CN111401275B - Information processing method and device for identifying grassland edge - Google Patents
Information processing method and device for identifying grassland edge Download PDFInfo
- Publication number
- CN111401275B CN111401275B CN202010200210.0A CN202010200210A CN111401275B CN 111401275 B CN111401275 B CN 111401275B CN 202010200210 A CN202010200210 A CN 202010200210A CN 111401275 B CN111401275 B CN 111401275B
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- sensing image
- image
- regression model
- segmentation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an information processing method and device for identifying grassland edges, wherein the method comprises the steps of obtaining a first remote sensing image; preprocessing the first remote sensing image; obtaining a regression model based on the preprocessed first remote sensing image; acquiring a second remote sensing image; determining a first segmentation threshold of the second remote sensing image based on the regression model; carrying out threshold segmentation on a plurality of wave bands contained in the second remote sensing image based on the first segmentation threshold; and carrying out image fusion on the segmentation result images of the plurality of wave bands to obtain the grassland edge of the second remote sensing image. The method provided by the invention can be used for storing the details and the edge information of the image to the maximum extent, has higher extraction accuracy on the edge information, and has higher application value of the provided data; and by identifying the grassland edge information, the change rate of land desertification and soil improvement can be analyzed, and a data basis is provided for grassland ecological research and grassland area measurement.
Description
Technical Field
The invention belongs to the technical field of detection, and particularly relates to an information processing method and device for identifying grassland edges.
Background
China has abundant natural grassland resources, and the area of the national key natural grassland is about 3.37 hundred million hectares. In recent years, due to the fact that climate, relevant policies and people know the situation that ecological damage such as grassland vegetation degradation and land desertification is caused, in order to make quick and effective emergency response to grassland range change and land desertification degree, damage to the grassland ecological environment is reduced to the greatest extent, and the range change situation of grasslands must be effectively monitored in real time.
With the successful launching operation of high-resolution satellites in China, high-resolution remote sensing images are gradually used as main data sources for ecological protection, and the high-resolution remote sensing images are large in data volume, rich in information volume and high in edge density. Edge features are one of the most important features of images and are widely applied to the fields of image segmentation, image classification, image matching, pattern recognition and the like. The edge detection technology is a core part of an edge-based contour extraction algorithm, and the quality of an edge detection result directly influences the quality of edge contour extraction, and is very important for understanding grassland degradation and land desertification. Therefore, automatically and effectively identifying grass edge information is a hot spot of research.
There are some methods for identifying grass edge information, but these methods still have certain limitations. For example, 1) an improved edge detection algorithm based on the Canny algorithm, an improved method is proposed from the following aspects: processing image smoothing by using methods such as smoothing filtering, median filtering or mixed filtering; the image gradient is calculated from the gradient templates in the four directions of horizontal, vertical, 45 degrees and 135 degrees, so that the noise sensitivity is improved; the Otsu algorithm is adaptively set with high and low thresholds according to the image. However, the existing improved Canny algorithm has unsatisfactory effect in actual grassland edge detection, does not have universality, cannot automatically solve the problem of edge detection of a large number of grasslands, has large edge density and more gray changes of remote sensing images, cannot accurately segment image edges by using adaptive acquisition threshold of Otsu algorithm, and is easy to lose part of edge details.
2) Inputting a gray-scale numerical image based on a multi-structure element and multi-scale morphological edge detection algorithm, calculating multi-scale and omnibearing structure elements, calculating a morphological edge detection operator through elements, and performing opening and closing operation on the image to complete the edge detection algorithm. However, the existing multi-structure and multi-scale research methods are poor in noise resistance and edge contour accuracy, image edges in all directions cannot be well detected, detected edge contours are incomplete, details are not rich enough, and the calculation amount is large.
3) The remote sensing image artificial grassland based on the BP neural network extracts the model, and the part RGB data of the remote sensing image is used as a BP neural network training set, and the color difference characteristic is divided into the artificial grassland and the non-artificial grassland to train, and the value of the definition is set and extracted, and the part of the artificial grassland with irregular shape is extracted. However, the existing BP neural network scheme is affected by the definition of the extracted image, the data volume and the size of the image size, the clearer the extracted image is, the lower the accuracy is, and meanwhile, the uncertainty of the neural network structure also causes the reduction of the network training accuracy.
Disclosure of Invention
The invention provides an information processing method and device for identifying grassland edges. According to the information processing method for identifying the grassland edge, the threshold segmentation is carried out on the plurality of wave bands contained in the second remote sensing image, the segmentation result graphs of the plurality of wave bands are fused, the details and the edge information of the image can be stored to the maximum extent, the extraction accuracy of the edge information is high, and the provided data has high application value; and by identifying the grassland edge information, the change rate of land desertification and soil improvement can be analyzed, and a data basis is provided for grassland ecological research and grassland area measurement. The method can obtain better grassland and desert edge detection effects especially in remote sensing images containing salt and pepper and Gaussian noise, has strong self-adaptability, and is favorable for researching grassland range change.
