Search

Humanitarian organizations must have fast and reliable data to respond to disasters. Deep learning approaches are difficult to implement in real-world disasters because it might be challenging to collect ground truth data of the damage situation (training data) soon after the event. The implementation of recent self-paced positive-unlabeled learning (PU) is demonstrated in this work by successfully applying to building damage assessment with very limited labeled data and a large amount of unlabeled data. Self-PU learning is compared with the supervised baselines and traditional PU learning using different datasets collected from the 2011 Tohoku earthquake, the 2018 Palu tsunami, and the 2018 Hurricane Michael. By utilizing only a portion of labeled damaged samples, we show how models trained with self-PU techniques may achieve comparable performance as supervised learning., Comment: 7 pages, 1 figure, Artificial Intelligence for Humanitarian Assistance and Disaster Response Workshop, NeurIPS 2021

National land use database construction projects in county(or city)level was started in 1999.As an important part of “Digital Land Project”, Government in various level paid much attention to database construction, preliminary achievement has been made, but difference between area with various social and economic condition is objective and obvious. These areas with comprehensive basic material have already finished building land use database, which has been applied in land management, on the contrary, those region with incomplete basic data encountered enormous difficult, owing to historical reasons, present situation of data can not meet the request of establishing database,so the work is still walk back and forth. these original data must perform part or overall updating investigation in order to meet the need for database construction. In this paper, we discuss a series of new technology(or methods) adopted in the course of land use current situation updating partially (or overall renewing)and database construction, at the same time, bring up the choice method of the technological route under various basic materials situations, finally, suggestion solution is made out in the light of common problem.

Taking Xuzhou city as the case study area, EO-1 Hyperion hyperspectral remote sensing image captured in 2004 was chosen as the data source to extract impervious surface area (ISA) by spectral mixture analysis using LSMM model. In order to demonstrate the advantages of hyperspectral images and LSMM for ISA extraction, hyperspectral image was compared with multispectral Landsat TM image, and LSMM was compared with traditional hard classifiers, indicating that hyperspectral image is more effective than multispectral data for urban impervious surface extraction, and unmixing is superior to hard classification methods.

Remote sensing of Urban Heat Islands (UHIs) usually used Land Use and Land Cover (LULC), Normalized Difference Vegetation Index (NDVI), normalized difference built-up index (NDBI) and Vegetation-Impervious Surface Area (ISA)-Soil (V-I-S) model separately in the past. The purpose of this paper is to examine the relationship between surface thermal patterns with land cover types, NDVI and VIS model as a whole. LST was derived using three different algorithms from remotely sensed images at two different scales and LUCC map was obtained by SVM classification method. The results demonstrate that LST retrieval from Landsat ETM+ image by improved MWA algorithm is more suitable for analyzing relationship between urban structures with urban heat islands in Xuzhou. By analyzing the relationship between LST and land cover type, it also indicated that SVM classified result and VIS have a good uniformity. The analytical correlation between UHIs and different indices indicated that correlation between indices with UHIs change slightly among the different fraction in different city structures. The special distribution of heat island shows that building area, bare land, semi-bare land and land under development are warmer than other surface type, and higher temperature in the UHIs was related to certain land cover types and distributed with a scattered pattern. By making an integration analysis of NDVI, NDBI and LST, it was confirmed that strong positive correlations between LST and NDBI existed, as with negative correlations between LST and NDVI.