The Method Study Of LIDAR Data Filtering

Airborne LIDAR is becoming more and more popular at home and abroad for its high efficiency. It provides a reliable approach to obtain a wide range of terrain data in a short time for the surveying and mapping industry. Today, the foreign and domestic research on LIDAR, instead of studying the hard device as it did before, began to focus on the related processing of the LIDAR data. As the original data of laser points cloud provided by the airborne LIDAR data cannot be processed immediately, filtering becomes the most essential process when working with LIDAR data. The quality of the data directly relates to the quality of the DEM and terrain data generated subsequently. For this reason, this research looks at how to develop a quick and accurate method to transform LIDAR points-cloud data. The following is the methodology employed in this paper:1. The components and the measuring principle of the airborne laser radar system were introduced. The three-dimensional coordinates of the laser points on the LIDAR system were determined through the principles of solid geometry. The feature of airborne LIDAR was analyzed and a comparison to the aerial photogrammetry and the InSAR technique respectively was given.2. The airborne LIDAR data was analyzed. Its characteristics and components including the LAS data structure were studied. The application of LIDAR data was introduced.3. The filtering principle of airborne LIDAR data and the problems during the experiment were analyzed. Deeper study and analysis on common filtering algorithms was discussed to understand the advantages and disadvantages of the present algorithms.4. The mobile surface fitting filtering algorithm was studied in depth, and some improvements in certain steps were proposed. Although the algorithm could filter the low points, it affe (?)ed the structure of initial fitting surface before the filtering process. A new algori(?)an was proposed to remove the low points before choosing the seed region in the beginning. The new algorithm was based on grid partition and used least squares method with several points to fit the quadric instead of using the coefficients of the quadric calculated by six points directly. An experiment was taken to testify the feasibility of the improved algorithm and a filter errors statistical analysis was also performed. A discussion on the types of data that the algorithm could be applied and the performance under different sizes of grid were made.The results show that the improved filtering algorithm could classify more ground points correctly, thus significantly reducing type I errors, without type II errors increasing obviously. Meanwhile, the total errors decreased accordingly and a better filtering result was achieved.