njurs / river_detection

River detection code

Matlab River Detection Code

Environment

MATLAB R2008a (or higher)

Add MATLAB Tools

Use 2DSpecTools for band pass filter.

Use OBNLMpackage for denoising.

Use DIPimage for path opening.

Install tools

2DSpecTools and OBNLMpackage tools have been offered on the GitHub.

Download DIPimage from http://www.diplib.org/download and install it.

Set Path

• Click set path in Matlab GUI.
• Add 2DSpecTools, OBNLMpackage, DIPimage tools, Matlab River Detection code folder to MATLAB environment.
• Save.

Open Matlab River Detection code folder in MATLAB

Change the path of DIPimage in pathopening.m

Run river detection code

Here, we provide run_river_detection.m to detect river in single image and run_batch_river_detection.m for batch detection.

• Open Matlab River Detection code folder in MATLAB.
• Open run_river_detection.m or run_batch_river_detection.m.
• Write the image path. test_sentinel2_image.tif is provided as test image.
• Set input 6 parameters.

parameters	description
	WV: WorldView image
	SPOT: SPOT image
sensor	Sentinel2: Sentinle-2 image after NDWI calculation
	Landsat: Landsat panchromatic image
	LandsatNDWILandsat: image after NDWI calculation
inverse	set 1 to convert dark rivers into bright rivers (e.g., for panchromatic image); set 0 to keep bright rivers (e.g., for NDWI images)
width	small river width for Gabor filter, default = 2
ppo_length	path opening length, default = 20
histCountThreshold	a pixel count threshold to stretch image pixel values, default = 1000
Smooth (optional)	used for denoise algorithm. Because denoise algorithm is too slow and scale dependent, is abandoned for large images, default = 0.7

You can add your customized sensor (image) easily. The only requirement is to reset the band pass frequency based on the spatial resolution of your input image.

• Click run.

If you see these commands in the command window, the river detection code is running successfully. You will get 3 processed images. test_sentinel2_image_bandpass_gabor_cpo20.tif is the final result.

Acknowledgements

Corresponding Author

• Kang Yang ([email protected], ph: 13814179324, School of Geography and Ocean Science, Nanjing University)
• Xin Lu ([email protected], School of Geography and Ocean Science, Nanjing University)
• Yao Lu ([email protected], School of Geography and Ocean Science, Nanjing University)

For more information, please see the paper

Yang, K. , Karlstrom, L., Smith, L.C., Li, M., 2017. Automated high resolution satellite image registration using supraglacial rivers on the Greenland Ice Sheet. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 10(3): 845-856.

Yang, K. , Li, M., Liu, Y., Cheng, L., Huang, Q., Chen, Y., 2015. River Detection in Remotely Sensed Imagery Using Gabor Filtering and Path Opening. Remote Sensing, 7(7): 8779-8802.

Yang, K. , Li, M., Liu, Y., Cheng, L., Duan, Y., Zhou, M., 2014. River Delineation from Remotely Sensed Imagery Using a Multi-Scale Classification Approach. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 7(12): 4726-4737.