CN-121982228-A - Method and system for constructing coastal zone landscape map based on image processing

CN121982228ACN 121982228 ACN121982228 ACN 121982228ACN-121982228-A

Abstract

The invention relates to the technical field of image recognition, and discloses a coastal zone landscape map construction method and system based on image processing, wherein the method comprises the following steps: obtaining a coastal zone image, preprocessing, atmospheric correction and geometric registration to obtain a corrected image, generating a vegetation index image and a water body index image based on multiband index calculation, fusing to generate a preliminary mask image, carrying out zone division and label fusion by adopting a watershed algorithm to obtain a segmented mask image, extracting texture and color feature calculation similarity, generating a water body reflection inhibition graph and extracting a boundary path if the similarity is lower than a threshold value to generate a refined boundary feature image, and finally carrying out multi-frame difference analysis and contour vectorization to generate a final vegetation distribution map. The method realizes effective distinction between the water body and the semi-submerged vegetation under the conditions of tidal change and water body reflection interference.

Inventors

CHEN YOURU
LU SHENYI
LIANG XIAOYI

Assignees

深圳翰博设计股份有限公司

Dates

Publication Date: 20260505
Application Date: 20251216

Claims (8)

1. The method for constructing the coastal zone landscape map based on the image processing is characterized by comprising the following steps of: acquiring a coastal zone image; Preprocessing according to the coastal zone region image to obtain a denoised multispectral image, and performing atmospheric correction and geometric registration processing on the denoised multispectral image to obtain a corrected image; Calculating the multiband difference value index of the correction image to obtain a vegetation index image and a water body index image, and generating a preliminary mask image based on the vegetation index image and the water body index image; performing region division processing on the preliminary mask image by adopting a watershed algorithm to obtain a candidate segmented region image, and performing region water judgment and pixel-level label fusion on the candidate segmented region image to obtain a segmented mask image; extracting texture features and color features of the segmentation mask image to obtain a texture feature set and a color feature set, and calculating similarity according to the texture feature set and the color feature set to obtain a similarity score; When the similarity score is lower than a preset similarity threshold, generating a water body reflection inhibition chart, and carrying out boundary path extraction and contour compression processing on the water body reflection inhibition chart to obtain a refined boundary feature chart; and carrying out multi-frame difference analysis and boundary extraction processing on the refined boundary feature map to obtain a dynamic water area boundary path, and carrying out boundary contour vectorization and vegetation distribution generation processing on the dynamic water area boundary path to obtain a final vegetation distribution map.
2. The method for constructing a coastal zone landscape map based on image processing according to claim 1, wherein the preprocessing is performed according to the coastal zone region image to obtain a denoised multispectral image, and the atmospheric correction and geometric registration processing are performed on the denoised multispectral image to obtain a corrected image, and the method comprises the steps of: performing noise suppression on the coastal zone region image by adopting median filtering to obtain a denoised multispectral image; According to the denoised multispectral image, performing atmospheric column thickness calculation and atmospheric scattering correction to obtain a preliminary correction image; And extracting stable characteristic points in the preliminary correction image based on angular point detection, calculating pixel corresponding positions of the stable characteristic points in a target image by taking the stable characteristic points as references, further obtaining pixel offset, and carrying out coordinate translation on the preliminary correction image based on the pixel offset to obtain a correction image.
3. The method for constructing a coastal zone landscape map based on image processing according to claim 1, wherein the performing multiband difference index calculation on the corrected image to obtain a vegetation index image and a water index image, and generating a preliminary mask image based on the vegetation index image and the water index image comprises: Extracting a near infrared band reflectance value and a red light band reflectance value of the correction image, performing difference calculation on the near infrared band reflectance value and the red light band reflectance value to obtain a vegetation value, marking the vegetation value as a vegetation region if the vegetation value is larger than a preset vegetation threshold value, and outputting the vegetation region to a vegetation index image; Extracting a green light wave band reflectance value and a near infrared wave band reflectance value of the corrected image, performing difference calculation on the green light wave band reflectance value and the near infrared wave band reflectance value of the corrected image to obtain a water body numerical value, marking the water body numerical value as a water body area if the water body numerical value is larger than a preset water body threshold value, and outputting the water body area to a water body index image; And carrying out preset weight fusion processing on the vegetation index image and the water index image to obtain a preliminary mask image.
