A detailed course of land use classification--taking the image of high score as an example (I.)

1 Preface

I believe that as I have just contact with remote sensing computer people, suddenly switch to start to get remote sensing, at first, will certainly confused. Not to mention the production of land-use classifications. Here, I will use the most direct graphical process operation to give the children like me the first knowledge of remote sensing children's shoes to explain the whole process of making land use.
First, I'll use a flowchart to summarize the whole process of making land use. As shown in 1:

Figure 1 Land use classification flowchart
This process operation is suitable for general image making land use process. For high-score image data, however, some modifications are needed to make the land-use process. At the end of the article, I'll attach the flowchart of high-score image data to land use. Now, let's start the process of land-use production for high-score imagery data.
Note: In the tutorial, I have used envi5.3 in some places. If you want to get envi5.3, please leave your email in the comments, then I will send a link to share. This software sharing is only used as scientific research, if commercial, please use genuine, otherwise the consequences.

2 cropping

The image is cropped first, here I have two purposes for clipping the image:
1, reduce the next workload
2, so you can choose a higher resolution of the DEM (DEM may not overwrite the image, here I use SHP map to cut the image, you can determine the download of the DEM smaller)
The cropping steps are as follows:

Figure 2 Cropping tool

Figure 3 Selecting a cropped image

Figure 4 Result output
This allows the image to be cropped out, showing the image I cropped:

Figure 5 cropping a picture
In the same vein, all the images in the study area are cropped (there is a problem with the image stitching directly, see the positive Correction section for details).

3 Positive shot correction

Here we place the Ortho correction in front of the stitching because the result of our ortho correction needs to be an initial document for atmospheric correction. Here we need to explain that the flowchart we just started with can also be pre-calibrated, that is, there is no dependency between the ortho correction and the atmospheric correction. The production of high-score image data can be a special case.
There is also a point to note: The Ortho correction may not be able to image mosaic (splicing), unable to operate will pop up as shown in 2 (in fact, we have no problem with the image, 3-5 shows that two images are the same). The reason why we are stitching is because there may be more than one image in the study area, so if the ortho correction requires an image to be corrected, and then splicing. This will take a lot of time. Therefore, we often think of the first image splicing, and then the stitching of an image to be corrected. If you're not as well as I am, then you can do the correction before stitching.

Figure 6 Stitching failure Tips

Figure 7 image data Information

Figure 8 Image coordinate information

Figure 9 Spectral band Information of the image
I have not thought of the question at the moment. Here we do not struggle to find out why this problem arises, then first with the trouble point of the operation, first to correct the positive shot.
Because I have no control points, there is no reference to the raster image, so the RPC file is used for ortho correction. If you have an existing raster image or control point information, you can use the RPC orthorectification using Reference image Tool (envi5.3 version, here because I used the envi5.1, so gave up this way, but this way than I do the way the positive effect is excellent, it is recommended to use this kind of ortho correction method). Detailed procedures refer to the blog: Automatic acquisition of control points of the RPC Ortho Correction tool. The way I use the RPC orthorectification Workflow tool is as shown in the following procedure:

Figure 10 Selecting the RPC orthorectification Workflow tool

Figure 11 File Selection
Here the Dem file defaults to the 2010 900m resolution of the DEM, here, according to the actual situation, if there is a higher resolution of the DEM, select a high-resolution DEM effect is better. I've cropped the image here, so the image range is definitely smaller than the DEM range. Here I choose a dem with a 30m resolution.

Figure dem and cropping image
The middle white border is a two-piece cropped image that is not stitched together. Next, select Crop Images and dem for file selection.

Figure 13 cropping file selection
Click "Next". The RPC settings are followed.

Figure RPC Settings
Because I don't have a control point here, I'm not dealing with the settings for GPC. Make the settings for the next option directly.

Figure the Advanced option setting
Advanced settings, wherein the radiation correction algorithm is: The nearest neighbor method, linear interpolation, three convolution method, here we can choose three convolution method.

Figure statistics Option settings
Statistics set us here to select all.

Figure 17 radiometric correction Output Options

(Here's a comparison test tomorrow)

4 radiometric calibration and atmospheric correction

Before I start, I'll explain the radiometric calibration and atmospheric correction. Similarly, the article refers to the above reference to the text of the geometric correction, ortho correction, geometric registration, image registration, spatial registration, radiation calibration, atmospheric correction, the concept of radiation correction to explain the definitions.

