Atmospheric Correction and Calculation of some physical indices of - - PowerPoint PPT Presentation

Atmospheric Correction and Calculation of some physical indices of VNREDSat-1A’s images

Supervisor:

• Prof. Phạm Văn Cự

Department of Geography, Hanoi University of Science Vietnam National University. 334 Nguyen Trai Street, Hanoi City, Vietnam. Student: Phạm Thị Thanh Hoa Intake 2014-2016 Academic year: M1

Internship Report - PHAM Thi Thanh Hoa

1

Outline

 1. Principle of Radiometric Correction 1.1. Conversion of Raw Image to Spectral Radiance Images. 1.2. Correction of Spectral Radiance to the Reflectance at the TOA and to the Surface Reflectance.  2. Calculation of Some Physical Indices of VNREDSat-1A Images  3. Preprocessing Methodology  4. Conclusion

Internship Report - PHAM Thi Thanh Hoa

2

• 1. Principle of Radiometric Correction

 L = Ls + Lp  Where: L: The value of digital number (DN) of a signal L Ls: Direct reflect Radiance of object Lp: “Noise” Signal.  Radiometric correction of satellite images is about to remove the “Noise” signal Lp. By considering the surface of object is Lambertian surface (smooth and insignificantly different in elevation), the process will be done by following steps:

• Radiometric correction at the receiver of the

satellite.

• Atmospheric correction

Internship Report - PHAM Thi Thanh Hoa

3

• 1. Principle of Radiometric Correction

 Four levels of images are:  (1) Raw Image including digital number of pixels

• f objects.

 (2) Result of radiometric correction of receiver: converting images from DN (after receiver) spectral radiance L (before receiver) (which is presented as R on the Figure 2).  (3) Result of converting R images to the reflectance at top of the atmosphere ρTOA.  (4) Result of converting images of reflectance at top of atmosphere to the surface.

Internship Report - PHAM Thi Thanh Hoa

4

1.1. Conversion of Raw Images to Spectral Radiance Images

 - Formula for VNREDSat-1A Correction: R ≡ Lλ = (DN / GAIN) + Bias (1) Where Gain and Bias is linearly changing parameters of the receiver after being accredited. For VNREDSat – 1A, Bias = 0.  Processing Steps:

Internship Report - PHAM Thi Thanh Hoa

5

1.1. Conversion of Raw Images to Spectral Radiance Images

 Practical case for Hue:

Parameters VNREDSat-1A (2013) Image Code Number V130514_034239 Acquisition Date 14/05/2013 Processing Level 1A Satellite Angle (Degree) Incident angle: 20.34 Azimuth: 358.93 Solar Angle (Degree) Azimuth: 80.98 Sun elevation: 74.8 Multi-Spectral Band Spectral Order XS1 XS2 XS3 NIR VNREDSat-1 (2013) Absolute Gain Value (W/m2/sr/µm) 1.63825 1.62130 1.84789 2.51121 Spatial Resolution (m) 10 10 10 10

Internship Report - PHAM Thi Thanh Hoa

6

1.1. Conversion of Raw Images to Spectral Radiance Images

 Practical Case for Hue: Composite RGB DN Image (Left) and Spectral Radiance image (Right) in May 14th, 2013. Co

Internship Report - PHAM Thi Thanh Hoa

7

1.2. Correction of Spectral Radiance to the Reflectance at the Top of the atmosphere and to the Surface Reflectance

We consider ground as a Lambertian surface, then: (3) Where:

 - Diffuse spectral reflectance;

v



• View path transmittance from the Earth to the receiver;

S



• View path transmittance from the Sun to the Earth;

ES - Solar spectral radiation; dES – Astronomical distance between the Sun and the Earth;

S



• Solar azimuth angle.

we obtain the reflectance at the TOA:

) ( . . .

