Cartosat-1-DEM-Mosaic

Name of the program: DEMOverlap_mosaic

Title of the manuscript: A novel framework for seamless mosaic of Cartosat-1 DEM scenes

Author details:

1. Rajeshreddy Datla, Scientist-SE, Signal processing division, Advanced Data Processing Research Institute (ADRIN), Indian Space Research Organization (ISRO), Department of Space, Government of India, Akbar Road, Tarbund, Manovikas Nagar, Secunderabad 500009, India Telephone number:+919704143731 E-mail:rajesh@adrin.res.in

2. Dr. C Krishna Mohan, Professor, Room No. 103, Block - A Department of Computer Science and Engineering Indian Institute of Technology Hyderabad Kandi, Sangareddy - 502285, Telangana (State), India.

Instructions to use this software:

A. Software with source code:

The complete software with source code is provided in the folder "DEMOverlap_mosaic". It consists of THREE sub-folders, namely,

1. DEMFilterLib: It contains separate functions for the DEM mosaic, namely, Mosaic_Average, Mosaic_Conventional, and Mosaic_Proposed, that correspond to average method, feathering-based blend method and Proposed method, respectively.

   1.1) Mosaic_Average function: It creates an virtual grid based on the extents of the input DEM scenes and assigns the elevation value to each cell of the grid, by computing the average of the elevation values in the overlapping regions wherever a cell contains more than one elevation value. Finally, gdal_translate is used to crop the result to ROI extent.

   1.2) Mosaic_Conventional function: It is basically feathering-based blend method. This method considers two inputs at a time for the mosaic, according to its nature of the methodology (scalability property). It collects all the input DEM scene filenames in a list and iterates by cascading the output generated from the first two inputs with the third input for the mosaic process. Then, considering the resulted output with fourth input for the mosaic and so on. It computes the overlapping region and the valid-data(non-NODATA) cells along the scan and pixel direction for every iteration. It derives a weighting mechanism based on the number of valid-data cells and uses these weights in the mosaic process to compute a new elevation value. It does not use meta-data of the input scenes.

   1.3) Mosaic_Proposed: Cartosat-1 mission has a unique path-row referencing scheme. This method makes use of this referencing-scheme which is available in the form of meta-data of each input DEM scene. It arranges the input scenes in a hierarchical manner and establishes a sequence. This method uses this sequence during mosaic process and creates a seamless mosaic. Hoever, It also computes the overlapping region and the valid-data(non-NODATA) cells along the scan and pixel direction for every iteration. It derives a weighting mechanism based on the number of valid-data cells and uses these weights in the mosaic process to compute a new elevation value.

2. DEMFIlterRef: It has the main function to invoke the mosaic methods.

3. in_conv: It reads the DEM in geotiff format into binary form along with the required meta-data population.

All the three mosaic methods use "gdal_translate" command to crop the mosaic output to an ROI extent. In this work, ROI extent is 1 degree x 1 degree (i.e., 110x 110 sq.km approximately).

B. Folder sturcture of input and output data:

Sample input and output files are located in the Testdata.zip and Mosaic_Output_Files.zip, respectivley. The file name of an input DEM scene has the following naming convention: P5008536_1P_20061202_556295 where, P5 represents Sensor name followed by 6-digit orbit number,1P represents the payload followed by date in YYYYMMDD format; followed by path(3-digit) and row(3-digit); Input files (DEMs) must be placed in a folder with a tile name, e.g., 64G_E81_N21 and place it under the inputfolder. This inputfolder and the outputfolder must be passed as the first and second arguments, respectively followed by the option which indicates mosaic method as explained in section C.

The format of the input are in Geotiff and the output files generated by this software are also saved in Geotiff. In this repo, the output files provided in Mosaic_Output_Files.zip are in the png format to have glimpse of outputs and to consume less space.

C. Usage of this software: Create a c++ source project using IDE (e.g., Microsoft visual studio 2010, Netbeans) Load the .sln file (solution file) from the IDE. First build the DEMFilterLib.cpp to generate DEMFilterLib.lib and use this lib as dependency while building DEMFIlterRef.cpp. The two files are supposed to get generated are: DEMFilterLib.lib and DEMFIlterRef.exe ,after building the corresponding source cpp files. Include the paths of the header files and the lib files (DEMFilterLib.lib and in_convNew.lib) in the configuration.

Software can be executed through command prompt also, apart from the IDE, as mentioned below:

Usage: DEMFIlterRef {inputfolder} {outputfolder} {1 | 2| 3}

inputfolder: Location of the parent folder of the tile folder (e.g., 64G_E81_N21). Tile folder contains input DEM scenes. output folder: It must have read/write permission. It must be different from the inputfolder. The DEM mosaic outputs will be created in this folder. The ouputs with the following names, namely, 64G_E81_N21_average.tif, 64G_E81_N21_conventional.tif, and 64G_E81_N21_proposed.tif will be created by the mosaic process using average method, conventional feathering-based blend method, and the proposed method, respectively. {1|2|3}: option 1 for Proposed method; 2 for average method ;3 for conventional feathering-based blend method. This software is compiled and executed in Microsoft visual studio 2010. This software is implemented with an intension to include these methods as static library for better reusage.