Remote Sensing Datasets

Remote Sensing Datasets - آزمايشگاه سنجش از راه دور rslab

Building Damage Detection Datasets

Dataset (1), 2023 Turkey earthquake:

On February 06, 2023, at 4:17 a.m. (1:17 UTC), an earthquake of 7.8 (on the Richter scale) hit southern and central Turkey as well as southwestern Syria. Fig.1 illustrates the WorldView II image of the city of Nurdağı after the earthquake and the corresponding ground truth (GT) which shows that the city was densely built and undergoes post-earthquake damage.

Dataset (1)

Paper

Dataset (2), 2023 Morocco earthquake:

On September 08, 2023, at 22:11 UTC, a 6.8 magnitude earthquake struck the High Atlas Mountain, 71 km southwest of Morocco. Fig.2 displays the WorldViewII image of the Talat N'Yaaqoub region and the corresponding GT.

Dataset (2)

Paper

Dataset (3),2023 Libya Flood:

On September 11, 2023, heavy rain led to the collapse of two dams in Libya, and about 30 million m3 of water flowed into the city of Derna, which caused much damage and destroyed many buildings. Fig.3 shows the GeoEye-1 image of a part of the city of Derna and the corresponding GT in the floodway area.

Dataset (3)

Paper


(a)	(b)
Fig. 1. Turkey earthquake dataset: (a) RGB post image and (b) ground truth.


(a)	(b)
Fig. 2. Morocco earthquake dataset: (a) RGB post image and (b) ground truth.


(a)	(b)
Fig. 3. Libya flood dataset: (a) RGB post image of the floodway showed by green polygon and (b) ground truth.

3D Multiple Building Change Detection Datasets

The first dataset is a set of images and cloud points taken by UAV in 2011 and 2016 from a densely built urban area in the city of Mashhad. This study makes use of ortho and DSM images obtained from cloud points at both times. The spatial resolution of this data (10 cm) allows for accurate identification and change detection in buildings. To generate ground truth, Ortho images, and Google Earth images corresponding to them were used. Building polygons were drawn in the Global Mapper software and the different layers of "No building change, Newly built, Demolished, and Taller" were separated and numbered with different codes.

The second dataset is from the 22nd district of Tehran. This is one of Tehran's newest urban areas, located northwest of the city. It is two pairs of GeoEye-1 satellite images with a spatial resolution of 0.5 meters from 2009 and 2013. The semi-global matching technique was used to generate point clouds from these satellite images. Ground truth was generated using satellite images from both time and Google Earth. The polygons were drawn and labeled in the Global Mapper.

Mashhad Dataset (Part-1, Part-2, Part-3)

Tehran Dataset

Paper


Time1	Time2	Ground Truth
Mashhad Dataset


Time1	Time2	Ground Truth
Tehran Dataset

Iran (North) Change Detection Datasets

This dataset includes Sentinel-2 satellite data from urban areas of northern Iran, specifically those in Gilan, Mazandaran, and Golestan provinces, where urban areas could be quite diverse and complex. These urban areas are located mostly alongside forests and agricultural lands but are also partially located near ports, coastal areas, and on foothills and mountainous terrains. The Sentinel-2 satellite images taken in 2017 and 2021 as well as their corresponding ground truth, which was obtained by examining Google Earth images taken at the same time.

For downloading related dataset please email hasanlou@ut.ac.ir

Related Paper

The nation-wide wetland inventory map for Iran.

SAR Datasets

(1) SAR Sentinel-1 Satellite Imagery in VV & VH polarization

The Sentinel-1 SAR imagery, in the Interferometric Wide (IW) mode, C-band, with dual polarization VV (Vertical transmit Vertical receive), VH (Vertical transmit Horizontal receive), was acquired on 1 March 2017 from Kuh Sefid district, which is a village located in the Central District of Qom County, Qom, Iran. This area, which has a hot and dry climate with annual precipitation of about 115.5 mm, the highest temperature of 39.7 °C in July and the lowest temperature of 0.4 °C in December, is affected by severe salinity hazard mainly due to the vicinity to the Salt Lake Qom. The surface soil texture in this region varies between silt loams to silty clay loam, which has a yellowish brown to dark brown color.

