MartinBorgt / jarvis

jarvis

Goal

Detect and describe the shape of following objects from the high resolution satellite image

1. Buildings - large building, residential, non-residential, fuel storage facility, fortified building
2. Misc. Manmade structures
3. Road
4. Track - poor/dirt/cart track, footpath/trail
5. Trees - woodland, hedgerows, groups of trees, standalone trees
6. Crops - contour ploughing/cropland, grain (wheat) crops, row (potatoes, turnips) crops
7. Waterway
8. Standing water
9. Vehicle Large - large vehicle (e.g. lorry, truck,bus), logistics vehicle
10. Vehicle Small - small vehicle (car, van), motorbike

Image file information

WorldView-3 products are delivered to the customer as relative radiometrically corrected image pixels. Their values are a function of how much spectral radiance enters the telescope aperture and the instrument conversion of that radiation into a digital signal.

Data sample counts

• Total images: 450 * 4
• Training data: 25
  • 6010_1_2, 6010_4_2, 6010_4_4
  • 6040_1_0, 6040_1_3, 6040_2_2, 6040_4_4
  • 6060_2_3
  • 6070_2_3
  • 6090_2_0
  • 6100_1_3, 6100_2_2, 6100_2_3
  • 6110_1_2, 6110_3_1, 6110_4_0
  • 6120_2_0, 6120_2_2
  • 6140_1_2, 6140_3_1
  • 6150_2_3
  • 6160_2_1
  • 6170_0_4, 6170_2_4, 6170_4_1

Image files

File	Size (Rows x Cols)	Bands	Resolution	Color depth
xxx_x_x	3349 x 3396	Red, Green, Blue	0.31m	11 bits
xxx_x_x_A	134 x 136	8 SWIR Bands	7.5m	14 bits
xxx_x_x_M	837 x 849	8 Multispectral Bands	1.24m	11 bits
xxx_x_x_P	3348 x 3396	Panchromatic, greyscale, single band	0.31m	11 bits

Sensor bands information

Band	Type	Wavelength
Panchromatic	Panchromatic	450 - 800 nm
Coastal	Multispectral	400 - 450 nm
Blue	Multispectral	450 - 510 nm
Green	Multispectral	510 - 580 nm
Yellow	Multispectral	585 - 625 nm
Red	Multispectral	630 - 690 nm
Red Edge	Multispectral	705 - 745 nm
Near-IR1	Multispectral	770 - 895 nm
Near-IR2	Multispectral	860 - 1040 nm
SWIR-1	SWIR	1195 - 1225 nm
SWIR-2	SWIR	1550 - 1590 nm
SWIR-3	SWIR	1640 - 1680 nm
SWIR-4	SWIR	1710 - 1750 nm
SWIR-5	SWIR	2145 - 2185 nm
SWIR-6	SWIR	2185 - 2225 nm
SWIR-7	SWIR	2235 - 2285 nm
SWIR-8	SWIR	2295 - 2365 nm

Using Docker

Docker is a container engine that stabilizes the runtime environment. A Dockerfile is included in the project. And the image has been pushed to DockerHub. https://cloud.docker.com/swarm/junjchen90/repository/docker/junjchen90/jarvis-machine/general

After installed Docker, run the following command in the cloned repo's directory:

docker run -ti --name app -v `pwd`:/app junjchen90/jarvis:latest

Is will start a bash and you're in the container. (The command pulls images from DockerHub, starts it as a container named "app" and mount the current working directory to container's /app directory)

launch jupyter

jupyter notebook --ip 0.0.0.0 --allow-root --NotebookApp.iopub_data_rate_limit=10000000000

Classification methods

References

Abburu S, Golla S B. Satellite image classification methods and techniques: A review[J]. International journal of computer applications, 2015, 119(8).