4. The method for constructing a landscape map of a coastal zone based on image processing according to claim 1, wherein the performing region division processing on the preliminary mask image by using a watershed algorithm to obtain a candidate divided region image, and performing region water body judgment and pixel-level label fusion on the candidate divided region image to obtain a divided mask image comprises: constructing an inverted gray image based on the preliminary mask image and a watershed algorithm, and simulating a gray expansion process by taking a gray minimum value in the inverted gray image as a starting point to obtain a candidate segmentation area image; Acquiring an average water index value corresponding to each candidate region in the candidate segmented region image, outputting the candidate region into a water region image if the average water index value is larger than a preset water index threshold, otherwise outputting a potential vegetation region image; And carrying out pixel-level superposition processing and label giving according to the water body region image and the potential vegetation region image to obtain a segmentation mask image.
5. The method for constructing a coastal zone landscape map based on image processing according to claim 4, wherein the performing texture feature and color feature extraction processing on the segmentation mask image to obtain a texture feature set and a color feature set, and performing similarity calculation according to the texture feature set and the color feature set to obtain a similarity score comprises: Extracting pixel positions marked as potential vegetation areas in the segmentation mask image to obtain a potential vegetation area image; constructing a gray level co-occurrence matrix based on the potential vegetation area image, and calculating contrast and correlation according to the gray level co-occurrence matrix to obtain a texture feature set; Converting the potential vegetation area image into an HSV color space, and calculating a hue histogram and a saturation histogram according to the HSV color space to obtain a color feature set; and vector splicing is carried out on the texture feature set and the color feature set to obtain a numerical vector, and similarity calculation is carried out on the numerical vector to obtain a similarity score.
6. The method for constructing a coastal zone landscape map based on image processing according to claim 5, wherein when the similarity score is lower than a preset similarity threshold, generating a water body reflection inhibition map, and performing boundary path extraction and contour compression processing on the water body reflection inhibition map to obtain a refined boundary feature map, comprising: When the similarity score is lower than a preset similarity threshold, calculating a water body reflection value based on the potential vegetation area image to obtain a water body reflection inhibition value, and generating a graph according to the water body reflection inhibition value to obtain a water body reflection inhibition chart; according to the water body reflection inhibition graph, boundary pixel screening and path refinement processing are carried out to obtain a refined boundary path; and carrying out contour tracing on the refined boundary path to obtain a boundary closed path, and carrying out fold line compression processing on the boundary closed path to obtain a refined boundary feature map.
7. The method for constructing a coastal zone landscape map based on image processing according to claim 1, wherein the performing multi-frame difference analysis and boundary extraction processing on the refined boundary feature map to obtain a dynamic water boundary path, and performing boundary contour vectorization and vegetation distribution generation processing on the dynamic water boundary path to obtain a final vegetation distribution map comprises: According to the refined boundary feature image, multi-frame region clipping and adjacent frame reflectance difference value calculation are carried out to obtain a time sequence difference image; Carrying out dynamic change point identification and boundary path extraction processing on the time sequence difference map to obtain a dynamic water area boundary path; Sequentially tracking the boundary paths of the dynamic water area based on the eight neighborhood directions, and performing broken line compression operation to remove redundant points to obtain a dynamic water area boundary adjustment profile; And carrying out vegetation index calculation on the dynamic water area boundary adjustment profile to obtain a final vegetation index, and generating a final vegetation distribution diagram based on the final vegetation index.
8. The coastal zone landscape map construction system based on image processing is characterized by comprising: the data acquisition module is used for acquiring a coastal zone area image; The correction module is used for preprocessing according to the coastal zone region image to obtain a denoised multispectral image, and carrying out atmospheric correction and geometric registration processing on the denoised multispectral image to obtain a corrected image; the preliminary mask image module is used for carrying out multiband difference value index calculation on the correction image to obtain a vegetation index image and a water body index image, and generating a preliminary mask image based on the vegetation index image and the water body index image; the segmentation mask image module is used for carrying out region division processing on the preliminary mask image by adopting a watershed algorithm to obtain candidate segmentation region images, and carrying out region water judgment and pixel-level label fusion on the candidate segmentation region images to obtain segmentation mask images; The similarity module is used for extracting texture features and color features of the segmentation mask image to obtain a texture feature set and a color feature set, and performing similarity calculation according to the texture feature set and the color feature set to obtain a similarity score; the feature map module is used for generating a water body reflection inhibition map when the similarity score is lower than a preset similarity threshold value, and carrying out boundary path extraction and contour compression processing on the water body reflection inhibition map to obtain a refined boundary feature map; and the vegetation distribution map module is used for carrying out multi-frame difference analysis and boundary extraction processing on the refined boundary feature map to obtain a dynamic water area boundary path, and carrying out boundary contour vectorization and vegetation distribution generation processing on the dynamic water area boundary path to obtain a final vegetation distribution map.