For downloading related paper please go to this website in "Forest, Soil and Vegetation Remote Sensing" section, paper entitle "Soil Salinity Mapping Using Dual-Polarized SAR Sentinel-1 Imagery".

Download MATLAB data files:

Features (12.0 KB)

Indices Names (1.0 KB)

Target (1.0 KB)


The color composite image of the Sentinel-1 data from Kuh Sefid region (Red: VV, Green: VH, Blue: VH/VV)

The VH polarization of Sentinel-1 Satellite imagery acquired from Kuh Sefid district	The VV polarization of Sentinel-1 Satellite imagery acquired from Kuh Sefid district

(2) PolSAR UAVSAR Change Detection Images- San Francisco

Two pairs of single-look quad-polarimetric SAR images acquired by the UAVSAR airborne sensor in L-band over an urban area in San Francisco city on 18 September 2009, and May 11, 2015, are used for the experiments. The dataset #1 have length and width of 200 pixels and the dataset #2 length and width of 100 pixels.

For downloading related paper, please go to this website in "Change Detection Algorithms" section, paper entitle "Land Cover Changes Detection in Polarimetric SAR Data Using Direct, Similarity and Distance Based Methods".

Download MATLAB data files:

Dataset #1, 2009 (22.7 MB)

Dataset #1, 2015 (22.8 KB)

Dataset #1, Ground truth (1.0 KB)

Dataset #2, 2009 (5.6 MB)

Dataset #2, 2015 (5.6 KB)

Dataset #2, Ground truth (1.0 KB)


a	b	c

d	e	f
Pauli images of datasets, (a) dataset #1 in 2009, (b) dataset #1 in 2015, (c) ground truth of dataset #1, (d) dataset #2 in 2009, (e) dataset #2 in 2015, (f) ground truth of dataset #2

(3) PolSAR UAVSAR Change Detection Images- Los Angeles

Two L-band UAVSAR datasets, these two datasets belong to the city of Los Angeles, California, were acquired on April 23, 2009, and May 11, 2015, by the JAV Propulsion Laboratory/National Aeronautics and Space Administration UAVSAR. There are 786´300 pixels in the first data-set and 766´300 pixels in the second data-set. Figures 1-a, 1-b, 1-d, and 1-e show the RGB (Red: |HH – VV|; Green: 2|HV|; Blue: |HH + VV|) Pauli images of the two data-sets of the PolSAR images. The GT images connected with these datasets, shown in Figures 1-c, and 1-f, were prepared by using Google Earth images.

Download Dataset #1:

Download Dataset #2:

(a)

(b)

(c)

(d)

(e)

(f)

Figure 1. Pauli decomposition of UAVSAR images taken over Los Angeles, California on (a) and (d) April 23, 2009. (b) and (e) May 11, 2015. (c) and (f) ground truths, where white means change area and black means no-change area. Top: dataset#1. Bottom: dataset#2.

)4) Earthquake Damage Region Detection Images

On November 12, 2017 at 17:35 local time, an earthquake with a moment magnitude of 7.3 struck Ezgeleh, Kermanshah Province, Iran. for this end, we presented a novel framework for rapid damage region assessment. the proposed method is applied in two main phase: (1) built-up area detection, (2) multi-level damage detection. The proposed method was utilized Sentinel-1 and Sentinel-2 datasets.

For downloading MATLAB Kermanshah earthquake dataset please go to this website in "Target Detection Algorithms" section, paper entitle "Earthquake Damage Region Detection by Multi-Temporal Coherence Map Analysis of Radar and Multispectral Imagery"

Download MATLAB data files:

Dataset of Qasr-Shirin

Dataset of Sarpol-Zahab

The case study areas from Sentinel-2 data that acquired on the 5th of 2017, (a) Qasr-Shirin, and (b) Sarpol-Zahab.