In order to solve the above technical problem, an embodiment of the present invention provides the following technical solutions:
a first aspect of the invention provides an information processing method for identifying a grass edge, the method comprising,
acquiring a first remote sensing image;
preprocessing the first remote sensing image;
obtaining a regression model based on the preprocessed first remote sensing image;
acquiring a second remote sensing image;
determining a first segmentation threshold of the second remote sensing image based on the regression model;
performing threshold segmentation on a plurality of wave bands contained in the second remote sensing image based on the first segmentation threshold;
and carrying out image fusion on the segmentation result images of the plurality of wave bands to obtain the grassland edge of the second remote sensing image.
Preferably, the method further comprises the step of comparing the second remote sensing image with the grassland edge information with other remote sensing images with the grassland edge information to obtain the overall ecological change condition of the area corresponding to the second remote sensing image;
wherein the second remote sensing image corresponds to the same region as the other remote sensing images.
Preferably, the preprocessing the first remote sensing image comprises,
carrying out wave band synthesis on the wave band data of the first remote sensing image to form a wave band synthesis image;
deriving the wave band synthetic image into a format image;
and cutting, screening and mean value smooth filtering the format image to obtain a preprocessed first remote sensing image.
Preferably, the regression model is obtained based on the preprocessed first remote sensing image, and includes,
obtaining a gray level co-occurrence matrix and a second segmentation threshold value based on the preprocessed first remote sensing image;
establishing a regression model to be trained based on the gray level co-occurrence matrix and a third segmentation threshold;
finishing the training of the regression model to be trained based on the relation between the second segmentation threshold and the third segmentation threshold to obtain a trained regression model;
and determining the model training precision of the trained regression model to obtain the regression model.
Preferably, the gray level co-occurrence matrix comprises a plurality of feature statistics, and the training of the regression model to be trained is completed based on the relationship between the second segmentation threshold and the third segmentation threshold to obtain a trained regression model, including,
carrying out multiple combinations on the mean value and the variance of the feature statistics and bringing the combination into the regression model to be trained to obtain a plurality of third segmentation threshold values;
comparing the obtained plurality of third segmentation threshold values with the second segmentation threshold values respectively;
and determining the regression model to be trained corresponding to the third segmentation threshold value most relevant to the second segmentation threshold value as the regression model after training.
Preferably, the threshold segmentation is performed on a plurality of wave bands contained in the second remote sensing image based on the first segmentation threshold, and comprises,
analyzing a plurality of wave bands in the second remote sensing image;
and respectively carrying out threshold segmentation on the plurality of wave bands based on the first segmentation threshold.
Preferably, the image fusion of the segmentation result maps of the plurality of bands includes image fusion of the segmentation result maps of the plurality of bands by using a wavelet transform method.
A second aspect of the invention provides an information processing apparatus for identifying grass edges, the apparatus comprising at least a memory having a computer program stored thereon, a processor performing the steps of:
acquiring a first remote sensing image;
preprocessing the first remote sensing image;
obtaining a regression model based on the preprocessed first remote sensing image;
acquiring a second remote sensing image;
determining a first segmentation threshold of the second remote sensing image based on the regression model;
carrying out threshold segmentation on a plurality of wave bands contained in the second remote sensing image based on the first segmentation threshold;
and carrying out image fusion on the segmentation result images of the plurality of wave bands to obtain the grassland edge of the second remote sensing image.
Preferably, the processor further performs the steps of:
comparing the second remote sensing image with the grassland edge information with other remote sensing images with the grassland edge information to obtain the overall ecological change condition of the area corresponding to the second remote sensing image;
wherein the second remote sensing image corresponds to the same region as the other remote sensing images.
Preferably, the processor further performs the steps of:
carrying out wave band synthesis on the wave band data of the first remote sensing image to form a wave band synthesis image;
deriving the wave band synthetic image into a format image;
and cutting, screening and mean value smooth filtering the format image to obtain a preprocessed first remote sensing image.
Based on the disclosure of the above embodiments, it can be known that the embodiments of the present invention have the following beneficial effects:
according to the information processing method for identifying the grassland edge, the threshold segmentation is carried out on the plurality of wave bands contained in the second remote sensing image, the segmentation result graphs of the plurality of wave bands are fused, the details and the edge information of the image can be stored to the maximum extent, the extraction accuracy of the edge information is high, and the provided data has high application value; and by identifying the grassland edge information, the change rate of land desertification and soil improvement can be analyzed, and a data basis is provided for grassland ecological research and grassland area measurement. The method can obtain better grassland and desert edge detection effects especially in remote sensing images containing salt and pepper and Gaussian noise, has strong self-adaptability, and is favorable for researching grassland range change.