Description

Method and system for constructing coastal zone landscape map based on image processing Technical Field The invention relates to the technical field of image recognition, in particular to a coastal zone landscape map construction method and system based on image processing. Background Along with the development of remote sensing imaging and image recognition technology, the use of multi-source remote sensing images to interpret and construct coastal zone landscapes has become an important direction in the fields of geographic information processing and ecological monitoring. Classification methods based on spectral indices are commonly used in the art, with normalized vegetation index (NDVI) being the most typical one. According to the method, the NDVI value of each pixel point is calculated by analyzing the reflectivity difference between the red light wave band and the near infrared wave band in the image so as to reflect the vegetation coverage degree. The method has a good vegetation identification effect in the static area and is suitable for rough classification in a large-scale scene. However, when the coastal zone scene is faced with frequent tides and dynamic change of the boundary of the water body, the fluctuation of the NDVI value can be not only from the real vegetation change, but also can be interfered by the influence of water body reflection, so that when a large number of areas with similar reflectivity exist, the water body is easy to be mistakenly identified as low-density vegetation, and the reality of landscape distribution is destroyed. To sum up, the prior art has the problem of misjudgment of vegetation when a large number of areas with similar reflectivity exist. Disclosure of Invention The invention provides a coastal zone landscape map construction method and system based on image processing, which are used for effectively distinguishing water from semi-submerged vegetation under the conditions of tidal change and water reflection interference. In order to solve the technical problems, the present invention provides a method for constructing a coastal zone landscape map based on image processing, comprising: acquiring a coastal zone image; Preprocessing according to the coastal zone region image to obtain a denoised multispectral image, and performing atmospheric correction and geometric registration processing on the denoised multispectral image to obtain a corrected image; Calculating the multiband difference value index of the correction image to obtain a vegetation index image and a water body index image, and generating a preliminary mask image based on the vegetation index image and the water body index image; performing region division processing on the preliminary mask image by adopting a watershed algorithm to obtain a candidate segmented region image, and performing region water judgment and pixel-level label fusion on the candidate segmented region image to obtain a segmented mask image; extracting texture features and color features of the segmentation mask image to obtain a texture feature set and a color feature set, and calculating similarity according to the texture feature set and the color feature set to obtain a similarity score; When the similarity score is lower than a preset similarity threshold, generating a water body reflection inhibition chart, and carrying out boundary path extraction and contour compression processing on the water body reflection inhibition chart to obtain a refined boundary feature chart; and carrying out multi-frame difference analysis and boundary extraction processing on the refined boundary feature map to obtain a dynamic water area boundary path, and carrying out boundary contour vectorization and vegetation distribution generation processing on the dynamic water area boundary path to obtain a final vegetation distribution map. Preferably, the preprocessing is performed according to the coastal zone region image to obtain a denoised multispectral image, and the atmospheric correction and geometric registration processing are performed on the denoised multispectral image to obtain a corrected image, which includes: performing noise suppression on the coastal zone region image by adopting median filtering to obtain a denoised multispectral image; According to the denoised multispectral image, performing atmospheric column thickness calculation and atmospheric scattering correction to obtain a preliminary correction image; And extracting stable characteristic points in the preliminary correction image based on angular point detection, calculating pixel corresponding positions of the stable characteristic points in a target image by taking the stable characteristic points as references, further obtaining pixel offset, and carrying out coordinate translation on the preliminary correction image based on the pixel offset to obtain a correction image. Preferably, the calculating the multiband difference value index of the corrected image