	8
(a)	(b)
Ground Truth of (a) Qasr-Shirin, (b) Sarpol-Zahab.

(5) The Soil Moisture Active Passive Validation Experiment 2016 (SMAPVEX16-MB)

The Soil Moisture Active Passive Validation Experiment 2016 (SMAPVEX16-MB) campaign was conducted over an agricultural test site (Figure) measuring 26 km × 48 km and located in southern Manitoba (Canada). Numerous SAR experiments have been conducted over this site, spanning several decades (Bhuiyan et al., 2018). The site is dominated by annual agricultural production with a diverse crop mix that includes canola, corn, soybean, wheat and oats. The SMAPVEX16-MB was conducted during the summer of 2016 and was focused on the monitoring of soil surface and crop. Soil surface moisture and roughness, and vegetation water content (VWC) were mainly collected. Soil surface moisture are available at the depth of 5, 10, 20, 50 cm. The distribution of the ground sites is shown in Figure.For more information "H. McNairn et al., "SMAPVES-16 Experimental Plan," 1–71 (2016)".

SV16M RISM Soil Station Product Specifications

SV16M Temporary Soil Moisture Station Product Specifications

SV16M CropScan Product Specifications

SV16M Soil Roughness Product Specifications

SMAPVEX16 Manitoba Experiment Plan

SMAPVEX16 Manitoba Soil Moisture Data

McNairn, H., K. Gottfried, and J. Powers. 2018. SMAPVEX16 Manitoba Station Soil Moisture Data, Version 1. [Indicate subset used]. Boulder, Colorado USA. NASA National Snow and Ice Data Center Distributed Active Archive Center. doi: https://doi.org/10.5067/GMMWKUUCYYIR. [Date Accessed].

SMAPVEX16 Manitoba CropScan Data

McNairn, H., K. Gottfried, and J. Powers. 2018. SMAPVEX16 Manitoba CropScan Data, Version 1. [Indicate subset used]. Boulder, Colorado USA. NASA National Snow and Ice Data Center Distributed Active Archive Center. doi: https://doi.org/10.5067/Y4W64RE5RWBF. [Date Accessed].

SMAPVEX16 Manitoba Surface Roughness Data

McNairn, H., K. Gottfried, and J. Powers. 2018. SMAPVEX16 Manitoba Surface Roughness Data, Version 1. [Indicate subset used]. Boulder, Colorado USA. NASA National Snow and Ice Data Center Distributed Active Archive Center. doi: https://doi.org/10.5067/C18GQDVVRHOY. [Date Accessed].

mantaghe_new

Study area and land cover of the SMAPVEX16-MB ground campaign in Manitoba, Canada

Hyperspectral Datasets

Here you can find information over some public available hyperspectral scenes. All of them are Earth Observation images taken from airborne or satellites.

(1) Indian Pines

This scene was gathered by AVIRIS sensor over the Indian Pines test site in North-western Indiana and consists of 145´145 pixels and 224 spectral reflectance bands in the wavelength range 0.4–2.5 10^-6 meters. This scene is a subset of a larger one. The Indian Pines scene contains two-thirds agriculture, and one-third forest or other natural perennial vegetation. There are two major dual lane highways, a rail line, as well as some low density housing, other built structures, and smaller roads. Since the scene is taken in June some of the crops present, corn, soybeans, are in early stages of growth with less than 5% coverage. The ground truth available is designated into sixteen classes and is not all mutually exclusive. We have also reduced the number of bands to 200 by removing bands covering the region of water absorption: [104-108], [150-163], 220. Indian Pines data are available through Pursue's univeristy MultiSpec site.

