Search

It is of interest to localize a ground-based LiDAR point cloud on remote sensing imagery. In this work, we tackle a subtask of this problem, i.e. to map a digital elevation model (DEM) rasterized from aerial LiDAR point cloud on the aerial imagery. We proposed a contrastive learning-based method that trains on DEM and high-resolution optical imagery and experiment the framework on different data sampling strategies and hyperparameters. In the best scenario, the Top-1 score of 0.71 and Top-5 score of 0.81 are obtained. The proposed method is promising for feature learning from RGB and DEM for localization and is potentially applicable to other data sources too. Source code will be released at https://github.com/wtseng530/AVLocalization., Accepted for publication in the ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences (online from July 2022)

The availability and accessibility of remote sensing (RS) data, cloud processing platforms and provided information products and services has increased the size and diversity of the RS user community. This development also generates a need for validation approaches to assess data quality. Validation approaches employ quality criteria in their assessment. Data Quality (DQ) dimensions as the basis for quality criteria have been deeply investigated in the database area and in the remote sensing domain. Several standards exist within the RS domain but a general classification established for databases has been adapted only recently. For an easier identification of research opportunities, a better understanding is required how quality criteria are employed in the RS lifecycle. Therefore, this research investigates how quality criteria support decisions that guide the RS lifecycle and how they relate to the measured DQ dimensions. Subsequently follows an overview of the relevant standards in the RS domain that is matched to the RS lifecycle. Conclusively, the required research needs are identified that would enable a complete understanding of the interrelationships between the RS lifecycle, the data sources and the DQ dimensions, an understanding that would be very valuable for designing validation approaches in RS. (VLID)2625878

Se proponen novedosas fórmulas para evaluar la certeza de la cartografía

Forest structure is a key element to determine the capacity of mountain forests to protect people and their assets against natural hazards. Airborne laser scanning offers new ways for describing forest structure in 3D. This study aimed at developing a generic automated approach for assessing and quantifying forest structure using landscape metrics on height class patches of the normalized crown model (nCM). These patches were built up from objects that were obtained by segmentations. Two separate multi-resolution segmentations were carried out: level 1 objects represented tree crowns and collectives of tree crowns, level 2 objects represented forest stands. Level 1 objects were classified into four height classes and overlaid with level 2 stands in order to calculate landscape metrics as the Shannon Evenness Index (SHEI) and the Division Index (DIVI). The SHEI could not sufficiently represent the vertical layering of the stands. Canopy density values of each height class were used instead. The DIVI proved to be a suitable measure to distinguish between dense and open crown closure. By means of the DIVI and canopy density values, 85% of the forest area could be correctly assigned to one of the six discrete forest structure types. With the approach presented, resource and natural hazard managers can easily assess the structure of different forests and as such can better take into account the protective effect of forests.