4.1 Radiation Calibration Instructions

As many sources have said, different scholars have explained the different definitions of radiometric calibration:
① calibration is to transform the measured value of the sensor to the absolute brightness of the surface reflectance, surface temperature and other physical quantities relative value of the processing process (Zhao Ying, such as "the principle and method of remote sensing Application analysis")
② Remote Sensor Calibration is to establish the quantitative relationship between the digital quantization output value DN of remote sensing sensors and the radiation luminance values in the field of view (statement Peng, also the degree Niang explanation)
③ radiometric Calibration is the process of converting a voltage or digital value recorded by a sensor into absolute radiant brightness (Liang Shunlin "Quantitative Remote Sensing")
④ radiometric Calibration is the conversion of the recorded original DN value to the atmospheric outer surface reflectance , to eliminate the error caused by the sensor itself
Radiation calibration has many methods: laboratory calibration, on-star calibration, site calibration. See the ESRI China Community envi section, Dsbin sensor calibration http://bbs.esrichina-bj.cn/ESRI/viewthread.php?tid=56191
In other words: Convert the DN value into a formula for radiant brightness and atmospheric apparent reflectance:
1. Convert the initial DN value to radiant brightness, where lb is the radiant luminance value, in W/CM2.ΜM.SR (w/cm. Micron. Spherical degree), gain and bias are gain and deviation, the unit and radiation luminance values are the same, as can be seen, the radiation luminance and DN values are linear relations.
2. The radiant luminance value is converted to the atmospheric apparent reflectance, the lλ in the formula is the radiant luminance value, D is the diurnal distance of the astronomical unit, the esunλ is the solar apparent emissivity mean, and the θs is the solar altitude angle in degrees. In general, however, this part of the work does not require users to do their own, the relevant coefficients are included in the data header file or metadata.

4.2 Atmospheric Correction Instructions

The objective of atmospheric correction is to eliminate the influence of atmospheric and illumination factors on the reflection of ground objects, and to obtain real physical model parameters such as reflectivity, emissivity or surface temperature in a broad sense. It is used to eliminate the influence of water vapor, oxygen, carbon dioxide, methane and ozone in the atmosphere on the reflection of ground objects, and to eliminate the scattering of atmospheric molecules and aerosols. In most cases, the atmospheric correction is also the process of retrieving the true reflectivity of the object.
Under what circumstances do we need to do an atmospheric correction, or have we purchased or otherwise acquired an image for atmospheric correction? In layman's terms, if we need quantitative inversion or access to earth information, accurate recognition of objects and so on, we need to use the image to reflect the actual radiation on the sun, then we need to do atmospheric correction. The general purchase of the image, the document will be indicated by radiation correction, in fact, this radiation correction refers to the coarse radiation correction, just do the system atmospheric correction, and the meaning of the system geometry correction is the same.
Atmospheric correction and radiation schools are interrelated, in envi to do atmospheric correction premise is the need for radiation calibration results, the actual process is the image of the radiation brightness value, the final conversion to the surface reflectance.

4.3 Radiation correction Instructions

is the correction of systematic, random radiation distortion or distortion resulting from external factors, data acquisition and transmission systems, eliminating or correcting the process of image distortion caused by radiation errors. (including radiometric calibration and atmospheric correction) three relationships can be attributed to:

Fig. 18 Radiation correction, radiometric calibration, atmospheric correction diagram
OK, here's a thank you for the "Pour Me not" explanation. Next we continue our tutorial-radiometric correction.

4.4 radiometric Calibration

Then we begin to calibrate the radiation. First, the radiometric calibration is performed. Use radiometric correction->>radiometric calibration (radiometric correction) in ENVI. As shown in the following examples:

Fig. 19 Radiometric Calibration
The alert appears as shown when I select the selected image for the next step:

Figure 20 Warning
？？？ What does that mean? I have been able to solve this problem by reviewing the relevant literature. In fact, the purpose of radiometric calibration is to convert the DN value to the radiant luminance value (already mentioned in the front). Therefore, the gain value and offset value of the sensor must be entered into the ENVI to convert to a radiant luminance value. Edit the gain and offset values in ENVI by editing the ENVI header file. Envi 5.3 (The specific way to refer to the preface) through the raster management in the Edit Envi header tool to modify the two values, the following:

Figure Set Raster Metadata
You can add the gain and offset tools by clicking Add:

Figure Add Gain and Offset tool
Once added, we can see that the interface has been modified to gain and offset values. As shown in the following:

Figure Gain and offset interface
Next we begin to add two values, here actually there is a way to modify (manual), the specific operation, not in the narrative, the specific operation please refer to the blog Envi high score one PMS camera multispectral data atmospheric correction, here I use automatic file selection method for quick Add. First of all, I'm here to make a list of the 2015 domestic terrestrial observation satellite field absolute radiation calibration coefficient table for manual use:

Fig. 24 2015 The radiometric calibration coefficients of domestic satellites
In the next step is to modify the two values, here we change the gain value as an example, the image of the spectral function and other parameters obtained, can be obtained through the China Resources Satellite Application Center.

Figure 25 Adding a gain file
Then we can find that the gain value of each band has been entered into the interface, as follows:

Figure 26 Adding gain values
Similarly, modify the offset value. The view metadata can be used to confirm whether the bands have been assigned to gain and offset.

Figure 27 Checking the parameter values for each band
By examining the spectral information, we find that the image needs atmospheric correction, and the atmospheric correction requires the following conditions:
1, the data is after the calibration of the radiant luminance (emissivity) data (UW/CM2*NM*SR)
2, the data has a central wavelength (wavelenth) value, if the hyperspectral also need band width (FWHM).
3, data type support four kinds: floating-point type, long integer type, integer type, unsigned integer type
4. Spectral range: 400~2500nm
We can see that there is no center wavelength in our image, so we also need to input the center wavelength information. Next we will add the center wavelength, for the center wavelength add here I use the manual way to add. In a specific way I can use ENVI to open the spectral curve of the image to determine. The steps are as follows:
Use the spectral library to add spectral curve files.