Drawings
FIG. 1 is a flow chart illustrating an information processing method for identifying grass edges in an embodiment of the present invention;
fig. 2 is a schematic structural diagram of an information processing apparatus for identifying a grass edge in an embodiment of the present invention.
Detailed Description
The following detailed description of specific embodiments of the present invention is provided in connection with the accompanying drawings, which are not intended to limit the invention.
It will be understood that various modifications may be made to the embodiments disclosed herein. Accordingly, the foregoing description should not be construed as limiting, but merely as exemplifications of embodiments. Other modifications will occur to those skilled in the art within the scope and spirit of the disclosure.
The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the disclosure and, together with a general description of the disclosure given above, and the detailed description of the embodiments given below, serve to explain the principles of the disclosure.
These and other characteristics of the invention will become apparent from the following description of a preferred form of embodiment, given as a non-limiting example, with reference to the accompanying drawings.
It should also be understood that, although the invention has been described with reference to some specific examples, a person of skill in the art shall certainly be able to achieve many other equivalent forms of the invention, having the characteristics as set forth in the claims and hence all coming within the field of protection defined thereby.
The above and other aspects, features and advantages of the present disclosure will become more apparent in view of the following detailed description when taken in conjunction with the accompanying drawings.
Specific embodiments of the present disclosure are described hereinafter with reference to the accompanying drawings; however, it is to be understood that the disclosed embodiments are merely examples of the disclosure that may be embodied in various forms. Well-known and/or repeated functions and constructions are not described in detail to avoid obscuring the disclosure in unnecessary or unnecessary detail. Therefore, specific structural and functional details disclosed herein are not to be interpreted as limiting, but merely as a basis for the claims and as a representative basis for teaching one skilled in the art to variously employ the present disclosure in virtually any appropriately detailed structure.
The specification may use the phrases "in one embodiment," "in another embodiment," "in yet another embodiment," or "in other embodiments," which may each refer to one or more of the same or different embodiments in accordance with the disclosure.
The embodiments of the present invention will be described in detail below with reference to the accompanying drawings,
as shown in fig. 1, a first embodiment of the present invention provides an information processing method for identifying grass edges, the method comprising,
s1, acquiring a first remote sensing image;
s2, preprocessing the first remote sensing image;
s3, obtaining a regression model based on the preprocessed first remote sensing image;
s4, acquiring a second remote sensing image;
s5, determining a first segmentation threshold value of the second remote sensing image based on the regression model;
s6, threshold segmentation is carried out on a plurality of wave bands contained in the second remote sensing image based on the first segmentation threshold;
and S7, carrying out image fusion on the segmentation result images of the plurality of wave bands to obtain the grassland edge of the second remote sensing image.
In this embodiment, the first remote sensing image and the second remote sensing image are different and can be from a high-resolution satellite respectively, and the regression model is mainly obtained based on the first remote sensing image, so as to further identify the grassland edge information of the second remote sensing image. Specifically, after the regression model is obtained based on the first remote sensing image, a first segmentation threshold value of the second remote sensing image is determined based on the regression model; then, threshold segmentation is carried out on a plurality of wave bands contained in the second remote sensing image based on the first segmentation threshold; and finally, carrying out image fusion on the segmentation result images of the plurality of wave bands to obtain the grassland edge of the second remote sensing image.
According to the information processing method for identifying the grassland edge, the threshold segmentation is carried out on the plurality of wave bands contained in the second remote sensing image, the segmentation result graphs of the plurality of wave bands are fused, the details and the edge information of the image can be stored to the maximum extent, the extraction accuracy of the edge information is high, and the provided data has high application value; and by identifying the grassland edge information, the change rate of land desertification and soil improvement can be analyzed, and data basis is provided for grassland ecological research and grassland area measurement. The method can obtain better grassland and desert edge detection effects especially in the remote sensing images containing salt and pepper and Gaussian noise, has strong adaptability, and is beneficial to the research on the grassland range change.
In another embodiment provided by the invention, the method further comprises the steps of comparing the second remote sensing image with the grassland edge information with other remote sensing images with the grassland edge information to obtain the overall ecological change condition of the area corresponding to the second remote sensing image;
wherein the second remote sensing image corresponds to the same region as the other remote sensing images.
In this embodiment, the second remote sensing image with grassland edge information may be compared with other remote sensing images with grassland edge information to obtain the overall ecological change condition of the area corresponding to the second remote sensing image. For example, the frequency of the detection of the grass edge information may be once a year, and by comparing the remote sensing image with grass edge information obtained this year with the remote sensing image with grass edge information obtained last year (the remote sensing image with grass edge information obtained this year and the remote sensing image with grass edge information obtained last year correspond to the same area), the ecological change of the area corresponding to the remote sensing image can be known, for example, the grass area is increased or decreased this year compared with the last year.