Download MATLAB data files:

Indian Pines (6.0 MB)

corrected Indian Pines (5.7 MB)

Indian Pines ground truth (1.1 KB)

Ground truth classes for the Indian Pines scene and their respective samples number
#	Class	Samples
1	Alfalfa	46
2	Corn-notill	1428
3	Corn-mintill	830
4	Corn	237
5	Grass-pasture	483
6	Grass-trees	730
7	Grass-pasture-mowed	28	Sample band of Indian Pines dataset
8	Hay-windrowed	478
9	Oats	20
10	Soybean-notill	972
11	Soybean-mintill	2455
12	Soybean-clean	593
13	Wheat	205
14	Woods	1265
15	Buildings-Grass-Trees-Drives	386
16	Stone-Steel-Towers	93	Ground truth of Indian Pines dataset

(2) Salinas

(2-1) Salinas scene

This scene was collected by the 224-band AVIRIS sensor over Salinas Valley, California, and is characterized by high spatial resolution (3.7-meter pixels). The area covered comprises 512´217. As with Indian Pines scene, we discarded the 20 water absorption bands, in this case bands: [108-112], [154-167], 224. This image was available only as at-sensor radiance data. It includes vegetables, bare soils, and vineyard fields. Salinas ground truth contains 16 classes.

Download MATLAB data files:

Salinas (26.3 MB)

corrected Salinas (25.3 MB)

Salinas ground truth (4.2 KB)

Ground truth classes for the Salinas scene and their respective samples number
#	Class	Samples
1	Brocoli_green_weeds_1	2009
2	Brocoli_green_weeds_2	3726
3	Fallow	1976
4	Fallow_rough_plow	1394
5	Fallow_smooth	2678
6	Stubble	3959
7	Celery	3579	Sample band of Salinas dataset
8	Grapes_untrained	11271
9	Soil_vinyard_develop	6203
10	Corn_senesced_green_weeds	3278
11	Lettuce_romaine_4wk	1068
12	Lettuce_romaine_5wk	1927
13	Lettuce_romaine_6wk	916
14	Lettuce_romaine_7wk	1070
15	Vinyard_untrained	7268
16	Vinyard_vertical_trellis	1807	Ground truth of Salinas dataset

(2-2) Salinas-A scene

A small sub scene of Salinas image, denoted Salinas-A, is usually used too. It comprises 86´83 pixels located within the same scene at [samples, lines] = [591-676, 158-240] and includes six classes.

Download MATLAB data files:

Salinas-A (1.5 MB)

corrected Salinas-A (1.5 MB

Salinas-A ground truth (587 Bytes)

Ground truth classes for the Salinas-A scene and their respective samples number
#	Class	Samples
1	Brocoli_green_weeds_1	391
2	Corn_senesced_green_weeds	1343	Sample band of Salinas-A dataset
3	Lettuce_romaine_4wk	616
4	Lettuce_romaine_5wk	1525
5	Lettuce_romaine_6wk	674
6	Lettuce_romaine_7wk	799	Ground truth of Salinas-A dataset

(3) Pavia Centre and University

These are two scenes acquired by the ROSIS sensor during a flight campaign over Pavia, nothern Italy. The number of spectral bands is 102 for Pavia Centre and 103 for Pavia University. Pavia Centre is a 1096´1096 pixels image, and Pavia University is 610´610 pixels, but some of the samples in both images contain no information and have to be discarded before the analysis. The geometric resolution is 1.3 meters. Both image ground truths differentiate 9 classes each. It can be seen the discarded samples in the figures as abroad black strips. Pavia scenes were provided by Prof. Paolo Gamba from the Telecommunications and Remote Sensing Laboratory, Pavia university (Italy).

(3-1) Pavia Center scene

Download MATLAB data files:

Pavia Center Part1 (65.00 MB)

Pavia Center Part2 (58.90 MB)

Pavia Center ground truth (34.1 KB)

Ground truth classes for the Pavia center scene and their respective samples number
#	Class	Samples
1	Water	824
2	Trees	820
3	Asphalt	816
4	Self-Blocking Bricks	808	Sample band of Pavia Centre dataset
5	Bitumen	808
6	Tiles	1260
7	Shadows	476
8	Meadows	824
9	Bare Soil	820	Ground truth of Pavia Centre dataset

(3-2) Pavia University scene

Download MATLAB data files:

