Range-based intensity normalization of ALS data over forested areas using a sensor tracking method from multiple returns

• DOI: 10.31223/osf.io/k32qw
• Publication Date: 2020-07-10T08:12:35Z
• Authors: Jean-Romain Roussel; Jean-François Bou...; Alexis Achim
• Keywords: bepress|Physical Sciences and Mathematics; EarthArXiv|Life Sciences|Forest Sciences|Other Forestry and Forest Sciences; EarthArXiv|Physical Sciences and Mathematics|Environmental Sciences|Natural Resources Management and Policy; EarthArXiv|Life Sciences|Forest Sciences|Forest Management; bepress|Engineering; EarthArXiv|Physical Sciences and Mathematics|Environmental Sciences; bepress|Life Sciences|Forest Sciences; EarthArXiv|Engineering; 0211 other engineering and technologies; EarthArXiv|Education; bepress|Life Sciences|Forest Sciences|Other Forestry and Forest Sciences; 02 engineering and technology; Education; bepress|Life Sciences; Engineering; bepress|Physical Sciences and Mathematics|Environmental Sciences|Natural Resources Management and Policy; Other Forestry and Forest Sciences; EarthArXiv|Life Sciences; Physical Sciences and Mathematics; bepress|Physical Sciences and Mathematics|Environmental Sciences; bepress|Life Sciences|Forest Sciences|Forest Management; Forest Sciences; EarthArXiv|Life Sciences|Forest Sciences; EarthArXiv|Physical Sciences and Mathematics|Environmental Sciences|Environmental Monitoring; Natural Resources Management and Policy; Life Sciences; 15. Life on land; Forest Management; bepress|Physical Sciences and Mathematics|Environmental Sciences|Environmental Monitoring; EarthArXiv|Physical Sciences and Mathematics; bepress|Education; Environmental Sciences; Environmental Monitoring

Airborne laser scanning (ALS) point-clouds are used in forest inventory to map properties of the resource. In most cases, only the (x,y,z) coordinates of the point cloud are used to build predictive models of forest structure. Despite being recorded and provided by data suppliers, the intensity values associated with each point are rarely used as an input to such models because raw intensity values vary within and between datasets not only due to variations of the target reflectivity, but also due to variations in settings and conditions that may either modify the emitted energy or the distance between the sensor and the targets. Some studies have proposed data-driven methods of normalization, but these are often impossible to apply in practice because of the need to acquire additional reference data. Other studies have proposed model-driven methods of normalization, but in this case the difficulty of application lies with need to know the position of the sensor at any given time during the survey. In this study we applied a method to track the sensor position using pulses associated with multiple returns, which was then used to apply a model-driven correction of the intensity values within several datasets. This normalization method is based only a simple model-driven equation and does not require any reference data, thus making it applicable to any dataset. Our results demonstrate the very high accuracy of the sensor positioning method with an error under 0.5%. Using this information, the model-driven range correction then performed a satisfactory normalization of intensities within different datasets. We provide an open-source and ready-to-use tool to facilitate the application of the normalization method, which in turn could promote a better use of intensity values in ALS studies.