In a specific embodiment, the second remote sensing image with the grassland edge information may be regarded as obtained this year, and other remote sensing images with grassland edge information may be regarded as obtained last year, for example, obtained last year, and so on, which is not limited in this disclosure. For convenience of description, in the following embodiments of the present invention, the description is given by taking the example obtained in the past year for the last year, that is, the second remote sensing image with the grassland edge information is compared with the remote sensing image with the grassland edge information obtained in the past year.
For example, 1) performing exclusive or on the remote sensing image with grassland edge information of this year and the remote sensing image with grassland edge information of the past year to obtain the whole ecological change condition (whole change rate) of this year compared with the past year, including the whole conditions of land desertification and soil improvement;
for example, the exclusive or between 2013 and 2012 reflects the overall change, including the rate of land desertification and soil improvement.
2) Subtracting the remote sensing image with the grassland edge information of the past year from the remote sensing image with the grassland edge information of the present year to obtain the ecological change condition (namely the change rate of the desertification of the land) of the desertification of the land of the present year, namely the change rate of the grassland of the last year to the sandy land of the present year;
for example, the subtraction of 2013 from 2012 reflects the rate of land desertification, i.e., the rate of change of grass in 2012 to sand in 2013.
3) Subtracting the remote sensing image with the grassland edge information of the current year from the remote sensing image with the grassland edge information of the previous year to obtain the ecological change condition (namely the soil improvement change rate) of soil improvement of the current year, namely the sand of the last year becomes grassland in the current year; or subtracting the soil desertification change rate from the integral change rate obtained by XOR to obtain the soil improvement change rate;
for example, the marginal information of year 2012 minus year 2013 reflects the rate of soil improvement, i.e., the rate of change of sand to grass in year 2012 in 2013.
In the embodiment of the invention, a logical relationship is established by using a remote sensing image with grassland edge information to analyze the ecological condition; and by combining edge information arrangement, comparison and analysis in the past year, the dynamic change trend of the land desertification in the research area is mastered, and technical and theoretical support is provided for formulating policies for preventing and controlling desertification and long-term development planning.
The change rate of land desertification and soil improvement can be analyzed through annual edge information change, the edge change is analyzed, the range change condition of the desert and the grassland can be obtained every year, the land desertification is intensified or improved, the implementation effect of the current control, protection, improvement and other related measures is evaluated, the policy and the development plan for more effectively preventing and controlling the desertification are conveniently provided for the research area, and the technical support is provided for the understanding of the related departments on the ecological condition; and the policy and the development plan for more effectively preventing and controlling the desertification are convenient to put forward.
In other embodiments of the present invention, the preprocessing the first remote sensing image includes,
carrying out wave band synthesis on the wave band data of the first remote sensing image to form a wave band synthesized image;
exporting the wave band synthetic image as a format image;
and cutting, screening and mean value smooth filtering the format image to obtain a preprocessed first remote sensing image.
In this embodiment, the specific process of preprocessing the first remote sensing image is mainly used. The first remote sensing image comprises a plurality of wave band data, and the wave band data of the first remote sensing image can be subjected to wave band synthesis to form a wave band synthesis image; for example, only the data of the red, green, and blue three bands of the first remote sensing image may be band-synthesized, specifically, the band data may be band-synthesized in real color by the arcbios 10.2 software in the order of red, green, and blue (RGB) to form a band-synthesized image. Then exporting the wave band synthetic image in a certain format to form a format image; specifically, the band-pass synthesized image may be represented by a 1: the proportion of 100000 is derived as the TIF format, and a format image is formed. And finally, cutting, screening and mean value smoothing filtering the format image to obtain a preprocessed first remote sensing image. Specifically, the derived TIF image was cropped using MATLAB2016a, cropping the image size to 128 x 128; screening the cut image, deleting the white image, and reserving the image containing the image information; and performing mean smoothing filtering on the screened image, setting the values of the filtered space bandwidth and the filtered chromaticity bandwidth to be hs =20 and hr =20 respectively, and performing mean smoothing filtering to obtain a preprocessed first remote sensing image. The invention uses the average value smoothing filtering, can effectively inhibit the trivial details in the image and retain the edge information, and eliminates the interference of non-target pixels, namely noise pixels.
In an embodiment of the present invention, the obtaining a regression model based on the preprocessed first remote sensing image includes,
obtaining a gray level co-occurrence matrix and a second segmentation threshold value based on the preprocessed first remote sensing image;
establishing a regression model to be trained based on the gray level co-occurrence matrix and a third segmentation threshold;
finishing the training of the regression model to be trained based on the relation between the second division threshold and the third division threshold to obtain a trained regression model;
and determining the model training precision of the trained regression model to obtain the regression model.