Pavia University (33.2 MB)

Pavia University ground truth (10.7 KB)

Ground truth classes for the Pavia University scene and their respective samples number
#	Class	Samples
1	Asphalt	6631
2	Meadows	18649
3	Gravel	2099
4	Trees	3064	Sample band of Pavia University dataset
5	Painted metal sheets	1345
6	Bare Soil	5029
7	Bitumen	1330
8	Self-Blocking Bricks	3682
9	Shadows	947	Ground truth of Pavia University dataset

(4) Cuprite

This data sets can be retrieved from AVIRIS NASA site. Among the many datasets available, the. mat archive posted here corresponds to the f970619t01p02_r02_sc03.a.rfl reflectance file. Cuprite is the most benchmark dataset for the hyperspectral unmixing research that covers the Cuprite in Las Vegas, NV, U.S. There are 224 channels, ranging from 370 nm to 2480 nm. After removing the noisy channels (1-2 and 221-224) and water absorption channels (104-113 and 148-167), we remain 188 channels. Aregion of 250x190 pixels is considered, where there are 14 types of minerals. Since there are minor differences between variants of similar minerals, we reduce the number of endmembers to 12, which are summarized as follows "#1 Alunite", "#2 Andradite", "#3 Buddingtonite", "#4 Dumortierite", "#5 Kaolinite1", "#6 Kaolinite2", "#7 Muscovite", "#8 Montmorillonite", "#9 Nontronite", "#10 Pyrope", "#11 Sphene", "#12 Chalcedony".

Download MATLAB data file:

Cuprite (95.3 MB)

Data in the ENVI format with 224 channels: Cuprite_S1_F224.zip (14.0Mb)

Data in the ENVI format with 188 channels: Cuprite_S1_R188.zip (12.9Mb)

Data in the Matlab format with 224 channels: Cuprite_S1_F224.mat (15.4Mb)

Data in the Matlab format with 188 channels: Cuprite_S1_R188.mat (13.7Mb)

Ground Truth: GroundTruth (438Kb) only includes GT: endmembers.


False greyscale image of Cuprite sample.	The ground truth for Cuprite (12 endmembers).

(5) Kennedy Space Center (KSC)

The NASA AVIRIS (Airborne Visible/Infrared Imaging Spectrometer) instrument acquired data over the Kennedy Space Center (KSC), Florida, on March 23, 1996. AVIRIS acquires data in 224 bands of 10 nm width with center wavelengths from 400 - 2500 nm. The KSC data, acquired from an altitude of approximately 20 km, have a spatial resolution of 18 m. After removing water absorption and low SNR bands, 176 bands were used for the analysis. Training data were selected using land cover maps derived from color infrared photography provided by the Kennedy Space Center and Landsat Thematic Mapper (TM) imagery. The vegetation classification scheme was developed by KSC personnel in an effort to define functional types that are discernable at the spatial resolution of Landsat and these AVIRIS data. Discrimination of land cover for this environment is difficult due to the similarity of spectral signatures for certain vegetation types. For classification purposes, 13 classes representing the various land cover types that occur in this environment were defined for the site.

Download MATLAB data file: Kennedy Space Center (KSC) (56.8 MB)

Download MATLAB ground truth file: KSC gt (3.2 kB)

Kennedy Space Center Data

(6) Botswana

The NASA EO-1 satellite acquired a sequence of data over the Okavango Delta, Botswana in 2001-2004. The Hyperion sensor on EO-1 acquires data at 30 m pixel resolution over a 7.7 km strip in 242 bands covering the 400-2500 nm portion of the spectrum in 10 nm windows. Preprocessing of the data was performed by the UT Center for Space Research to mitigate the effects of bad detectors, inter-detector mis-calibration, and intermittent anomalies. Un-calibrated and noisy bands that cover water absorption features were removed, and the remaining 145 bands were included as candidate features: [10-55, 82-97, 102-119, 134-164, 187-220]. The data analyzed in this study, acquired May 31, 2001, consist of observations from 14 identified classes representing the land cover types in seasonal swamps, occasional swamps, and drier woodlands located in the distal portion of the Delta.

