Data Paper Zone II Versions EN1 Vol 4 (2) 2019
Value-added Landsat image data of Hainan Island from 1970 to 2017
>>
： 2019 - 01 - 14
： 2019 - 02 - 01
： 2019 - 05 - 20
587 4 0
Abstract & Keywords
Abstract: Since its establishment as Hainan Special Economic Zone, Hainan Island has achieved great development in urbanization and great changes have taken place in the natural environment. In recent years, many policies have been put forward for the construction of Hainan Island, such like Hainan International Tourism Island Construction, Maritime Silk Road and the ASEAN strategy. In this situation, the long time-sequence, high-quality satellite remote sensing data sets provide important reference to Urban Construction, Environmental Protection, Agricultural Planning, and Ecotourism. In order to learn the changes of Hainan Island, in macroscope perspective, and understand the evolution of its natural resources, since the establishment of the special economic zone, this study selected the cloudless or less cloudy Landsat series satellite remote sensing images from 1970 to 2017. Through the deep processing service system, we obtained five images which can cover the entire island of Hainan in 1970, 1988, 1998, 2008 and 2017. And the main jobs included image registration, projection conversion, band fusion and image mosaicking. With less cloud influence, this product has a large time span and can be used for scientifically analyzing the temporal and spatial changes of Hainan Island over the past 40 years, as well as in industrial application monitoring and government macro decision.
Keywords: Landsat; Hainan Island; value-added products; band fusion; remote sensing service
Dataset Profile
 English title Value-added Landsat image data of Hainan Island from 1970 to 2017 Corresponding author Cheng Bo (chengbo@radi.ac.cn) Data author(s) Liang chenbin, Cheng bo, He guojin Time range 1970-2017 Geographical scope Hainan Island (1807’N – 2007’N, 10830’E – 11110’E) Spatial resolution 60 m, 30 m Data volume 1.00G Data format *.tif (GeoTIFF, 8 bit float) Data service system Source(s) of funding Major Science and Technology Program of Hainan Province (ZDKJ2016021); National Key Research and Development Program of China (2017YFC1405600); Strategic Priority Research Program of the Chinese Academy of Sciences (XDA190090300); National Natural Science Foundation of China (61731022). Dataset/Database composition This dataset covers two parts of data, including Landsat Hainan Island mosaic product in GeoTIFF format from 1970 to 2017 and image lists in XLS format. In addition, a data sample is provided to show the band fusion data and image mosaic data.
1.   Introduction
Hainan has a land area of 35,300 km2, a coastline of 1,823 km and a sea area of 2 million km in the South China Sea. It is located at the southernmost tip of China and is the frontier of the world and the main channel for East-West exchanges. As the largest special economic zone in China, Hainan has a special location advantage and strategic position1.. Since the establishment of the special economic zone in Hainan Province in 1988, the nature environment and urban construction of Hainan Island have undergone tremendous changes. It is necessary to quickly grasping the dynamic information of nature source and human activities development, which is in favor of mastering the natural, agricultural and tourism resources of Hainan Province, strengthening resource and environmental management, promoting the sharing of information resources, improving the level of scientific and macroscopic decision management, achieving coordinated and sustainable development of resources, environment, economy and society in Hainan Province and promoting the construction of Hainan International Tourism Island, the 21st Century Maritime Silk Road and the Hainan Free Trade Zone.
Satellite remote sensing technology has unique advantages for conducting macro-research on resource and environment, which must rely on high-quality satellite remote sensing basic data2.. With the development of remote sensing application technology, the demand for data products gradually changed from basic to value-added type. In order to provide standardized value-added products more efficiently and quickly, to meet the user's rapid application requirements and enable the users to put more energy into deeper research and practical applications3., He Guojin et al4.5. proposed Ready To Use (RTU), the concept of satellite remote sensing products, which is to process satellite remote sensing data into products that users can directly use. Value-added produces eliminate the need for users to make cumbersome pre-use before use and improve the efficiency of the work.
Based on the Landsat series remote sensing data, this paper used the remote sensing data deep processing method to obtain 5 value-added products covering Hainan Island from 1970 to 2017. After a series of pre-processing, such as image registration, projection conversion, etc., and the image mosaic, we got value-added produces covering Hainan Island. On the one hand, we have adopted false color synthesis. Under this color synthesis, vegetation and buildings have obvious and distinct characteristics, which is beneficial to the relevant departments to monitor the natural environment and urban construction of the island. On the other hand, these data products These products have completed the pre-processing of data, with a large time span and a wide spatial range, which has laid a foundation for the further research of other researchers.