2 s S ES S TOA

Cos E d L    

(4)

From formula (4), we acknowledge that this obtained reflectance haven’t calibrated:

• The effect of atmosphere.
• The effect of topography
• The effect of azimuth of the receiver on the satellite.

Internship Report - PHAM Thi Thanh Hoa

8

1.2. Correction of Spectral Radiance to the Reflectance at the Top of the atmosphere and to the Surface Reflectance

 Processing Steps:

Internship Report - PHAM Thi Thanh Hoa

9

1.2. Correction of Spectral Radiance to the Reflectance at the Top of the atmosphere and to the Surface Reflectance

 Practical Case for Hue: Raw Image (Left), Reflectance at TOA Image (Middle) and Reflectance at the Surface (Right).

Internship Report - PHAM Thi Thanh Hoa

10

1.2. Correction of Spectral Radiance to the Reflectance at the Top of the atmosphere and to the Surface Reflectance

 Statistical Assessment: Vegetation Spectral Profile at surface (left) and at TOA (right):

VNREDSat-1 Image in May 14th, 2013 Statistical Result DN 𝝇𝑼𝑷𝑩 𝝇𝑻𝑮 Mean Variance CV 129.2872 1137.603 0.260879 0.148704 0.001546 0.256666 0.087790 0.001546 0.444793

Internship Report - PHAM Thi Thanh Hoa

11

• 2. Calculation of some Physical Indices
• f VNREDSat – 1A Images.

 Vegetation indices can be calculated directly from remote sensing images, and then are used for various purposes like vegetation-covering assessment, biomass assessment, crop monitoring, etc.  Some Vegetation Indices:  𝑂𝐸𝑊𝐽 = 𝑂𝐽𝑆−𝑆

𝑂𝐽𝑆+𝑆

 𝑇𝐵𝑊𝐽 =

𝑂𝐽𝑆−𝑆 𝑂𝐽𝑆+𝑆+𝑀 (1 + 𝑀)

 𝑄𝑊𝐽 = 𝑂𝐽𝑆−𝑏𝑆−𝑐

𝑏2+1

 Where: NIR: Near infrared spectral bands Red: Red spectral bands. a, b, L: parameters

Internship Report - PHAM Thi Thanh Hoa

12

• 2. Calculation of some Physical Indices
• f VNREDSat – 1A Images.

 NDVI: SAVI:

SAVI Minimum: -0.2339 Maximum: 0.7022 Mean: 0.2582 Std Dev: 0.1327 NDVI Minimum: -0.8127 Maximum: 0.9145 Mean: 0.4598 Std Dev: 0.2661

Internship Report - PHAM Thi Thanh Hoa

13

• 2. Calculation of some Physical Indices
• f VNREDSat – 1A Images.

 PVI:  The values of parameters a and b can be deduced from the linearity of brightness of soil in band 3 (red) and band 4 (nir)

PVI Minimum: -0.0423 Maximum: 0.4167 Mean: 0.1347 Std Dev: 0.0601

Internship Report - PHAM Thi Thanh Hoa

14

• 3. Preprocessing Methodology

 ENVI tool named “VNREDSat Calibration” under the function “Calibration Utilities” of Basic Tools.

Internship Report - PHAM Thi Thanh Hoa

15

Conclusion

 During this internship, there are two main matters:

• Calibrating DN images due to the effect of receiver of the satellite and the

atmosphere.

• Calculating some vegetation indices: NDVI, SAVI, and PVI.

 The images which are corrected and treated by DOS 4 obtain higher quality since their variant indices are close to zero. Hence, the final image is more reliable when we need to use it for identifying threshold of objects on the surface, and managing volatility.  IDL and ENVI are highly potential to dig deeper and develop more in the future.

Internship Report - PHAM Thi Thanh Hoa

16

THANK YOU FOR YOUR ATTENTION!

Internship Report - PHAM Thi Thanh Hoa

17

THANK YOU FOR YOUR ATTENTION!

Internship Report - PHAM Thi Thanh Hoa

18