In a specific embodiment, the gray level co-occurrence matrix includes a plurality of feature statistics, and the training of the regression model to be trained is completed based on the relationship between the second segmentation threshold and the third segmentation threshold, so as to obtain a trained regression model, including,
carrying out multiple combinations on the mean value and the variance of the feature statistics and bringing the combination into the regression model to be trained to obtain a plurality of third segmentation threshold values;
comparing the obtained third segmentation threshold values with the second segmentation threshold values respectively;
and determining the regression model to be trained corresponding to the third segmentation threshold value most relevant to the second segmentation threshold value as the regression model after training.
In this embodiment, the feature statistics included in the gray level co-occurrence matrix may be, for example, homogeneity (Homogeneity), contrast (Contrast), dissimilarity (similarity), entropy (Entropy), angular Second Moment (Angular Second Moment), correlation (Correlation), and the like. The mean and variance ratio of the feature statistics are the mean of homogeneity, the variance of homogeneity, the mean of contrast, the variance of contrast, the mean of non-similarity, the variance of non-similarity, the mean of entropy, the variance of entropy, the mean of angular second moment, the variance of angular second moment, the mean of correlation and the variance of correlation, etc., and the mean and variance ratio of the feature statistics can be extracted from the gray level co-occurrence matrix using python3.6 software.
A regression model to be trained is established based on the gray level co-occurrence matrix and the third segmentation threshold value such as,
k i ＝β 0 +β 1 x i1 +...+β p x ip +ε i ,i＝1,2,...,n
wherein, k is i Is a third division threshold, β 0 ，β 1 ，...，β p Is a coefficient of ∈ i Is an error, x i1 ,x i2 ,...,x ip Respectively, the mean or variance of the number of feature statistics, i being the total number.
And carrying out various combinations on the mean value and the variance of the characteristic statistics and bringing the combinations into the regression model to be trained to obtain a plurality of third segmentation threshold values. For example, in one embodiment, the variance of homogeneity, the mean of contrast, the mean of entropy, the variance of entropy, and the mean of correlation are taken as a first combination into the regression model to be trained, resulting in a third segmentation threshold (1); taking the mean value of the contrast, the variance of the contrast, the mean value of the non-similarity, the variance of the non-similarity and the mean value of the angular second moment as a second combination to be brought into the regression model to be trained to obtain a third segmentation threshold (2); taking the mean value of homogeneity, the mean value of non-similarity, the variance of entropy, the mean value of angular second moment and the mean value of correlation as a third combination to be brought into the regression model to be trained to obtain a third segmentation threshold (3); and taking the mean value of homogeneity, the variance of correlation, the variance of angular second moment and the mean value of contrast as a fourth combination to be brought into the regression model to be trained to obtain a third segmentation threshold (4) and the like, respectively comparing different third segmentation thresholds such as the obtained third segmentation threshold (1), the obtained third segmentation threshold (2), the obtained third segmentation threshold (3) and the obtained third segmentation threshold (4) with the obtained second segmentation threshold, and determining the regression model to be trained corresponding to the third segmentation threshold most correlated with the second segmentation threshold as the regression model after training, thus obtaining the regression model after training. And finally, determining the model training precision of the trained regression model to obtain the regression model. For example, the Root Mean Square Error (RMSE) may be used to determine the model training accuracy, which reflects how far a number of different third segmentation thresholds deviate from the second segmentation threshold, with smaller values of the root mean square error indicating higher measurement accuracy, and the regression model being obtained when the accuracy meets a predetermined criterion. The root mean square error formula is as follows:
wherein n is the number of the third threshold, X obs,i Representing a second division threshold, X model，i Indicating a third segmentation threshold.
For example, in one embodiment, after comparing the different third segmentation threshold with the second segmentation threshold, it is determined that the third segmentation threshold (4) is most correlated with the second segmentation threshold, and it is determined that the model training precision meets the preset standard by using the root mean square error, the mean value and the variance of the feature statistics (mean value of homogeneity, variance of correlation, variance of angular second moment, mean value of contrast) corresponding to the third segmentation threshold (4) are X in the regression model. That is, in the regression model, based on the mean homogeneity, the variance correlation, the variance angular second moment, and the mean contrast of the remote sensing image, the segmentation threshold corresponding to the remote sensing image can be obtained for the remote sensing image. For example, the regression model determined is:
K＝0.2016-0.06075X 1 -0.38878X 2 +0.43*10 -5 X 3 +0.47623*10 -8 X 4 -0.041063X 5
wherein: x 1 (mean of homogeneity), X 2 (variance of homogeneity), X 3 (variance of correlation), X 4 (variance of angular second moment), X 5 (average of contrast).
Specifically, a first segmentation threshold corresponding to the second remote sensing image may be determined for the second remote sensing image based on a mean value of homogeneity, a variance of correlation, a variance of angular second moment, and a mean value of contrast of the second remote sensing image.
