Search

CN-121982541-A - Soil erosion feature extraction method based on image segmentation

CN121982541ACN 121982541 ACN121982541 ACN 121982541ACN-121982541-A

Abstract

The invention discloses a soil water erosion feature extraction method based on image segmentation, and relates to the technical field of water erosion feature extraction. The method comprises the steps of obtaining an original soil remote sensing image of a target area, preprocessing the original soil remote sensing image to obtain a preprocessed image, extracting spectral features, texture features and exposure features of soil from the preprocessed image, fusing the extracted spectral features, texture features and exposure features to generate a water erosion sensitivity image layer, collecting digital elevation model data of the target area, calculating terrain factors, dividing different terrain function units according to the calculated terrain factors, presetting a differentiated layered image segmentation rule base for each unit, segmenting the water erosion sensitivity image layer based on the rule base to obtain a water erosion area segmentation image, constructing a water erosion contour extraction model based on an energy minimization theory, optimizing the segmentation image, and finally outputting accurate soil water erosion features. According to the method, the precision of extracting the soil erosion features is effectively improved by fusing the multi-feature and the layered segmentation strategy.

Inventors

  • YANG XUYAN
  • CHEN XU

Assignees

  • 陕西工业职业技术大学
  • 中国科学院东北地理与农业生态研究所农业技术中心

Dates

Publication Date
20260505
Application Date
20260128

Claims (10)

  1. 1. The soil erosion feature extraction method based on image segmentation is characterized by comprising the following steps of: step S1, acquiring an original soil remote sensing image of a target area, and preprocessing the original soil remote sensing image to obtain a preprocessed soil remote sensing image; s2, carrying out feature extraction on the preprocessed soil remote sensing image to obtain soil spectrum features, soil texture features and soil exposure features, and carrying out fusion to generate a water erosion sensitivity layer; Step S3, collecting digital elevation model data of a target area, calculating to obtain a terrain factor based on the digital elevation model data, and dividing the target area into different terrain function units according to the terrain factor; S4, constructing a terrain segmentation mechanism based on a layered segmentation rule base, and segmenting the water erosion sensitivity layer through the terrain segmentation mechanism to obtain a water erosion area segmentation map; And S5, constructing a water erosion contour extraction model based on an energy minimization theory, inputting the water erosion region segmentation map into the water erosion contour extraction model, and outputting to obtain soil water erosion characteristics.
  2. 2. The method for extracting soil erosion characteristics based on image segmentation according to claim 1, wherein the steps of obtaining an original soil remote sensing image of a target area, preprocessing the original soil remote sensing image to obtain a preprocessed soil remote sensing image comprise: Data collection is carried out, a satellite imaging sensor is selected, a high-resolution soil remote sensing image of a single time phase of a target area is obtained, and an original soil remote sensing image is obtained ; For original soil remote sensing image The first link of the pretreatment is a radiation correction link, and the original digital quantized value is converted into the earth surface reflectivity through an atmosphere correction model; performing geometric fine correction by a coordinate transformation function model: ; Wherein, the The original soil remote sensing image coordinates; in order to correct the geographical coordinates after the correction, Is a rational function model; Representing a ground control point set; through the pretreatment flow, the original soil remote sensing image Is converted into a pretreated soil remote sensing image to obtain a pretreated soil remote sensing image 。
  3. 3. The method for extracting soil erosion characteristics based on image segmentation according to claim 2, wherein the step of extracting characteristics of the preprocessed soil remote sensing image to obtain soil spectrum characteristics, soil texture characteristics and soil exposure characteristics, and performing fusion to generate an erosion sensitivity layer comprises the following steps: extracting multidimensional characteristics from the pretreated soil remote sensing image Three physical characteristic components with indication on the water erosion phenomenon are analyzed, wherein the three physical characteristic components comprise soil spectrum characteristics, soil texture characteristics and soil exposure characteristics; Spectral features of soil Is a linear combination of a vegetation index and a soil brightness index to comprehensively characterize spectral anomalies; calculating entropy Entropy by using gray level co-occurrence matrix to measure complexity and irregularity of texture and obtain soil texture characteristics ; Obtaining soil exposure characteristics by adopting normalized difference exposure index NDBI ; After the soil spectrum characteristics, the soil texture characteristics and the soil exposure characteristics are obtained, characteristic normalization processing is needed, and pixel-by-pixel weighted fusion is carried out on the normalized characteristics to generate a unified water erosion sensitivity image layer 。
  4. 4. A method for extracting soil erosion features based on image segmentation as defined in claim 3, wherein collecting digital elevation model data of a target area, and calculating a topography factor based on the digital elevation model data, comprises: Collecting digital elevation model data covering a target area, noted as The data source is derived from a public database of the topographic survey and drawing task of the spaceborne radar or the spaceflight plane radar, and is used for digital elevation model data Filling the data hole, smoothing noise pretreatment to obtain pretreated digital elevation model data with reliable quality, which is recorded as ; Based on preprocessing digital elevation model data Performing terrain factor calculation to provide quantitative index and gradient for unit division The calculation is as follows: ; Wherein, the And Respectively represent the elevation change rate of DEM in east-west and north-south directions, arctan ) Is an arctangent function; is a partial differential symbol, x and y respectively represent the preprocessed digital elevation model data The east-west direction and the north-south direction of the coordinate axes, the gradient Is a core factor for controlling the energy and erosion potential of surface runoff.
  5. 5. The method for extracting soil erosion features based on image segmentation according to claim 4, wherein the dividing the target area into different terrain functional units according to the terrain factors comprises: according to the topography and water erosion principle, the topography functional units are divided based on the topography factors obtained by calculation, the target area is divided into three basic topography functional units, and the topography functional units are assembled into Representing the top zone, the middle zone and the toe zone, respectively.
  6. 6. The method for extracting soil erosion characteristics based on image segmentation according to claim 5, wherein the step of presetting a differential decision threshold to form a hierarchical segmentation rule base based on different terrain functional units comprises the following steps: setting a critical differential decision threshold for each cell Threshold values and corresponding relations of all units form hierarchical segmentation rule base The core mapping relation of the rule base is simplified into I.e. for any belonging unit Is determined as the water erosion area, i.e. the threshold value The rule base is constructed based on the following core principles: in the slope top area with weak erosion signal Setting a relatively low decision threshold In the slope region with linear erosion dominant Setting a decision threshold of a medium level In the slope toe area with obvious sedimentation effect and blurred boundary Setting a relatively high decision threshold 。
  7. 7. The method for extracting soil erosion characteristics based on image segmentation according to claim 6, wherein the construction of a terrain segmentation mechanism based on a hierarchical segmentation rule base comprises the following steps: The method comprises the steps of constructing a terrain segmentation mechanism M, which is a mechanism for converting geographic knowledge into image segmentation decisions, wherein the core is that a mapping relation from a space position to a dynamic decision threshold is established, and the mapping relation is used for any pixel point in an image The operation flow of the mechanism is as follows: firstly, positioning a terrain unit according to pixels Determining the geographic coordinates of the associated terrain function ; Re-query rule base Acquiring the topography unit Preset decision threshold And assign it to the pixel This process may be formalized by the following mapping function: ; Wherein, the Is a pixel The minimum sensitivity threshold value required to be reached by the water erosion area is judged, and the threshold value is determined by the terrain unit where the threshold value is positioned; Is a query function that receives as input a terrain cell type and then returns from the hierarchical segmentation rule base the optimal empirical threshold preset for that type of terrain.
  8. 8. The method for extracting soil erosion characteristics based on image segmentation according to claim 7, wherein the segmentation of the erosion sensitivity layer by the terrain segmentation mechanism to obtain an erosion region segmentation map comprises: Dividing the water erosion sensitivity layer by a terrain dividing mechanism to obtain a dynamic decision threshold value of each pixel After that, the sensitivity to water erosion layer And executing segmentation judgment, wherein the judgment rule is that if the sensitivity value of a certain pixel is not lower than a threshold value required by a terrain unit where the pixel is positioned, the pixel is judged to be a water erosion area, and the segmentation judgment formula is as follows: ; Wherein, the Is a water erosion area division diagram in pixels The value of the water erosion area is 1, 0 is non-water erosion area, and the continuous water erosion sensitivity information is obtained According to a reasonable critical standard which varies with the terrain space Binarizing to generate a preliminary water erosion area segmentation map 。
  9. 9. The method for extracting soil erosion features based on image segmentation according to claim 8, wherein constructing the erosion contour extraction model based on the energy minimization theory comprises: constructing a water erosion profile extraction model, wherein the model takes a level set method as a framework to convert the boundary evolution problem into an energy functional Is a level set function Is an implicitly expressed scalar function whose zero level set Representing the outline of the water erosion area, the total energy functional is as follows: ; Wherein, the For the image data item, driving the contour to be attached to the gradient edge in the image; Introducing texture and spectrum continuity priori knowledge of a water erosion area for a structural constraint term; the smoothness of the contour is constrained for regularization items, so that unreasonable tortuosity is avoided; Is a weight coefficient and is used for balancing the contribution of each item; the model optimization is carried out by adopting a narrow-band level set method to carry out iterative solution to efficiently solve the energy functional, and after the extracted model of the soil water erosion characteristic is converged, the final level set function is obtained Extracting zero level set thereof : ; Wherein, the Is a definition field; representing elements in the collection as a point; Meaning that the set of all points forms a curve.
  10. 10. The method for extracting soil erosion features based on image segmentation according to claim 9, wherein the narrow-band level set method comprises: initializing, and dividing the binary image into two parts Conversion to a symbolic distance function as a level set function And construct an initial value centered on the zero level set and a width of Is a narrow band of the initial band; in iterative evolution, i.e. at the first In each iteration, the following steps are performed: Firstly, calculating the evolution speed, and calculating the evolution speed field of a level set function on all pixel points in a current narrow band The velocity field is determined by the gradient descent of the energy functional, containing driving forces from image gradients, texture constraints and curvature smoothing; with explicit euler format, the level set function is updated only within a narrow band: ; Wherein, the Is a level set function, superscript Is an iteration index; Represent the passing of the first Boundary states after iterative optimization; Is the time step; then carrying out narrowband updating and reinitialization, detecting whether the zero level set moves to be close to the boundary of the current narrowband, if so, reconstructing the narrowband by taking the new zero level set as the center, and carrying out the method Carrying out quick reinitialization; repeating the above process until the position change of the zero level set is less than a predetermined threshold Or to a maximum number of iterations At the moment, the water erosion profile extraction model is considered to be converged to obtain a final level set function 。

