Research On Extraction Of Vegetation Parameters Based On Hemispherography

In the research of remote sensing,agronomy,and environmental science,there appears a great demand for instant access to various indicators of growing vegetation and crop.Therefore,hemispherical photography has been favored by scholars.It not only avoids the cumbersome direct measurement method and the damage to the vegetation canopy,but also cuts the high cost of remote sensing measurement and has higher versatility.This thesis systematically studies the whole process of hemisphere photography,improves some of its procedure,and expands some other measurable parameters.In this thesis,the process of extracting vegetation parameters by hemisphere photography is studied from the aspects of vegetation canopy image acquisition,image processing,parameter inversion and precision analysis,also,several improvement schemes are proposed and verified as follows :(1)In order to ensure the diversity of experimental samples and reduce the accidental factors as the error in the research process,this thesis collected image of vegetation canopy in multiple regions and multiple types in the country.The conditions while taking these samples include day and night,cloudy and sunny,etc.,during multiple growth periods of crops.At the same time,the relevant parameters were recorded in the meanwhile using the LAI2000 vegetation canopy structure analyzer.The large size of samples provide an important reference for the subsequent parameter inversion and accuracy calculation.(2)In the stage of canopy image processing,this thesis focuses on the correlation algorithm between foreground pixel and background pixel.Firstly,several traditional gray threshold segmentation algorithms are studied.Sample verification proves these algorithms are not adaptive.Therefore,the multi-scale segmentation algorithm based on region growing is used for the first time among similar researches.The pixels with higher homogeneity are firstly clustered,and the clustered pixels are learned and trained by SVM.Then,the ability to distinguish the pixels is generalized.This method has significantly stronger adaptability than the traditional gray threshold segmentation algorithm,thus improves the overall accuracy of subsequent calculations.(3)In the stage of inverting vegetation parameters,this paper starts from the leaf area index,firstly obtains the optimal porosity by cutting the canopy field of view,and then derives the relationship among leaf area index,projection function,and the angle of view the according to Lang’s hypothesis and Lambert-Beer law.Comparing the projection functions from different angles of view,it is found that the projection function is always a fixed value when the viewing angle is about 57.5°,which further deduces the method of calculating the leaf area index using only the porosity under a single viewing angle,and then based on Miller integral formula,the calculation method of the leaf area index for adding canopy porosity from all angles of view is derived.(4)In this thesis,the calculation of some vegetation parameters needs to correct the view of distorted fisheye image into the view under the vertical projection plane.Therefore,the equidistant spherical model is used to simulate the process of imaging the fisheye lens,and the general formula for projecting the three-dimensional coordinate system to the imaging plane is obtained.After the actual fisheye image is applied,the correction effect is found remarkable.On this basis,cluster index,leaf density,DIFN and vegetation coverage were derived.