Remote Sensing Digital Image Processing [Feature Extraction] —— Xi'an

"Remote Sensing Digital Image Processing" course assignment

1. Course Requirements

1.1 Specific Requirements for Course Arrangement

(1) Master the general method of independent development or secondary development based on commercial software and programming language to realize image processing. Enable students to realize image correction, image enhancement, image classification and other operation modules based on relevant principles and technical methods, demand analysis report (30%);

(2) For engineering application fields such as image classification, thematic information extraction and change detection, be able to design and effectively implement an appropriate image processing flow plan, and be able to correctly evaluate the processing results. (40%);

(3) Give the effect of the implemented system in practical application, use software development tools and data processing and analysis software modules, and submit specific research area drawings (30%).

1.2 Basic Requirements for Course Design

(1) Data download and preprocessing. Download the SPOT image corresponding to the study area (spatial resolution 20m) to complete the preprocessing such as mosaic, cropping, and correction of the image of the study area; establish a data set for 3 consecutive years, with a period of 2000-2015;

(2) Establish feature extraction algorithm. Extract the urban impermeable surface, vegetation and water body in the research area, design three kinds of ground feature extraction algorithms, list the algorithm formula, realize the algorithm program, establish the algorithm accuracy verification, and realize the three kinds of ground feature extraction;

(3) Remote sensing mapping. According to different types of ground features, establish three kinds of ground feature mapping for three consecutive years, increase the number of maps, and make thematic maps according to the drawing standards;

(4) Complete the description of the algorithm, the work description of the mapping results of the large-scale operation, complete the report of the large-scale operation, and describe in detail the process of remote sensing mapping and the advantages of this scheme.

2. Preliminary preparation

2.1 Data Acquisition

Because we need to use SPOT image data for downloading, and the SPOT series satellites are an earth observation satellite system developed by the French Space Research Center (CNES), and SPOT satellites 1-7 have been launched. At present, CNES has opened the satellite data of spot1-5, and the time range of free download data is from 1986 to 2014.
Before officially downloading SPOT data, we need to understand the specific parameters of SPOT satellites, that is, the colors of image bands captured by different SPOT satellites are different, and the specific data are shown in the following figure:

The satellite parameters in the above picture are my personal summary of articles written in the blog for my own study and record use. The link of the article is: key summary: SPOT data download, image band data introduction, three methods of multispectral data simulation for true color [reprinted and integrated] the
blog also summarizes and records SPOT data download, image band data introduction, and multispectral data simulation The purpose of the three methods of true color is to have a basic concept of data selection before downloading SPOT data, and refer to the detailed parameters of the data, so that the downloaded data can be used and feature extraction operations can be performed.
The URL for SPOT data download is: REGARDS OSS

① First register as a user, and then get free data download times (up to 1000 times);

②After registering and logging in, click "Date Search" to enter the data search interface;

③Click the red box similar to the small map icon to enter the map selection interface;

④According to the area we are studying, first click on the "dashed box" in the red box to select the research area (take Xi'an as an example);

You can see that the available data are densely packed red squares on the map, and a frame corresponds to a remote sensing image. The image has black and white and false color (the reason for no true color is that the blue band is included in the SPOT satellite 6-7, and downloading requires a fee. The free spot only includes red, green, near-infrared, mid-infrared, and panchromatic bands. It is mentioned in the SPOT satellite parameters in the above picture and in my blog).
We can choose to click the red square to download the data we need, but before that, we still need to filter the data, and ensure that the cloud density in the downloaded data is below 5%, so that the feature recognition can be performed when extracting features. more precise.

⑤ After selecting the data, we click the button in the red box, and then click "COLUMNS" to filter the data;

⑥We put "Date" and "Cloud Free %" in the front, and click the button "First" on the right;

⑦ After that, we can see that the cloud density and time are arranged in front, which is convenient for us to choose;

After selecting the data we want, click the download button at the back (the exclamation mark download button is displayed here because the number of downloads has reached the upper limit).

