Lidar (Laser Scanner) is capable of collecting a large number of 3D points, in which abundant surface features are implied in the distribution of point cloud data. However, these surface features should be extracted to from explicit information, i.e., it is necessary to transfer the point cloud data into mathematical expressions or vector data descriptions. The proposed algorithm of surface reconstruction is based on the schemes of surface growing and surface feature fitting. It merges the co-plan points and extracts surface features from point cloud data. There are two factors to conduct the growing process: the angle between two normal vectors of adjacent patches and the distance of the point from the growing surface. Every merged cell is considered as a small patch, then the connect areas by region growing regarded as a surface feature. The reconstructive surface features in the proposed method include planar, spherical,cylindrical, and polynomial surface features. The experimental data include ground-based Lidar and airborne Lidar. The overall results show the successful application examples of the proposed algorithm. In the experiment, the initial parameters such as grid sizes, threshold of angle, threshold of distance and the growing seed position are also discussed and extracted from point cloud data using a semi- automatic method. The results of reconstructive surface provide points cluster with the same surface characteristics and fitting parameter of the surface features. The extracted surface features will be useful for 3D modeling

ABSTRACT Systematic errors of airborne Lidar data cause elevation offset of point clouds. Strip adjustment is one of the ways to reduce systematic errors. Using strip adjustment, the locations of conjugate blocks or tie points have to be detected first and they usually be manually selected and decided with laborious and time-consuming efforts. The purpose of this study is to develop a method for automatically selecting conjugate blocks or tie points. In this article, the tensor voting method is adopted for the extraction of planar features from Lidar data and an artificial neural network method is applied to match the planes with similar topologic properties. The Bintree method is used for increasing the success rate of classification based on the artificial neural network algorithm. The gravity centers of matched conjugate planes are regarded as tie points for strip adjustments in this study. The advantage of the current algorithm is that the choice of tie points can be executed automatically. The results of experiments of strip adjustments show the feasibility of our algorithm to improve the height accuracy of airborne Lidar data.

The Biomass condition is important index in forest fields. The used information contain number of stems, stand position, tree height, timber volume, canopy cover, etc. In literature reviews, several studies extracted canopies using field surveying, aerial imagery, radar data, LIDAR data and so on. However, those manners are labor intensive and hard to identify crown edges in concentrated forest. Following the development of airborne laser scanning systems, this study proposes a scheme to detect forest canopies using high density point clouds. This study analyzes canopy profiles using empirical mode decomposition method to find the vertical discontinuities, i.e. crown edges. Finally, those adjacent detected boundaries are merged to describe complete canopy boundaries. The experimental results indicate that the proposed scheme has the capability to extract stand position, tree height, and crown edges.

Cloud Ceiling is a parameter describing the cloud used in meteorology. This parameter is related to the amount of the cloud and the cloud height at a specific location. From the cloud observations, the cloud ceiling can be derived. This study investigates the probability of cloud caused risk to aerial mapping, based on the historic record of the cloud observations. Because the cloud status could vary drastically from day to day, month to month, and year to year, it is very difficult, if not impossible, to predicate the cloud conditions. This study utilizes more than 10 years cloud observation for the statistical analysis. The observation from Magong airport located in Penghu is analyzed as a sample. It is demonstrated that the monthly statistics does provide information to the cloud caused risk assessment. However, the cloud ceilings are still difficult to predict, as observed from the yearly changes of monthly probability.

Among models sufficiently employed for positioning and mapping tasks by using satellite imagery, rational function model (RFM) performs almost as equally well as rigorous sensor model when fulfilling some restrictions, and has gained increasing popularity due to its transparency and convenience on application side. One disadvantage of RFM is the issue of rank deficiency when estimating high-order RFM coefficients by exclusively using ground control points with insufficient geometry as referring to RFM. Although the approach of regularizing normal matrix by adding a small multiplication of the identity matrix has been commonly used to stabilize the system and obtain the solution, non-smoothing processing still gets bothered and suggests that more efficient methods are welcome and expected. To this end, the authors proposed a method in which correlated parameters are to be eliminated. The preliminary result shows that effectiveness of the proposed method towards the solution is highly performed. Apart from that, identifying the correlations among the parameters highlights the very unique contribution as compared to other alternatives.