Search

CN-116258975-B - Remote sensing extraction method for forest land information of power transmission and transformation project construction area

CN116258975BCN 116258975 BCN116258975 BCN 116258975BCN-116258975-B

Abstract

The invention discloses a remote sensing extraction method for forest land information in a power transmission and transformation project construction area, which belongs to the technical field of remote sensing extraction of forest land information, and comprises the steps of selecting high-resolution satellite remote sensing images when main crops in autumn of a task area are harvested according to the physical characteristics of the task area; dividing the image to generate an image object layer, calculating a normalized vegetation index of the image, determining three thresholds of the normalized vegetation index, brightness and gray level co-occurrence matrix contrast, and classifying the image object to obtain woodland information. The invention adopts the high-resolution remote sensing image to extract the woodland information according to the woodland characteristics, has the advantages of simple principle, strong applicability, high precision and technical process, and can be suitable for extracting the information of the large-scale waling places.

Inventors

  • LI PENG
  • ZHOU WEIQING
  • ZHANG ZIJIAN
  • ZHENG XIAOBIN

Assignees

  • 国网冀北电力有限公司电力科学研究院
  • 国家电网有限公司

Dates

Publication Date
20260505
Application Date
20230323

Claims (6)

  1. 1. The remote sensing extraction method for the forest land information of the power transmission and transformation project construction area is characterized by comprising the following steps of: Acquiring a remote sensing image according to the weather condition of a task area, acquiring a high-resolution satellite remote sensing image of the task area, wherein the imaging time of the satellite image is used for selecting an image when a main crop is harvested in autumn according to the weather characteristic of the task area, the remote sensing image comprises a blue wave band, a green wave band, a red wave band and a near infrared wave band, and performing pretreatment of radiation correction, wave band fusion and geometric correction on the acquired remote sensing image; Secondly, image segmentation, namely performing chessboard segmentation on the task area image firstly, and then performing multi-scale segmentation on the basis of the chessboard segmentation to generate an image object layer; calculating a normalized vegetation index, namely calculating the image data to obtain normalized vegetation index data, wherein the calculation formula of the normalized vegetation index is as follows: Wherein NDVI is normalized vegetation index, ρNIR is the pixel surface reflectance value of the near infrared band, ρR is the pixel surface reflectance value of the red band; the normalized vegetation index of the image object is the average value of the normalized vegetation index value of each pixel in the image object; Determining a threshold value, wherein a woodland image object is selected from the image data as a sample land block selected by the threshold value, the determined threshold value comprises a normalized vegetation index threshold value, a brightness threshold value and a gray level co-occurrence matrix contrast threshold value, and the method comprises the following steps of: The normalized vegetation index threshold value is selected to be N1 and N2, N1 is the minimum value in the non-shadow woodland sample NDVI value, and N2 is the maximum value of the shadow woodland sample NDVI value; The brightness threshold is I1, and I1 is the maximum value in HSI conversion-brightness values of the woodland samples; the contrast threshold of the gray level co-occurrence matrix is C1, and C1 is 1/2 of the sum of the maximum value and the minimum value in the forest land sample; step five, threshold classification, namely extracting the forest land of the image object based on the threshold, and specifically comprising the following steps: S1, extracting an image object with a normalized vegetation index value being greater than or equal to N1 and a brightness value being less than or equal to I1 as a woodland; S2, extracting an image object with a normalized vegetation index value being greater than or equal to N2, a brightness value being less than or equal to I1 and a gray level co-occurrence matrix contrast value being less than or equal to C1 as a woodland; s1, the woodland information extracted in the step S2 and the woodland information extracted in the step S jointly form final extracted woodland information; The method comprises the steps of carrying out precision verification by comparing with woodland information extracted by manual visual interpretation, specifically carrying out manual visual interpretation on a task area to extract woodlands, carrying out space superposition on woodland vectors extracted by manual visual interpretation and woodland vectors extracted by the method, marking the woodland area extracted by manual visual interpretation as MJ_R, marking the woodland area extracted by the method as MJ_M, marking the overlapped part area as MJ_R & M, and respectively calculating the precision of a producer and the precision of a user.
  2. 2. The remote sensing extraction method for forest land information in a power transmission and transformation project construction area according to claim 1, wherein the high-resolution satellite remote sensing image obtained in the first step is a sub-meter remote sensing image with spatial resolution better than 1 meter.
  3. 3. The remote sensing extraction method for forest land information in a power transmission and transformation project construction area according to claim 1, wherein in the second step, the size of a segmented object segmented by a chessboard is controlled to be 3-5 meters, the segmentation scale of multi-scale segmentation is 200, the shape parameter is 0.5, and the compactness parameter is 0.5.
  4. 4. The remote sensing extraction method for forest land information in a power transmission and transformation project construction area according to claim 1, wherein in the fourth step, the number of the selected forest land image sample land blocks is not less than 20, and the sample land blocks cover all forest land types in a task area and comprise forest land samples in mountain shadows.
  5. 5. The remote sensing extraction method for forest land information in a power transmission and transformation engineering construction area according to claim 1, wherein the gray level co-occurrence matrix contrast in the fourth step is a contrast value in the direction of 0 degrees.
  6. 6. The remote sensing extraction method of forest land information in a power transmission and transformation project construction area according to claim 1, wherein in the sixth step, a calculation formula of producer precision is as follows: The calculation formula of the user precision is as follows: 。