⑧The selected data to be downloaded is displayed in our account shopping cart, and the compressed package can be downloaded directly after the system is packaged;

2.2 Operating environment

Required software: ENVI 5.3, ArcMap 10.2;
Operating system: Window 11/10;
Computer configuration: 8G or more storage, 512G or more hard disk space;

3. Operation steps

The data needs to be processed by two softwares, so the following operations are required.

3.1 ENVI part

After downloading the images, we first need to mosaic the images, stitching multiple single images into a complete image.

①First load all the images to be mosaiced into ENVI, select the Map – Mosaicking – Georeferenced tool, import all the images, select the output, export the mosaicked images, and load the images into ENVI;

②Crop the region of interest;
after mosaicking an image, it is not our research area, so the image needs to be clipped.
Open the mosaicked image in ENVI, select the File – Save as – Save as tiff tool, select the shp file that you want to crop the area, and click OK to crop the area of ​​interest.

③ NDVI numerical calculation;
calculate the NDWI value of the cropped image, open the Band Math band calculation tool, and input the formula of the NDWI normalized difference water index (taking the water extraction method as an example, the same is true for vegetation and towns), where b1 we Select the Green band and b3 select the NIR band to calculate the NDWI index.

④Water body range extraction:
After calculating the NDWI value, it can be roughly judged that the value of the water body is less than -0.25 through the histogram of the pixel value distribution, so we use this benchmark for water body extraction.
I was going to use the threshold method for extraction, but the output image is always all gray, so I changed the method.
Right-click the NDWI image output in the previous step, select New Raster Color Slices, click the options in the figure below in order, set the maximum and minimum pixel values ​​in the pop-up box on the right, click OK after completion, right-click the generated Slices folder, and select Save To the shp file, the range of the water body can be exported.

⑤ Open the output water range in Arcgis, and the overall extraction effect is good.
3.2 ArcMap
Import the extracted vegetation, water body, and town data into ArcMap to make a thematic map;
follow the previously learned thematic map making steps, insert legends, compass, scale, etc., draw up the title and export the image That's it.
The following is the three-year thematic map of Xi'an City that I was in charge of:

year 2011:

2012:

year 2013:

Four. Summary

This time, the extraction of water bodies, vegetation, and towns was carried out through the NDWI index. The setting of the threshold may not be standard, resulting in images that may cause errors in the extraction of water bodies, vegetation, or towns due to the same threshold. Part of the influencing factor is the downloaded SPOT The color of the data image is different, there are deep and light, and the data is not captured at the same time, resulting in a large extraction error. In specific operations, we can also use the DN value to determine the approximate range of the threshold for more accurate extraction. The specific value can be obtained by using the tool threshold reference tool in ENVI, which is convenient for us to speculate the approximate range of the threshold value of the ground object.
After nearly 40 years of development, remote sensing information feature extraction techniques are gradually being perfected. However, the remote sensing information is rich, the source is wide, the complexity is high, and the technical level of remote sensing information feature extraction is far behind the user's requirements. There are mainly the following problems: 1. The existing remote sensing information extraction method is complicated and difficult to
operate .
2. Although there are many existing remote sensing information extraction methods, the remote sensing information extracted by many methods is not rich enough and not complete enough.
3. Feature extraction, especially the feature extraction of Jiding geographic information system is still in the development stage. The current methods and theories in this area need to be improved.
This course has benefited a lot after the completion of the major assignments. Not only have I learned the digital processing flow of remote sensing images, the basics of image processing, image transformation, image correction, image enhancement, image segmentation and feature extraction, image classification and change detection, etc. The key to ensuring the accuracy of image classification is feature extraction. In order to improve the accuracy of image classification, we can extract the nonlinear information of the original image data. However, the larger the data, the more time it takes to process the data, and the higher the requirements for the equipment. If you want to engage in such a career in your daily work and life in the future, you first need to have a meticulous spirit of learning, a strong exploratory power, and a strong desire to learn new knowledge. Only by not forgetting your original intention can you go further and further on this road.

one more word

For a more detailed NDVI extraction process, you can refer to this blogUsing ENVI to process SPOT remote sensing images to extract water bodies, vegetation, and impermeable surfaces

Friends who like it can like, forward + bookmark.