| 3D POINT CLOUD CLASSIFICATION BASED ON MULTISPECTRAL UAV IMAGERY ONLY WITH SPECTRAL INFORMATION |
| Paper ID : 1350-SMPR |
| Authors: |
|
bahram sadeghi *, farhad samadzadegan Department of Surveying and Geomatics Engineering,University College of Engineering,University Of Tehran,Tehran,Iran |
| Abstract: |
| Today, many analyses are based on point clouds in photogrammetry and remote sensing. Two reliable sources for generation point cloud are airborne laser scanning (ALS) and dense matching of aerial images. ALS is very suitable for forest areas, but in urban areas, especially buildings show poor performance. Because it can’t directly derive linear features such as lines and edges. Although the ALS point cloud classification has been well researched, there are few studies in relation to image matching point clouds especially in relation to multispectral images that contain rich spectral information. Working with multispectral images has challenges. One of the great challenges is the band to band registration. In this study, the SIFT algorithm is used to extract the corresponding features of each of the bands. After band to band registration and generating point cloud derived from the dense matching of multispectral images, Now, The point cloud filtering is achieved, with this point cloud divided into two parts: ground and non-ground points. Then, each of these points is classified into several parts using the color information or the spectral index used here from the NDVI. To classify this point cloud, only color and height information is used as a threshold. According to the results, the use of only spectral information extracted from multispectral images for point cloud classification has been highly effective. |
| Keywords: |
| Point cloud, classification, multispectral images, Band to band registration, Dense matching, SIFT, NDVI |
| Status : Paper Accepted (Poster Presentation) |