In other embodiments provided by the present invention, the threshold segmentation of the plurality of wavelength bands included in the second remote sensing image based on the first segmentation threshold includes,
analyzing a plurality of wave bands in the second remote sensing image;
and respectively carrying out threshold segmentation on the plurality of wave bands based on the first segmentation threshold.
In this embodiment, after the regression model is obtained, the mean value of the homogeneity, the variance of the correlation, the variance of the angular second moment, and the mean value of the contrast of the second remote sensing image are extracted, and the values of the mean value of the homogeneity, the variance of the correlation, the variance of the angular second moment, and the mean value of the contrast are brought into the determined regression model, so that the first segmentation threshold of the second remote sensing image can be obtained.
The second remote sensing image comprises a plurality of wave bands, and the wave bands in the second remote sensing image are analyzed, for example, RGB three wave bands in the second remote sensing image are analyzed. And performing threshold segmentation on the three RGB wave bands respectively based on the first segmentation threshold to obtain segmentation result graphs respectively, and performing image fusion on the three segmentation result graphs obtained after threshold segmentation on the three different RGB wave bands to obtain the grassland edge of the second remote sensing image. For example, in a specific embodiment, the segmentation result maps of the several bands may be image-fused by using a wavelet transform method.
And respectively carrying out threshold segmentation on the RGB wave bands according to the first segmentation threshold predicted by the regression model. The method is mainly used for distinguishing the edges of the grassland and the desert, and pixel points F (x, y) of images in three wave bands of RGB are divided into 2 types through a first segmentation threshold value predicted by a regression model. The threshold K is in the range of [0,0.255], where the grassland is composed of all pixels in the image with the band value in the range of [0, K ] and displayed as black in the segmented image, the desert is composed of all pixels with the band value in the range of [ K +1,0.255], white is displayed in the segmented image, and the 3-band segmented image is saved.
In one embodiment, for example, the first segmentation threshold calculated by the regression model is 0.17, and the R-band image is segmented by using the first segmentation threshold to obtain a first segmentation result map; in the first segmentation result graph, classifying pixel points which are less than or equal to 0.17 in the R-band image into one class, regarding the pixel points as grasslands, and displaying the pixel points as black in the first segmentation result graph; and classifying pixel points larger than 0.17 in the R-band image into one type, and displaying the pixel points as a desert in the first segmentation result image. Segmenting the G-band image by using a first segmentation threshold value to obtain a second segmentation result graph; in the second segmentation result graph, pixel points which are less than or equal to 0.17 in the G-band image are classified as grasslands, and the grasslands are displayed as black in the second segmentation result graph; and classifying pixel points larger than 0.17 in the G-band image into one type, and displaying the pixel points as a desert in the second segmentation result image. Segmenting the B-band image by using a first segmentation threshold value to obtain a third segmentation result graph; in the third segmentation result graph, pixel points which are less than or equal to 0.17 in the B-band image are classified as grasslands, and the grasslands are displayed as black in the third segmentation result graph; and classifying pixel points larger than 0.17 in the B-band image into one type, regarding the pixel points as deserts, and displaying the desert in the third segmentation result image. And finally, carrying out image fusion on the first segmentation result graph, the second segmentation result graph and the third segmentation result graph by adopting a wavelet transform method to obtain the grassland edge of the second remote sensing image.
Based on the same inventive concept, as shown in fig. 2, a second embodiment of the present invention provides an information processing apparatus 01 for identifying grass edges, the apparatus at least comprising a memory 001 and a processor 002, the memory storing thereon a computer program, the processor executing the following steps:
acquiring a first remote sensing image;
preprocessing the first remote sensing image;
obtaining a regression model based on the preprocessed first remote sensing image;
acquiring a second remote sensing image;
determining a first segmentation threshold of the second remote sensing image based on the regression model;
performing threshold segmentation on a plurality of wave bands contained in the second remote sensing image based on the first segmentation threshold;
and carrying out image fusion on the segmentation result images of the plurality of wave bands to obtain the grassland edge of the second remote sensing image.
In one embodiment provided by the present invention, the processor further performs the following steps:
comparing the second remote sensing image with the grassland edge information with other remote sensing images with the grassland edge information to obtain the overall ecological change condition of the area corresponding to the second remote sensing image;
wherein the second remote sensing image and the other remote sensing images correspond to the same region.
In other embodiments provided by the present invention, the processor further performs the steps of:
carrying out wave band synthesis on the wave band data of the first remote sensing image to form a wave band synthesis image;
exporting the wave band synthetic image as a format image;
and cutting, screening and mean value smooth filtering the format image to obtain a preprocessed first remote sensing image.
The above embodiments are only exemplary embodiments of the present invention, and are not intended to limit the present invention, and the scope of the present invention is defined by the claims. Various modifications and equivalents may be made by those skilled in the art within the spirit and scope of the present invention, and such modifications and equivalents should also be considered as falling within the scope of the present invention.