2. Data collection and processing
2.1   Data sources
Since 1972, Landsat Series Satellites has launched eight satellites, which are the most widely used and most effective remote sensing information sources in the world. Landsat series remote sensing data are often used to detect the earth's resources and environment, to investigate underground mineral, marine and groundwater resources, to supervise agriculture, forestry, animal husbandry and water resources, to forecast crop yields, to study the growth and landform of natural plants, to investigate and forecast various serious natural disasters (such as earthquakes) and environmental pollution, to take pictures of various targets, and to draw thematic maps (such as geological maps, geomorphological maps, hydrological maps) etc. Landsat has many advantages such as high spatial resolution, long time span, open use, etc.6., which has significant advantages for long-term sequence monitoring research.
The data sources of this dataset includes Landsat MSS/TM/OLI of Hainan Island in 1970, 1988, 1998, 2008 and 2017 with a few clouds and strips, a total of 25 images, which were downloaded by the Sanya Receiving Station of the Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences and the US Geological Survey (http://glovis.usgs.gov/). Meanwhile, in order to avoid cloud and fog interference, remote sensing data of adjacent years are used as supplement and replacement. The information of remote sensing image used is shown in Table 1. According to the imaging time, the Landsat image was mosaic and five remote sensing images covering Hainan Island from 1970 to 2017 were obtained.
Table 1   Data List of Remote Sensing Images in Hainan Island
11973-02-25134-047Landsat-1MSS60 m
21973-12-28134-046Landsat-1MSS60 m
31974-01-13132-046Landsat-1MSS60 m
41977-01-25133-046Landsat-2MSS60 m
51978-12-18132-046Landsat-3MSS60 m
61980-01-01133-047Landsat-3MSS60 m
71988-01-07124-047Landsat-5TM30 m
81988-06-08123-046Landsat-5TM30 m
91988-06-08123-047Landsat-5TM30 m
101988-06-15124-046Landsat-5TM30 m
111998-01-02124-047Landsat-5TM30 m
121998-01-11123-047Landsat-5TM30 m
131998-04-24124-046Landsat-5TM30 m
141998-08-23123-046Landsat-5TM30 m
151998-09-22125-047Landsat-5TM30 m
162008-08-25124-047Landsat-5TM30 m
172009-10-08123-046Landsat-5TM30 m
182009-10-24123-047Landsat-5TM30 m
192010-03-24124-046Landsat-5TM30 m
202015-11-17124-046Landsat-8OLI30 m
212015-11-17124-047Landsat-8OLI30 m
222016-08-08123-046Landsat-8OLI30 m
232017-03-02125-046Landsat-8OLI30 m
242017-04-21123-047Landsat-8OLI30 m
252017-04-26125-047Landsat-8OLI30 m
2.2   Data processing
The following pre-processing work was performed on the Landsat series of remote sensing data: projection conversion, image registration, band fusion, and image mosaic. The projection coordinates of this dataset are unified to UTM 49N, the coordinate system is WGS84, and the short-wave infrared, near-infrared, and red-light bands are used for band fusion. Finally, the image mosaic is used to obtain the data products covering the whole island of Hainan Island (as shown in Figure1).

Figure 1   Landsat Series Satellite Remote Sensing Image Data Processing Flow Chart
2.2.1   Projection conversion
Because the data is acquired in different ways, and the projection of the Landsat data is not uniform, the image projection coordinates are unified into UTM 49N for the smooth operation of the image mosaic. This paper used ENVI software to use Map Based method to project and convert the data of each band in the data list.
2.2.2   Image registration
Image registration refers to the process of fitting the optimal spatial transformation using overlapping regions of two or more images (reference and image to be registered) acquired by the same or different dates, the same sensor or different sensors.If image registration is defined from a mathematical point of view, the registration relationship between images can be mapped to the following equation7.:
（1）
As a digital image, $${\mathrm{I}}_{1}\left(\mathrm{x},\mathrm{y}\right)$$ and $${\mathrm{I}}_{2}\left(\mathrm{x},\mathrm{y}\right)$$ are respectively a reference image and a two-dimensional array matrix of images to be registered, which are used to characterize the gray level of the image. Where $$\left(\mathrm{x},\mathrm{y}\right)$$ is the spatial coordinate; $$\mathrm{f}$$ is the two-dimensional spatial coordinate transformation, such as affine transformation, similar transformation, etc., and its parameter matrix is determined by the similarity measure between the two images; $$\mathrm{g}$$ is the transformation of image brightness or saturation, and can also be achieved by adjusting the sensor hardware parameters.
