No abstract

This paper presents a method of validating the MODIS (Moderate-Resolution Imaging Spectroradiometer) surface reflectance (MOD09) based on ground spectral measurements and high-resolution remote sensing images. Given that the spatial resolution of the MODIS image is too low to directly compare ground measurements with the MODIS pixels, registration is first made between MODIS and a highresolution image. Afterwards, the same sizable sample area to one MODIS pixel is chosen in a high resolution image (a pixel in MODIS corresponds to many pixels in a high-resolution image), after which image classification is conducted in each sample area. Finally, reflectance is measured for each class in each area. Taking the ratio of every class area to each sample area as the weight, the weighted mean reflectance is then calculated for each sample area (a pixel for MODIS), which is treated as the measured reflectance of the corresponding pixel in the MODIS image. A contradistinctive analysis is carried out between measured reflectance and MODIS retrieval reflectance from the space in order to determine the accuracy of the MODIS land surface reflectance product and correct its error. The experiment indicates that a linear positive correlativity between the MODIS pixel retrieval reflectance from space and the measured (calculated) surface reflectance exists. From this, we can build an error correction model of the MODIS surface reflectance product by linear regression. The accuracy of the MODIS surface reflectance products after error correction is remarkably improved. Likewise, the relative error can be controlled by 10%, thereby satisfying the requirements of various remote sensing applications.

LIDAR is a new promising technique in obtaining instantly 3D point cloud data representing the earth surface information. In order to extract valuable earth surface feature information for further application, 3D sub-randomly spatial distributed LIDAR point cloud should be filtered and classified firstly. In this article, a new LIDAR data filtering and classification algorithm is presented. First, the points' neighboring relation and height-jump situation in TIN (triangulated irregular network) model for 3D LIDAR point cloud are analyzed. After that, the filtering algorithm based on TIN neighboring relation and height-jump is presented. Third, an assistant plane is designed in TIN neighborhood filtering algorithm in order to yield more effective filtering result. Then, the LIDAR points are classified into bare ground points, building points and vegetation points using the above filtering algorithms. The experiment is performed using the airborne LIDAR data, and the result shows that this method has better effect on filtering and classification of LIDAR point cloud data.

No abstract

No abstract