Download MATLAB data file: Botswana (78.9 MB)

Download MATLAB ground truth file: Botswana gt (4.0 kB)

Kennedy Space Center Data

(7) Samson

Samson is a simple dataset that is available from the website. In this image, there are 952x952 pixels. Each pixel is recorded at 156 channels covering the wavelengths from 401 nm to 889 nm. The spectral resolution is highly up to 3.13 nm. As the original image is too large, which is very expensive in terms of computational cost, a region of 95x95 pixels is used. It starts from the (252,332)-th pixel in the original image. This data is not degraded by the blank channel or badly noised channels. Specifically, there are three targets in this image, i.e. "#1 Soil", "#2 Tree" and "#3 Water" respectively.

Data in the ENVI format with 156 channels: Data_Envi.zip (1.47Mb)

Data in the Matlab format with 156 channels: Data_Matlab.rar (3.41Mb)

Ground Truth: GroundTruth.zip (275Kb) includes GT: abundances and GT: endmembers.

Samson and its ground truths

(8) Jasper Ridge

Jasper Ridge is a popular hyperspectral dataset. There are 512x614 pixels in it. Each pixel is recorded at 224 channels ranging from 380 nm to 2500 nm. The spectral resolution is up to 9.46nm. Since this hyperspectral image is too complex to get the ground truth, we consider a sub image of 100x100 pixels. The first pixel starts from the (105,269)-th pixel in the original image. After removing the channels 1-3, 108-112, 154-166 and 220-224 (due to dense water vapor and atmospheric effects), we remain 198 channels (this is a common preprocess for HU analyses). There are four endmembers latent in this data: "#1 Road", "#2 Soil", "#3 Water" and "#4 Tree".

Data in the ENVI format with 198 channels: jasperRidge2_R198.zip (2.87Mb)

Data in the ENVI format with 224 channels: jasperRidge2_F224.zip (2.98Mb)

Data in the Matlab format with 198 channels: jasperRidge2_R198.mat (2.84Mb)

Data in the Matlab format with 224 channels: jasperRidge2_F224.mat (3.01Mb)

Ground Truth: GroundTruth.zip (364Kb) includes GT: abundances and GT: endmembers.

Jasper Ridge and its ground truth

(9) Urban

Urban is one of the most widely used hyperspectral data used in the hyperspectral unmixing study. There are 307x307 pixels, each of which corresponds to a 2x2 m² area. In this image, there are 210 wavelengths ranging from 400 nm to 2500 nm, resulting in a spectral resolution of 10 nm. After the channels 1-4, 76, 87, 101-111, 136-153 and 198-210 are removed (due to dense water vapor and atmospheric effects), we remain 162 channels (this is a common preprocess for hyperspectral unmixing analyses). There are three versions of ground truth, which contain 4, 5 and 6 endmembers respectively, which are introduced in the ground truth.

Data in the ENVI format with 221 channels: Urban_F210.zip (19.5Mb)

Data in the Matlab format with 162 channels: Urban_R162.mat (16.9Mb)

Data in the Matlab format with 221 channels: Urban_F210.mat (21.8Mb)

Ground Truth: three versions, including 4, 5 and 6 endmembers respectively.

4 endmembers version: GroundTruth (3.7Mb ). The 4 endmembers are "#1 Asphalt", "#2 Grass", "#3 Tree" and "#4 Roof" respectively.

5 endmembers version: GroundTruth (3.65Mb). The 5 endmembers are "#1 Asphalt", "#2 Grass", "#3 Tree", "#4 Roof" and "#5 Dirt" respectively.

6 endmembers version: GroundTruth (3.92Mb). The 6 endmembers are "#1 Asphalt", "#2 Grass", "#3 Tree", "#4 Roof", "#5 Metal", and "6 Dirt" respectively.