In order to eliminate the geometric errors between images to be mosaic, before mosaic, a scene image is usually selected as the reference image, and the geometric rectification is carried out by bilinear interpolation method using ENVI software.
2.2.3   Band fusion
False color synthesis is carried out in short-wave infrared, near-infrared and red bands, which clearly reflects the characteristics of various objects. This color synthesis method makes the water and its boundary very clear, makes it easy to distinguish the river channel from the road, and makes it suitable for the investigation of the coast and its beach. At the same time, the structure of the neighborhood within the residential area is clear, but its outer boundary is not very clear. In addition, the band combination has the advantage of highlighting vegetation with standard false color images, but the color is not very saturated.
This paper used the Geospatial Data Abstraction Library (GDAL) to batch process the data in the data list to achieve band fusion. GDAL was first developed by Frank Warmerdam in 1988. It developed an abstract data model for expressing various data formats and provided related command-line tools for data conversion and processing services. GDAL provides support for most raster data and vector data, but the reading and writing of certain data needs to be processed according to certain rules8.9. .
2.2.4   Image mosaic
When performing image mosaic, minimize the amount of clouding of the mosaic by modifying the mosaic. In addition, due to different imaging conditions, the color difference between the images used is relatively serious, and the images need to be evenly colored to reduce the chromatic aberration between the images, so that the overall color tone can be consistent.
This article uses histogram matching to achieve uniformity, which is a method often used in color consistency processing between images. The method takes the histogram of the reference image as a standard, and makes the histogram of the image to be evenly colored the same or similar to it, so that the two images have similar colors and brightness10.. It fully reflects the grayscale distribution of the image. The histogram correction makes the transformed histogram similar to the given reference image histogram. Since the image can be digitized, its histogram can be represented by a discrete function, which can be used to count the relationship between the gray value in the digital image and the number of occurrences11..
If the digital image is represented by$$\mathrm{ }\mathrm{g}\left(\mathrm{x},\mathrm{y}\right)$$, the dimensions of the image row and column are $$\mathrm{M}$$ and $$\mathrm{N}$$, respectively. A normalized histogram can be defined as:
（2）
Where $${\mathrm{r}}_{\mathrm{k}}$$is the kth gray value; $$\mathrm{P}\left({\mathrm{r}}_{\mathrm{k}}\right)$$is the estimated probability that the gray level $${\mathrm{r}}_{\mathrm{k}}$$ appears in the image; $${\mathrm{n}}_{\mathrm{k}}$$ is the total number of pixels of the gray level $${\mathrm{r}}_{\mathrm{k}}$$.
3.   Sample description
This dataset includes five value-added Landsat image data of Hainan Island from 1970 to 2017, which are stored in the “data set” folder, and each period of data is named in the corresponding year, with a total amount of 1.00 GB. The spatial resolution of the images in other years is 30 m, except that in 1970, the spatial resolution of the images is 60 M. The projection coordinate of each image is UTM 49N and the coordinate system is WGS1984. The value-added Landsat image data of Hainan Island in 2017 is shown in Figure 2.

Figure 2   The value-added Landsat image data of Hainan Island in 2017
4.   Quality control and assessment
4.1   Image registration accuracy
In terms of statistical error, the effect of image registration was comprehensively and deeply analyzed. The currently applicable method is the Root-mean-square Error (RMSE), which objectively evaluates the registration accuracy by calculating the RMSE between the corrected image and the reference image. In general, the smaller the RMSE value, the higher the accuracy of the algorithm. In the study of the direction of image change detection, the mathematical expression of RMSE is:
（3）
Where$$\left({\mathrm{x}}^{\mathrm{\text{'}}\mathrm{\text{'}}},{\mathrm{y}}^{\mathrm{\text{'}}\mathrm{\text{'}}}\right)$$is the pixel coordinate of the registered image.
The polynomial model is selected for geometric correction, and the bidirectional linear interpolation method is selected for fine registration. The overall accuracy is less than 0.5 pixels, and the upper and lower adjacent images are selected to show the edge accuracy (Figure 3). From Figure 3, it can be seen intuitively that rivers, roads and other linear objects have higher precision of joining edges.

Figure 3   Image Edge Graph
4.2   Uniform mosaic effect
Histogram matching is used for uniform coloring, which is based on the histogram of the reference image, and the histogram of the image to be evenly colored is the same or similar to it. Adjust the color and brightness of the image to finally balance the image in color and brightness, eliminating the chromatic aberration between different images, as shown in Figure 4. Comparing the images before and after the mosaic, it can be found that if the color mosaic is used, the effect of the edge between the images is good, the color transition at the edge is natural, and there is no obvious color difference.