Claims (8)
1. An information processing method for identifying grass edges, the method comprising,
acquiring a first remote sensing image;
preprocessing the first remote sensing image;
obtaining a regression model based on the preprocessed first remote sensing image, including,
obtaining a gray level co-occurrence matrix and a second segmentation threshold value based on the preprocessed first remote sensing image;
establishing a regression model to be trained based on the gray level co-occurrence matrix and a third segmentation threshold;
the gray level co-occurrence matrix comprises a plurality of characteristic statistics, the training of the regression model to be trained is completed based on the relation between the second segmentation threshold and the third segmentation threshold, and a trained regression model is obtained, which comprises,
carrying out multiple combinations on the mean value and the variance of the feature statistics and bringing the combination into the regression model to be trained to obtain a plurality of third segmentation threshold values;
comparing the obtained plurality of third segmentation threshold values with the second segmentation threshold values respectively;
determining a regression model to be trained corresponding to a third segmentation threshold value most relevant to the second segmentation threshold value as the trained regression model;
determining the model training precision of the trained regression model to obtain the regression model;
acquiring a second remote sensing image;
determining a first segmentation threshold of the second remote sensing image based on the regression model;
performing threshold segmentation on a plurality of wave bands contained in the second remote sensing image based on the first segmentation threshold;
and carrying out image fusion on the segmentation result images of the plurality of wave bands to obtain the grassland edge of the second remote sensing image.
2. The method of claim 1, further comprising comparing the second remote sensing image with the grassland edge information with other remote sensing images with the grassland edge information to obtain the overall ecological change condition of the area corresponding to the second remote sensing image;
wherein the second remote sensing image and the other remote sensing images correspond to the same region.
3. The method of claim 1, wherein the pre-processing the first remotely sensed image comprises,
carrying out wave band synthesis on the wave band data of the first remote sensing image to form a wave band synthesized image;
deriving the wave band synthetic image into a format image;
and cutting, screening and mean value smoothing filtering are carried out on the format image to obtain a preprocessed first remote sensing image.
4. The method of claim 1, wherein thresholding several bands contained in the second remotely sensed image based on the first segmentation threshold comprises,
analyzing a plurality of wave bands in the second remote sensing image;
and respectively carrying out threshold segmentation on the plurality of wave bands based on the first segmentation threshold.
5. The method according to claim 1, wherein the image fusing the segmentation result maps of the plurality of bands comprises image fusing the segmentation result maps of the plurality of bands by using a wavelet transform method.
6. An information processing apparatus for identifying grass edges, the apparatus comprising at least a memory having a computer program stored thereon, a processor performing the steps of:
acquiring a first remote sensing image;
preprocessing the first remote sensing image;
obtaining a regression model based on the preprocessed first remote sensing image, including,
obtaining a gray level co-occurrence matrix and a second segmentation threshold value based on the preprocessed first remote sensing image;
establishing a regression model to be trained based on the gray level co-occurrence matrix and a third segmentation threshold;
the gray level co-occurrence matrix comprises a plurality of characteristic statistics, the training of the regression model to be trained is completed based on the relation between the second segmentation threshold and the third segmentation threshold, and a trained regression model is obtained, which comprises,
carrying out multiple combinations on the mean value and the variance of the feature statistics and bringing the combination into the regression model to be trained to obtain a plurality of third segmentation threshold values;
comparing the obtained third segmentation threshold values with the second segmentation threshold values respectively;
determining a regression model to be trained corresponding to a third segmentation threshold value most relevant to the second segmentation threshold value as the regression model after training;
determining the model training precision of the trained regression model to obtain the regression model;
acquiring a second remote sensing image;
determining a first segmentation threshold of the second remote sensing image based on the regression model;
performing threshold segmentation on a plurality of wave bands contained in the second remote sensing image based on the first segmentation threshold;
and carrying out image fusion on the segmentation result images of the plurality of wave bands to obtain the grassland edge of the second remote sensing image.
7. The apparatus of claim 6, wherein the processor further performs the steps of:
comparing the second remote sensing image with the grassland edge information with other remote sensing images with the grassland edge information to obtain the overall ecological change condition of the area corresponding to the second remote sensing image;
wherein the second remote sensing image corresponds to the same region as the other remote sensing images.
