slightly higher than linear interpolation . contour lines in Total18 grids of kilometre were sampled and compared with original contour lines , and the results show that degree of coincidence is 96% for linear interpolation , and 96.5% for nonlinear interpolation .accuracy of nonlinear interpolation is slightly higher than accuracy of linear interpolation.Fourthly, A method of absolute and relative position is put forward to calculate the original values of exterior elements , it solved the problem of exterior elements original value in self-calibrating bundle aerotriangulation.Fifthly f According to the theory of self-calibrating bundle aerotriangulation , exterior orientation elements of every airscape , original GCPs and tic points were calculated . the differences between original GCPs and calculated GCPs were computed and showed that the maximum X coordinate error was 4.604 meter , total X coordinate errors RMS were 1.449 meter ; the maximum Y coordinate error was 4.833 meter , total Y coordinate errors RMS were 1.528 meter; the maximum Z coordinate error was 4.244 meter ? total Z coordinate errors RMS were 2.052 meter. So we can think that the accuracy is much higher, because 1 mm in orthophoto means 10 meter on the ground for the orthophoto with the scale of 1:10, 000 , this fact show that the GCPs error in the orthophoto is less than lmm. tic points maximum x image coordinate error is 0.0152 pixel , and maximum y image coordinate is 0.0144 pixel the accuracy is much higher.Sixthly , Three methods of forest map making from ploted areoscapes to stand map are put forward throgh working experiences.Seventhly, Many factors of compartment and subcompartment in forest map are marked automatically by means of secondary development function of AutoCAD software.and productivity of forest map making is improved greatly.Eighthly, comparing area measurement error of orthorectified airscape and unorthorectified airscape in three kinds of terrain. The results show that measurement error of flat ground is lowest , that is 1.2%; 3.3% is in middle and low mountain area , and 11.5% is the most biggest measurement error in high mountain area. These show that area measurement error is much higher in the unorthorectified airscape , we should overlook it . and counting up the forest survery results of Maoershan mountain forest farm in 2004.Ninthly, using the methods of unsupervised classification and recoding to calculate degree of closing are put forward , this paper get much better results which show that the calculation vaule is the same as measurement vaule in No.l sample area , and the difference between calculation value and measurement value is only 0.1.Tenthly, Using Digital Vegetation Height Model (ab.DVHM) made of the difference between Digital Surfacing Model(ab.DSM) and Digital Elevation Model (ab.DEM) to measure stand mean height is put forward firstly in this paper. DVHM can reflect totally vegetation height distribution , and to a certain degree .reflect stand age-class distribution .measuring average height of every subcompartment in Janlagou forest region by means of DVHM.Finally, according to the theory of bundle aerotriangulation, using exterior orientation elements to calculate the elevation of crown top and root of tree is put forward firstly .this paperconducts the experiment on tree measurement , the results show that error can be less than 2 meter by comparing measurement tree height with computation tree height. Stereoscope is not used to measure trees in the method.更多还原