Figure spectral Library Viewer

Figure 29 Adding a spectral curve function file
The center wavelength can be obtained by clicking on the highest point of each band (Data value=1, it should be noted that the unit is nm).

In the same way to get the wavelength of other bands, here is a high score of the center wavelength of each band finishing:
Center Wavelength (UM)
band1:0.485
band2:0.56
band3:0.696
band4:0.797
This allows us to complete the conditions of the atmospheric correction and then perform other operations on the radiometric calibration. It is important to note that I need to limit the format of the radiometric calibration file because the file format recognized by the atmospheric correction is in BIL or BIP format and therefore needs to be selected.

Figure 30 file format
(then modified: In fact, this is only the image file for the header file parameter editing, so the file format can also not be selected here)
Next use the radiometric Correction > radiometric Calibration tool for radiometric calibration, step 19, select the file after the above error is found, pop up the Radiation Correction dialog box, as shown in:

Figure 31 radiometric calibration Output
In this way, the radiation calibration operation is complete!

4.5 Atmospheric Correction –flaash Tool

After the radiometric calibration is complete, we will begin the atmospheric correction. The successful completion of radiometric calibration directly affects the results of atmospheric correction. Here we use the Flaash Atmospheric correction tool to make atmospheric corrections to the imagery. Such as:

Figure Flaash Atmospheric correction tool
Opens and pops up the input parameters of the Flaash model tool. First we need to edit the input Radiance Image, which is the radiometric calibration file we just finished processing. After selecting the file pop-up radiance scale factors, here we choose the use of Single that option, scale factor here we have the radiation calibration there is a conversion, there is no need to modify it. As shown in the following:

Figure Radiance Scale factors
The next two are the output file paths that set the atmospheric correction

Figure 34 Setting the atmospheric correction output file path
Then we set the parameters of the sensors, sensor type Select GF-1, and then the parameters of the high number, such as the sensor height, image resolution will be automatically filled up, there is a ground elevation needs we are based on the study area to obtain. If we don't have the exact average elevation data in our hands, we can use the elevation imagery provided by ENVI to get the average elevation of the study area. The following are the specific methods:
Open Global Elevation Data (2010)

Figure GMTED2010
Then get it with the compute statistics tool

Figure Compute Statistics Tools
Select elevation Data

Figure 37 Selecting an elevation data file
Press Stats subset button, in the pop-up Select Statistics Subset dialog box, click the File button to select the research area.

Figure 38 Selecting a study area file
The Calculate Statistics Parameters dialog box appears, the default is basic statistics

Figure Basics Stats
Clicking OK will pop up the results of the basic statistics, here we see the average elevation we want.

Figure Mean Elevation
So we get the average elevation of the area we're studying is 135.592195m. Fill in the atmospheric correction is a note unit that needs to be converted to KM. Next we need to set the flight data parameter.
For flight data image acquisition time, it can be viewed through the view metadata. Unfortunately my image is nothing, so how to get the image acquisition time?

Figure Time
In fact, our image data tend to record data, in which there is an XML file, learned computer people know that the XML file is mainly used for data storage transmission, open it, you will be pleasantly surprised to find that what you want is inside. Such as:

Figure Receive time
It is important to note here that we are filling out Greenwich Mean time instead of Beijing. If we want to get to Beijing, we need to add 8 hours on this basis.

Figure Time Setting
When selecting the atmospheric model, refer to the tips given by Help.

figure-Select atmospheric Model
Because we do not have a moisture content value, we use the second option. We can see from the second table that we can judge by two variables: one is latitude, the other is the month. And we found that the table 90 degrees north latitude, not December, so through this law can be concluded that both are "on the small principle", here my research area is 35.7n~36.6n, time is April 23, so choose 30n:march, so the model select MLS.
The next need to set the aerosol model settings, as to do not do aerosol, we can be determined by our influence whether the aerosol can be done, if the KT algorithm set total defaults three upstream and downstream can not have value, then the image can not do aerosol inversion.

Figure 45 Aerosol KT algorithm
You need to set the advanced option, here we choose according to our own computer configuration.

Figure Advanced Settings
Then click the Apply button to complete the atmospheric correction. If the atmospheric correction is successful, we can check whether the light composing line of vegetation is standard.
Here I have a rough sketch of the common features of the spectral curve for your reference.

Fig. 47 Spectral Graph
Atmospheric correction calculation Time is more frequent, need to wait patiently, and the effect of calibration will not be automatically loaded into the view, you need to open data management to load. Here we choose Load cir false color load (right click) Compare the spectral curve before and after the atmospheric correction, and find that the atmospheric correction is successful.

At the end of the note, many operations in the front can be omitted, including the central wavelength of the editor. If your version is envi5.3, open the XML file directly and you can do the following. Since I started using the envi5.1 version, I want to manually edit more content.

A detailed course of land use classification--taking the image of high score as an example (I.)