Figure 4   Comparison of image effects before and after uniform mosaic
5.   Value and significance
This data set is a relatively complete Landsat mosaic product in Hainan Island for more than 40 years which has less cloudiness, high positioning accuracy and color balance, and does not require user pre-processing. The data set contains five mosaic products and image data lists, which can meet the needs of different users. It can serve the natural resources survey, ecological environment supervision, urban development planning, site selection of major projects, marine strategic channel monitoring and other fields in Hainan Province. It can also be used as key basic data for scientific research.
6.   Usage notes
Value-added Landsat image data of Hainan Island from 1970 to 2017 is a long-term mosaic product for Hainan Island. Except for the spatial resolution of images in 1970 is 60 m, the spatial resolution of all other years is 30 m, which can be used as the background data of Hainan Island. This dataset facilitates the study of the overall changes in Hainan Island since the establishment of the Hainan Special Economic Zone, and plays an important role in urban planning, farmland protection, vegetation detection, land cover and coastal zone monitoring. This data set will be supplemented regularly with new time series products to provide basic data support for the sustainable development of Hainan Province. The data format of this dataset is GeoTIFF, which includes three bands of short-wave infrared, near-infrared and red-light bands. This band combination has the advantages of clear water boundary, easy to highlight vegetation, and can directly reflect vegetation and city information. In addition, data products can be read and manipulated directly in remote sensing and GIS software such as ENVI, ArcGIS, PCI and QGIS for secondary production. The data list adopted is named as Value-added Landsat Image Data of Hainan Island List of Hainan Province in XLS format.
Acknowledgments
Thanks to the data support of Sanya Receiver Station, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences and USGS.
1.
Meng Guangyi, Yang Kaizhong, Zhu Fulin, et al. Hainan, China: From the Special Economic Zone to the Evolution of a Comprehensive Compound Free Trade Port[J]. Geographic research, 2018(12): 2363-2382.
2.
Xu Guanhua, Liu Qinhuo, Chen Liangxuan, et al. Remote Sensing and Sustainable Development in China: Opportunities and Challenges[J]. Journal of Remote Sensing. 2016, 20(5): 679-688.
3.
Wang Kunlong, Liu Dingsheng, Zhang Wenyi, et al. Automated analysis, design and implementation of satellite data value-added process[J]. Remote Sensing Technology and Application. 2005 (3): 355-360.
4.
He Guojin, Wang Lizhe, Ma Yan, et al. Earth observation big data processing: challenges and reflections[J]. Chinese Science Bulletin. 2014, 60(5-6): 470-478.
5.
HE G J, ZHANG Z M, JIAO W L, et al. Generation of ready to use (RTU) products over china based on Landsat series data[J]. Big Earth Data,2018, 2 (1): 56-64.
6.
Tang Dongmei, Fan Hui, Zhang Yao. Overview of timing change detection[J]. Journal of Earth Information Science, 2017, 19(8): 1069-1079.
7.
Xu Liyan. Remote Sensing Image Registration Method Based on Feature Points and Its Application[D]. Nanjing: Nanjing University of Science and Technology, 2012.
8.
Zhang Lina, Zhang Dongfang, Bai Yabin. Research on Fast Reading and Display Method of Remote Sensing Image Based on GDAL[J]. Journal of West Resources, 2014 (1): 188-190.
9.
Zhao Yan, Wang Siyuan, Bi Haijun, et al. Research on Key Technologies of Remote Sensing Image Browsing Based on GDAL[J]. Computer Engineering, 2012 (23): 15-18.
10.
Wang Xiaoli, Dai Huayang, Yu Tao, et al. Research on image stitching uniformity of UAV based on multi-resolution fusion[J]. Bulletin of Surveying and Mapping, 2013(6): 27-30.
11.
Wang Huibai, Zhang Penglou. Research on Relative Radiation Correction Method Based on Histogram Matching[J]. Digital User, 2014(23): 234.
Data citation
Liang C, Chen B & He G. Value-added Landsat image data of Hainan Island from 1970 to 2017. Science Data Bank, DOI: 10.11922/sciencedb.725 (2019).
Article and author information
Liang C, Chen B & He G. Value-added Landsat image data of Hainan Island from 1970 to 2017. China Scientific Data 4(2019). DOI: 10.11922/csdata.2019.0002.zh
Liang Chenbin
Data collection, value-added processing and data batch algorithm design.
Master's degree; Intelligent Processing of Remote Sensing Image and Deep Learning.
Cheng Bo
Design of remote sensing data value-added processing.