8. The apparatus of claim 6, wherein the processor further performs the steps of:
carrying out wave band synthesis on the wave band data of the first remote sensing image to form a wave band synthesis image;
exporting the wave band synthetic image as a format image;
and cutting, screening and mean value smooth filtering the format image to obtain a preprocessed first remote sensing image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010200210.0A CN111401275B (en) | 2020-03-20 | 2020-03-20 | Information processing method and device for identifying grassland edge |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010200210.0A CN111401275B (en) | 2020-03-20 | 2020-03-20 | Information processing method and device for identifying grassland edge |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111401275A CN111401275A (en) | 2020-07-10 |
| CN111401275B true CN111401275B (en) | 2022-11-25 |
Family
ID=71432702
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010200210.0A Active CN111401275B (en) | 2020-03-20 | 2020-03-20 | Information processing method and device for identifying grassland edge |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111401275B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113421277B (en) * | 2021-08-25 | 2021-12-14 | 中科星图股份有限公司 | Road extraction and abnormity monitoring method and device based on remote sensing image |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103578110A (en) * | 2013-11-12 | 2014-02-12 | 河海大学 | Multi-band high-resolution remote sensing image segmentation method based on gray scale co-occurrence matrix |
| CN110211138A (en) * | 2019-06-08 | 2019-09-06 | 西安电子科技大学 | Remote sensing image segmentation method based on confidence point |
| CN110879992A (en) * | 2019-11-27 | 2020-03-13 | 内蒙古工业大学 | Grassland surface covering object classification method and system based on transfer learning |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| SG10201405182WA (en) * | 2014-08-25 | 2016-03-30 | Univ Singapore Technology & Design | Method and system |
| CN110008908A (en) * | 2019-04-10 | 2019-07-12 | 中国科学院、水利部成都山地灾害与环境研究所 | A kind of grassland fences extracting method based on high score remote sensing image |
-
2020
- 2020-03-20 CN CN202010200210.0A patent/CN111401275B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103578110A (en) * | 2013-11-12 | 2014-02-12 | 河海大学 | Multi-band high-resolution remote sensing image segmentation method based on gray scale co-occurrence matrix |
| CN110211138A (en) * | 2019-06-08 | 2019-09-06 | 西安电子科技大学 | Remote sensing image segmentation method based on confidence point |
| CN110879992A (en) * | 2019-11-27 | 2020-03-13 | 内蒙古工业大学 | Grassland surface covering object classification method and system based on transfer learning |
Non-Patent Citations (4)
| Title |
|---|
| An improved segmentation of high spatial resolution remote sensing image using Marker-based Watershed Algorithm;Boren Li 等;《2012 20th International Conference on Geoinformatics》;20120820;第1-5页 * |
| 基于灰度共生矩阵的多波段遥感图像纹理特征的提取;田艳琴 等;《计算机科学》;20041231;第31卷(第12期);第162-163页 * |
| 遥感影像植被边界提取方法研究;苏晓军;《中国优秀硕士学位论文全文数据库 信息科技辑》;20140215;I140-463 * |
| 针对高分四号卫星影像的边缘检测技术;郭敬 等;《遥感信息》;20180630;第33卷(第3期);第108-115页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111401275A (en) | 2020-07-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhao et al. | Multi-scale optimal fusion model for single image dehazing | |
| CN110309781B (en) | House damage remote sensing identification method based on multi-scale spectrum texture self-adaptive fusion | |
| US11094040B2 (en) | Noise detection method for time-series vegetation index derived from remote sensing images | |
| Liang et al. | Material based salient object detection from hyperspectral images | |
| CN107330875B (en) | Water body surrounding environment change detection method based on forward and reverse heterogeneity of remote sensing image | |
| CN109154978A (en) | System and method for detecting plant disease | |
| US20120213440A1 (en) | Systems and Methods for Automatically Identifying Shadows in Images | |
| CN104598908A (en) | Method for recognizing diseases of crop leaves | |
| CN111507426A (en) | No-reference image quality grading evaluation method and device based on visual fusion characteristics | |
| CN106960182A (en) | A kind of pedestrian integrated based on multiple features recognition methods again | |
| Masood et al. | Plants disease segmentation using image processing | |
| US11475707B2 (en) | Method for extracting image of face detection and device thereof | |
| Ji et al. | In-field automatic detection of maize tassels using computer vision | |
| CN111259925A (en) | Method for counting field wheat ears based on K-means clustering and width mutation algorithm | |
| Almaadeed et al. | Partial shoeprint retrieval using multiple point-of-interest detectors and SIFT descriptors | |
| CN108492288B (en) | Random forest based multi-scale layered sampling high-resolution satellite image change detection method | |
| CN115690086A (en) | Object-based high-resolution remote sensing image change detection method and system | |
| CN103218819A (en) | Automatic selection method for optimal homogenous region of ultrasound image and based on quad-tree decomposition | |
| CN111401275B (en) | Information processing method and device for identifying grassland edge | |
| Kumar et al. | Comparative analysis for edge detection techniques | |
| CN116468958B (en) | Communication tower safety detection method and system | |
| CN111882573A (en) | Cultivated land plot extraction method and system based on high-resolution image data | |
| Ouattara et al. | A new image inpainting approach based on Criminisi algorithm | |
| CN112541471B (en) | Multi-feature fusion-based shielding target identification method | |
| Huang et al. | Classification of very high spatial resolution imagery based on the fusion of edge and multispectral information |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |