Uneingeschränkter Zugang

Extraction of Soil and Water Conservation Measures Information from Remote Sensing Images Based on Image Segmentation Algorithm Protection Research

Songyu Wu

Songyu Wu

Liaoning Dongke Electric Power Co., Ltd.Shenyang, China
Suche nach diesem Autor auf
Sciendo|Google Scholar
Wu, Songyu
, 
Feiyu Zhang

Feiyu Zhang

Liaoning Dongke Electric Power Co., Ltd.Shenyang, China
Suche nach diesem Autor auf
Sciendo|Google Scholar
Zhang, Feiyu
, 
Xiaoyan Qi

Xiaoyan Qi

Liaoning Dongke Electric Power Co., Ltd.Shenyang, China
Suche nach diesem Autor auf
Sciendo|Google Scholar
Qi, Xiaoyan
, 
Nan Wu

Nan Wu

Liaoning Dongke Electric Power Co., Ltd.Shenyang, China
Suche nach diesem Autor auf
Sciendo|Google Scholar
Wu, Nan
 und 
Chengjun Li

Chengjun Li

Liaoning Dongke Electric Power Co., Ltd.Shenyang, China
Suche nach diesem Autor auf
Sciendo|Google Scholar
Li, Chengjun
  
23. Sept. 2025
Extraction of Soil and Water Conservation Measures Information from Remote Sensing Images Based on Image Segmentation Algorithm Protection Research's Cover Image
Über diesen Artikel
Vorheriger Artikel
Nächster Artikel
Zitieren
COVER HERUNTERLADEN
Online veröffentlicht: 23. Sept. 2025
Eingereicht: 03. Jan. 2025
Akzeptiert: 25. Apr. 2025
DOI: https://doi.org/10.2478/amns-2025-0989
Schlüsselwörter
© 2025 Songyu Wu et al., published by Sciendo
This work is licensed under the Creative Commons Attribution 4.0 International License.

Figure 1.

The flow chart for the Multi-scale segmentation algorithm
The flow chart for the Multi-scale segmentation algorithm

Figure 2.

Curves of Area ratio changing with window size
Curves of Area ratio changing with window size

Figure 3.

Overall classification accuracy evaluation
Overall classification accuracy evaluation

Information extraction segmentation parameter Settings

Argument Level Ground floor Second floor Third floor Fourth floor
Band weight R 1 1 1 1
G 1 1 1 1
B 1 1 1 1
green 1 1 1 1
red 1 1 1 1
NirRed 1 1 1 1
SWIR 0.2 0.2 0.2 0.2
VF 0.2 0.2 0.2 0.2
dem 0.2 0.2 0.2 0.2
Slope 0.2 0.2 0.2 0.2
Segmentation scale 50 80 120 170
colour 0.7 0.8 0.8 0.9
shape 0.5 0.4 0.4 0.3
smoothness 0.6 0.6 0.6 0.6
compactness 0.6 0.6 0.6 0.6

User accuracy evaluation table for each land type

Different grid size Land type 1m 5m 10m 15m 20m 25m 30m 35m
Sloping land 72.61 93.50 98.73 80.22 87.82 84.11 96.78 93.15
Terraced fields 98.33 97.36 98.18 98.12 97.38 97.37 98.92 97.59
damland 100.0 97.69 98.91 100.00 100.00 98.82 100.00 100.00
Sichuan plateau 99.96 100.00 98.38 100.00 100.00 99.84 100.00 100.00
Forest land 98.96 98.64 70.51 96.73 95.98 98.94 98.93 99.78
Open forest land 99.72 99.16 97.92 99.65 99.65 99.65 99.65 99.65
meadow 41.47 43.14 99.72 55.21 35.13 32.66 37.81 33.15
Water body 21.37 65.25 95.12 99.32 99.32 99.45 99.27 99.83
Settlement place 98.12 87.91 94.59 87.91 88.96 100.00 100.00 100.00
highroad 97.03 96.71 86.57 97.81 97.81 97.81 97.81 97.81
Wild grass land 99.53 99.61 99.61 99.61 99.61 99.61 99.61 99.61
Bare land 99.54 99.71 97.64 99.71 99.71 99.71 99.71 99.71

Evaluation table of cartographic accuracy of each land type

Different grid size Land type 1m 5m 10m 15m 20m 25m 30m 35m
Sloping land 99.51% 99.22% 74.39% 99.71% 99.29% 99.52% 99.19% 99.28%
Terraced fields 96.49% 96.49% 95.94% 96.49% 96.49% 96.49% 96.49% 96.49%
damland 99.61% 99.59% 99.13% 99.59% 99.59% 99.59% 99.59% 99.59%
Sichuan plateau 82.43% 82.43% 97.95% 82.43% 82.43% 98.25% 98.25% 98.25%
Forest land 75.69% 83.77% 98.62% 84.51% 78.29% 76.91% 86.11% 83.90%
Open forest land 99.57% 99.12% 98.37% 99.57% 99.57% 99.57% 99.57% 99.57%
meadow 99.35% 99.47% 32.87% 99.47% 99.47% 99.47% 99.47% 99.47%
Water body 98.41% 98.41% 97.72% 98.41% 98.41% 98.41% 98.41% 98.41%
Settlement place 95.91% 96.78% 96.78% 99.01% 96.78% 96.78% 96.78% 96.78%
highroad 97.91% 97.91% 87.29% 97.91% 97.91% 97.91% 97.91% 97.91%
Wild grass land 97.82% 98.53% 98.53% 98.53% 98.53% 98.53% 98.53% 98.53%
Bare land 99.29% 99.40% 99.40% 99.40% 99.40% 99.40% 99.40% 99.40%

Haralick texture feature parameters, mathematical definition and description

Parameter name Mathematical definition Description
homogeneity h h=∑i,j=0N−1Pi,j1+(i−j)2 Describe the mean value of the image, that is, the degree to which the larger elements in the gray scale co-occurrence matrix are concentrated in the diagonal line. The more concentrated, the greater the homogeneity value, indicating the higher the mean value of the image
Contrast C C=∑i,j=0N−1Pi,j(i−j)2 In contrast to homogeneity, it measures the degree of local variation of the image, and when the image varies greatly in the local range, the contrast value is also large
Dissimilarity degree d d=∑i,j=0N−1Pi,j|i−j| Linear correlation with contrast, the higher the local contrast, the greater the difference
Mean value u ui,j=∑i,j=0N−1Pi,jN2 The average value is calculated by the joint occurrence frequency of the pixel values in the gray scale co-occurrence matrix and the adjacent pixel values
Standard deviation S S=∑i,j=0N−1Pi,ji,j−ui,j Similar to contrast and dissimilarity, standard deviation is expressed in the form of gray co-occurrence matrix, which is a measure of the deviation between pixel value and mean value
entropy e e=∑i,j=0N−1Pi,j(−lnPi,j)2 When the texture in the image is inconsistent, the gray difference vector element value is small and the entropy value is large
Angular second moment a a=∑i,j=0N−1Pi,j2 It describes the homogeneity and consistency of image gray distribution. When the image is homogeneous or consistent texture, the angular second order moment value is larger
correlation Co Co=∑i,j=0N−1Pi,ji−uij−ujσi2•σj2 Measure the degree of linear dependence on the gray level of adjacent pixels and reflect the directionality of linear ground objects in the image. When linear ground objects are arranged in a certain direction, the correlation of this direction is higher than that of other directions
Angular second moment GLDV ∑k=0n−1V(k)2 The homogeneity, also known as energy, that describes the grayscale difference of an image
Entropy GLDV ∑k=0n−1−V(k)ln(V(k)) A measure of whether the texture features in an image are cluttered
Contrast GLDV ∑k=0n−1k2V(k) Measure the local change degree of the image, and the contrast value is also large when the image is very large
Mean value GLDV ∑k=0n−1kV(k) The average value is calculated by the joint occurrence frequency of the pixel values in the gray scale co-occurrence matrix and the adjacent pixel values
Sprache:
Englisch
Zeitrahmen der Veröffentlichung:
1 Hefte pro Jahr
Fachgebiete der Zeitschrift:
Biologie, Biologie, andere, Mathematik, Angewandte Mathematik, Mathematik, Allgemeines, Physik, Physik, andere
Zeitschrift RSS Feed