Point cloud data enhancement by matching multispectral images

Abstract

Generally, remote sensing images were two-dimensional multispectral data. The images can be classified more efficiently and precisely via ground features with different characteristics differ in spectrum. As processing in LIDAR technology, point cloud data with three-dimensional coordinates contain rich information. LIDAR usually acquires data using only single band, and lacks of multispectral information such as multispectral images. Therefore, this research acquires visible light and near infrared images, via close-range photogrammetry method, then, matching images automatically by free online services to generate visible light and near infrared point clouds with three-dimensional coordinates and color information. At last, one can use three-dimensional affine coordinate transformation to combine different sources of point clouds, and compare the results with LIDAR data, as an assessment for positioning precision. The experiment shows, image matching by near infrared images, point cloud data can increase 27%, much more than only using visible images; image matching by color infrared composition (NIR+R+G), point cloud data can increase 21%, much more than only using visible light images. As the results shows, multispectral point cloud data are helpful to enhance point clouds data.