Urban and its ground truths

(10) The China Dataset

The China dataset belongs to a farmland near the city of Yuncheng Jiangsu province in China, which was acquired on May 3, 2006, and April 23, 2007, respectively. This scene is mainly a combination of soil, river, tree, building, road and agricultural field. For this dataset, all changes related to the type of land cover and river. In addition, this dataset belongs to Hyperion sensors.

For downloading MATLAB China dataset please go to this website in "Hyperspectral Change Detection" section, paper entitle "Hyperspectral Change Detection: An Experimental Comparative Study".

Hyperspectral Change Datasets (USA and China )


(a)	(b)	(c)	(d)
(a) and (b) present false-colour composites of the original hyperspectral images acquired in 2006 and 2007 of the China data set , respectively; (c) binary change map ground truth ; (d) multiple change map ground truth.

(11) The USA Dataset

The USA dataset belongs to an irrigated agricultural field of Hermiston city in Umatilla County, Oregon, OR, the USA, which was acquired on May 1, 2004, and May 8, 2007, respectively. The land cover types are soil, irrigated fields, river, building, type of cultivated land and grassland. For this dataset, all changes related to the type of land cover and river. In addition, this dataset belongs to Hyperion sensors.

For downloading MATLAB USA dataset please go to this website in "Hyperspectral Change Detection" section, paper entitle "Hyperspectral Change Detection: An Experimental Comparative Study".

Paper

Hyperspectral Change Datasets (USA and China )


(a)	(b)	(c)	(d)
(a) and (b) False-colour composite of the original hyperspectral images acquired in 2004 and 2007 of the USA data set, respectively; (c) binary change map ground truth; (d) multiple change map ground truth.

(12) The Washington DC Mall

The figure here shows a simulated color IR view of an airborne hyperspectral data flight line over the Washington DC Mall provided with the permission of Spectral Information Technology Application Center of Virginia who was responsible for its collection. The sensor system used in this case measured pixel response in 210 bands in the 0.4 to 2.4 µm region of the visible and infrared spectrum. Bands in the 0.9 and 1.4 µm region where the atmosphere is opaque have been omitted from the data set, leaving 191 bands. The data set contains 1208 scan lines with 307 pixels in each scan line. It totals approximately 150 Megabytes. The image at left was made using bands 60, 27, and 17 for the red, green, and blue colors respectively.

The Washington DC Mall Part 1 (45.00 MB)

The Washington DC Mall Part 2 (40.30 MB)

The Washington DC Mall Wavelength Table (3.0 kB)

The Washington DC Mall Ground Truth (381.0 kB)

(a)
(b)
The (a) false color composite of the original hyperspectral images of the Washington DC Mall dataset and (b) is related ground truth.

(13) The Cooke City Dataset (Target Detection Dataset)

Recently, Khoshboresh-Masouleh and Hasanlou (2020) introduced a revised replacement signal model for improving hyperspectral sub-pixel target detection in multiple target signatures. In this paper, the hyperspectral image with coverage areas of approximately 2.0 km² in Cooke City town, MT, USA is used as the experimental data. It was provided via the HyMap airborne hyperspectral imaging sensor by Snyder et al., (2008). HyMap sensor flown at 1.4 km above the ground level. The Cooke City dataset is illustrated in Fig. 1a, where there are 280×800 pixels. Each pixel is observed at 126 bands covering the electromagnetic spectra from 0.45 to 2.48 μm and the ground sample distance is 3.0 m. In this study, the three vehicles (Fig. 1b- d) and four fabric panels (Fig. 1e-h) in different sizes and colors were selected as targets. The reference spectra of the vehicles and the fabric panels were measured by an Analytical Spectral Devices, Inc. FieldSpec Pro, and a Cary 500 spectrophotometer, respectively. The vehicles were occupied at most a few pixels, but fabric panels F1 and F2 were nearly a full pixel while fabric panels F3 and F4 were occupied less than a pixel. The public open-source dataset used to support this study is available at http://dirsapps.cis.rit.edu/blindtest/.

For more information, please see the related paper (Khoshboresh-Masouleh and Hasanlou, 2020).