Description

Remote sensing extraction method for forest land information of power transmission and transformation project construction area Technical Field The invention relates to the technical field of forest land information remote sensing extraction, in particular to a forest land information remote sensing extraction method for a power transmission and transformation project construction area. Background The construction area of the power transmission and transformation project has long points, complex construction period, various landforms, large relief fluctuation and multiple crossing of ecologically sensitive and fragile areas. Timely and accurately extracting forest land information, monitoring forest land change, and helping to protect and manage ecological environment and maintain forest land ecological system stability in construction period of power transmission and transformation project. The traditional method for acquiring the woodland information is low in efficiency, poor in timeliness and high in cost, and the remote sensing information technology becomes an important means for acquiring the woodland information by virtue of the advantages of real time, rapidness and the like, such as medium-spatial resolution satellite remote sensing images of GF-1, MODIS, sentinel-2, landsat series and the like and high-spatial resolution satellite remote sensing images of IKONOS, quickBird, worldView series and the like. The medium-resolution image data has lower spatial resolution, poor recognition effect on zero scattered forest land, and the high-resolution satellite remote sensing image has obvious advantages in forest land information extraction. In recent years, researchers develop research on woodland information extraction methods by adopting various methods based on different types of remote sensing image data. However, these researches are mainly based on medium-high resolution images, and the researches on the forest land extraction method of high-resolution, especially sub-meter resolution remote sensing images are less, and the technical method capable of being applied to large-area business is lacking. On the remote sensing image, the woodland is easy to be confused with the information of crops and grasslands, so how to effectively distinguish the woodland from the crops and the grasslands by using the remote sensing image is a key and difficult point for extracting the distribution information of the woodland. Disclosure of Invention Technical problem to be solved Aiming at the problems existing in the prior art, the invention aims to provide a remote sensing extraction method for the forest land information of a power transmission and transformation project construction area, which can rapidly and accurately extract the forest land information in the business application by utilizing a high-resolution (better than 1 meter) remote sensing image, thereby being more practical. Technical proposal In order to solve the problems, the invention adopts the following technical scheme. The woodland information referred to by the invention is spatial position, range and area information of the woodland. A remote sensing extraction method for forest land information of a power transmission and transformation project construction area comprises the following steps: Step one, acquiring a remote sensing image according to the task area climatic conditions And acquiring a high-resolution satellite remote sensing image of the task area, wherein the imaging time of the satellite image is used for selecting an image of the main crop in autumn when harvesting according to the physical characteristics of the task area. At this time, the woodland does not turn yellow, and grasslands and farmlands begin to wither yellow, so that the interference of grasslands and farmlands on information extraction of the woodland can be reduced to the greatest extent. The image includes blue, green, red, and near infrared bands. The image should have been pre-processed for radiation correction, band fusion, geometric correction, etc. Step two, image segmentation The task area image is subjected to chessboard segmentation firstly, and then multi-scale segmentation is performed on the basis of the chessboard segmentation. And generating an image object layer. Step three, calculating normalized vegetation index (NDVI) And calculating the image data to obtain normalized vegetation index (NDVI) data. The calculation formula of the normalized vegetation index (NDVI) is: Wherein NDVI is a normalized vegetation index, ρ NIR is a pixel surface reflectance value of a near infrared band, and ρ R is a pixel surface reflectance value of a red light band; the normalized vegetation index of the image object is the average value of the normalized vegetation index values of each pixel in the image object. Step four, determining a threshold value And selecting a threshold sample, namely selecting a woodland image object in the image data as a sample la