Description

Soil erosion feature extraction method based on image segmentation Technical Field The invention relates to the technical field of water erosion feature extraction, in particular to a soil water erosion feature extraction method based on image segmentation. Background Soil erosion is one of the key factors that lead to land degradation and ecological environment degradation. The method realizes the automatic extraction of the soil erosion characteristics in a large range and high precision by using the remote sensing technology, and has great significance for water and soil conservation planning and effect evaluation. Along with the popularization of high-resolution remote sensing images, how to accurately identify and outline the spatial range and the fine outline of a soil erosion area from a complex ground surface coverage background becomes a research hot spot and a technical difficulty in the field. The traditional soil erosion feature extraction method based on image segmentation is to pre-process an acquired remote sensing image, preliminarily identify an erosion region by calculating a spectrum index and combining texture feature analysis, then segment the erosion region by adopting an image segmentation method or an edge detection operator based on a fixed threshold value, and directly generate the boundary of the erosion region to be output as a final soil erosion feature. However, after the traditional method is finished, due to the mixed pixels, noise and ground object boundary blurring phenomenon in the remote sensing image, the directly segmented contour always presents a saw-tooth shape, a fracture shape or an irregular shape, so that the due smoothness and continuity of a real water erosion boundary are seriously deviated, the extracted water erosion characteristics have obvious deviation in spatial position, range and geometric precision of the contour, and the requirements of high-precision quantitative analysis and treatment engineering planning are difficult to meet. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a soil erosion feature extraction method based on image segmentation, which solves the problems existing in the background art. In order to achieve the purpose, the invention is realized by the following technical scheme that the soil erosion feature extraction method based on image segmentation comprises the following steps: step S1, acquiring an original soil remote sensing image of a target area, and preprocessing the original soil remote sensing image to obtain a preprocessed soil remote sensing image; s2, carrying out feature extraction on the preprocessed soil remote sensing image to obtain soil spectrum features, soil texture features and soil exposure features, and carrying out fusion to generate a water erosion sensitivity layer; Step S3, collecting digital elevation model data of a target area, calculating to obtain a terrain factor based on the digital elevation model data, and dividing the target area into different terrain function units according to the terrain factor; S4, constructing a terrain segmentation mechanism based on a layered segmentation rule base, and segmenting the water erosion sensitivity layer through the terrain segmentation mechanism to obtain a water erosion area segmentation map; And S5, constructing a water erosion contour extraction model based on an energy minimization theory, inputting the water erosion region segmentation map into the water erosion contour extraction model, and outputting to obtain soil water erosion characteristics. Preferably, obtaining an original soil remote sensing image of a target area, preprocessing the original soil remote sensing image, and obtaining a preprocessed soil remote sensing image includes: Data collection is carried out, a satellite imaging sensor is selected, a high-resolution soil remote sensing image of a single time phase of a target area is obtained, and an original soil remote sensing image is obtained ; For original soil remote sensing imageThe first link of the pretreatment is a radiation correction link, and the original digital quantized value is converted into the earth surface reflectivity through an atmosphere correction model; performing geometric fine correction by a coordinate transformation function model: Wherein, the The original soil remote sensing image coordinates; in order to correct the geographical coordinates after the correction, Is a rational function model; Representing a ground control point set; through the pretreatment flow, the original soil remote sensing image Is converted into a pretreated soil remote sensing image to obtain a pretreated soil remote sensing image。 Preferably, extracting features of the preprocessed soil remote sensing image to obtain soil spectrum features, soil texture features and soil exposure features, and fusing to generate the water erosion sensitivity layer comprises: extracting multidimensional charact