The Original Images (93.00MB)

The image dataset part one (55.00 MB)

The image dataset part two (50.00 MB)

The targets (9.60 MB)

Original Data

Cooke City dataset with target locations identified by the color squares (left). Plots showing the reference spectra of targets (right). The greyscale image (a), and the targets (b-h).

Hypercube display

(14) DLR HySU (HyperSpectral Unmixing) dataset

The DLR HySU (HyperSpectral Unmixing) dataset provides a publicly available benchmark to assess the performance of spectral unmixing algorithms. The dataset consists of airborne data acquired by a HySpex imaging spectrometer and a 3K RGB camera system over DLR premises at Oberpfaffenhofen, complemented by in-situ spectra recorded with an SVC field spectrometer. Synthetic targets of five different materials (bitumen, red metal, blue fabric, red fabric and green fabric) and sizes (0.25, 0.5, 1, 2 and 3 m) are deployed over a homogeneous background within the survey area in order to simulate various mixing scenarios. For a straightforward validation of unmixing algorithms, spectral libraries of the five target materials and the areas of all targets are also provided. The DLR HySU benchmark dataset is especially designed to test the main steps of spectral unmixing, namely dimensionality estimation, endmember extraction with and without pure pixel assumption, and abundance estimation. Other applications that can be investigated with DLR HySU include target detection, denoising, and super-resolution methods.

Dataset: DLR HySU (.zip, 5 MB)

DLR HySU dataset

Ocean Datasets

(1) Sea Surface Salinity Modeling

Level 3-monthly average SMAP SSS (BETA: version 2.0, validated release) data were obtained from RSS from April 2015 to April 2017. Sea surface parameters data sets namely sea surface temperature (SST), precipitation (P), evaporation (E), chlorophyll-a concentration (chl-a), geostrophic currents (u_cur, v_cur), brightness temperatures (T_b,4, T_b,h and T_b,v) and wind speed (u_ws, v_ws), longitude (lon), latitude (lat) and Julian date (JD) were also extracted for the same period and introduced as the inputs. All the variables have been rescaled to sustain a similar spatial and temporal resolution. Overall, 2328 samples were considered as total data which were randomly divided into two subsets of 70% and 30% for training and test data sets, respectively.

For more information, please see the related paper (Rajabi-Kiasari and Hasanlou, 2020).

The Train Target (6 KB)

The Train Inputs (96 KB)

The Test Target (3 KB)

The Test Inputs (43 KB)

The Label Target (1 KB)

The Label Inputs (1 KB)

Study area (Persian Gulf).

Pearson's correlation coefficients (r) between the utilized features.

Box plots of the z-scores registered by the boruta feature selection algorithm to identify the most significant predictors. Blue corresponds to the shadow inputs while the green represents the z-score distributions of confirmed inputs with notably large importance.

Reference

http://www.ehu.eus/ccwintco/index.php?title=Hyperspectral_Remote_Sensing_Scenes
http://www.escience.cn/people/feiyunZHU/Dataset_GT.html
https://earthexplorer.usgs.gov
https://www.dlr.de/eoc/en/desktopdefault.aspx/tabid-12760/22294_read-73262/
D. Snyder, J. Kerekes, I. Fairweather, R. Crabtree, J. Shive, and S. Hager, "Development of a Web-based Application to Evaluate Target Finding Algorithms," Proceedings of the 2008 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), vol. 2, pp. 915-918, Boston, MA, 2008, https://doi.org/10.1109/IGARSS.2008.4779144.
Gerg, I. (2008) ‘An evaluation of three endmember extraction algorithms: ATGP, ICA-EEA, and VCA'. The Pennsylvania State University.
Chang, C.-I. (2013) ‘Hyperspectral Imaging: Techniques for Spectral Detection and Classification'. Springer Science & Business Media.


(a)	(b)	(c)
Iran North Sentinel-2 Change Detection Datasets, (a) 2017 data, (b) 2021 dada, and (c) ground truth.