CN108846832B - Multi-temporal remote sensing image and GIS data based change detection method and system - Google Patents
Multi-temporal remote sensing image and GIS data based change detection method and system Download PDFInfo
- Publication number
- CN108846832B CN108846832B CN201810538861.3A CN201810538861A CN108846832B CN 108846832 B CN108846832 B CN 108846832B CN 201810538861 A CN201810538861 A CN 201810538861A CN 108846832 B CN108846832 B CN 108846832B
- Authority
- CN
- China
- Prior art keywords
- pattern
- period
- water body
- vegetation
- spots
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a change detection method and system based on multi-temporal remote sensing images and GIS data, and aims to solve the problems of the multi-temporal remote sensing images and the GIS data in the segmentation, classification and change detection methods. The method comprises the following steps: step 1, remote sensing image segmentation and pixel level change information extraction, which mainly comprises remote sensing image segmentation and pixel level change detection; step 2, classifying the object-oriented remote sensing images, wherein the classification mainly comprises a change pattern spot decision, automatic sample selection and an object-oriented SVM decision tree classification model; and 3, detecting the change after object-oriented classification. Compared with the existing method for extracting the change information by using the remote sensing image and the GIS data, the method can realize the automatic change information detection method flow applicable to a large area, can fully mine the correlation between the remote sensing image and the GIS data, and improves the accuracy of the change information extraction by using the attribute information of the GIS data, such as the category and the like.
Description
Technical Field
The invention relates to the field of remote sensing and Geographic Information Systems (GIS), in particular to a method for automatically detecting change Information of remote sensing images and GIS data, and particularly relates to a method for automatically detecting object-oriented change Information by utilizing multi-temporal remote sensing images and GIS data.
Background
In the field of remote sensing, change detection is a process of obtaining a change area by comparing differences of characteristics such as spectral reflectance values, textures and the like by using remote sensing images of the same area in different time phases. The remote sensing image is used for change detection, and is widely applied to the fields of environment monitoring, agricultural investigation, city expansion research, disaster monitoring, damage assessment and the like. In the GIS field, GIS data is a symbolic representation of surface features, geomorphology and terrain, and is the result of field measurements and image interpretation. The GIS data not only contains geometric position information, but also contains semantic information such as abundant ground feature types, attributes and the like. With the development of remote sensing technology and the wide application of GIS in the fields of environment, agriculture, city, disaster relief and reduction, geographical national conditions and the like, which relate to the spatial information industry, the extraction of change information by using remote sensing images and GIS data is more and more concerned.
The traditional method for extracting change information by using remote sensing images and GIS data mainly removes unchanged areas by using a change detection method based on an image processing technology through new and old two-time phase remote sensing images to generate an image consisting of pixels which are likely to change. And carrying out grid vectorization processing on the change image, and carrying out superposition analysis on the change image and GIS data to obtain a final change detection result. The method combines the remote sensing image change detection technology and the GIS analysis technology, and the detection precision of the method depends on the result of the remote sensing image change detection. Because the method does not fully utilize the category attribute information of the GIS data, the utilization rate of the correlation between the remote sensing image and the GIS data is low, and the accuracy of extracting the change information is difficult to ensure.
Disclosure of Invention
Aiming at the defects of the traditional change information extraction method based on multi-temporal remote sensing images and GIS data, the invention aims to provide an automatic and large-area applicable multi-temporal remote sensing image and GIS data change information detection method flow.
The technical scheme adopted by the invention is as follows: a change detection method based on multi-temporal remote sensing images and GIS data specifically comprises the following steps:
step 1, remote sensing image segmentation and pixel level change information extraction, comprising the following substeps;
step 1.1, according to GIS vector data in the T1 period and grid data in the T2 period, utilizing GIS vector data pattern spot type attribute information and geometric shape information to register the grid data in the T2 period to obtain a grid pixel set corresponding to a vector pattern spot in the GIS data in the T1 period;
step 1.2, performing subdivision on each obtained grid pixel set by using a multi-scale division algorithm to obtain a T2-period image spot, namely a T2-period image division result;
step 1.2, carrying out change detection on the raster data in the T1 period and the T2 period by using a pixel-level change detection algorithm to obtain a pixel-level change detection result;
step 2, classifying the remote sensing images facing the object, and comprising the following substeps;
step 2.1, performing superposition analysis on the image segmentation result and the pixel level change detection result in the period T2, counting the number of changed pixels and unchanged pixels in each image spot, and when the ratio of the number of the changed pixels to the number of the unchanged pixels in each image spot is greater than a threshold value m, determining that the image spot is a suspected changed image spot, otherwise, determining that the image spot is an unchanged image spot;
step 2.2, performing feature statistics on each obtained unchanged pattern spot in the T2 period, and selecting different feature combinations according to specific object types to obtain a proper training sample;
step 2.3, constructing an object-oriented SVM decision tree classification model, training the SVM decision tree classification model through the training samples in the step 2.2, and classifying all suspected change patches by using the trained model to obtain the classification result of each patch;
and 3, carrying out object-oriented classified change detection by using the types of the vector pattern spots in the GIS data in the period T1 and the classified pattern spots of the suspected change pattern spots in the period T2 to obtain the pattern spots with inconsistent types in the two periods and the type attribute information thereof, namely the final change detection result.
Further, the object-oriented SVM decision tree classification model constructed in the step 2.3 is used for more accurately dividing 5 types of ground objects, namely water, garden ploughing grass, forest land, artificial structures and bare land, and the specific implementation manner is as follows,
step 2.3.1, dividing the pattern spots in the period T2 into water body pattern spots and non-water body pattern spots according to a certain threshold value for the normalized water body index characteristic value of each pattern spot;
step 2.3.2, selecting the mean value and the variance in the spectral characteristics of the water body pattern spots obtained in the step 2.3.1, and performing reclassification by using an SVM classification model to obtain more accurate water body pattern spots and non-water body pattern spots;
step 2.3.3, dividing the normalized vegetation index characteristic value of the non-water body pattern spots obtained in the step 2.3.1 and the step 2.3.2 into vegetation pattern spots and non-vegetation pattern spots according to a certain threshold value;
step 2.3.4, selecting the mean value and the variance in the spectral characteristics of the vegetation pattern spots obtained in the step 2.3.3, and classifying by using an SVM classification model to obtain garden cultivation and grass pattern spots and forest land pattern spots;
and 2.3.5, selecting the mean value and the variance in the spectral characteristics of the non-vegetation patches obtained in the step 2.3.3, and classifying by using an SVM classification model to obtain artificial structure patches and bare area patches.
Further, the pixel level variation detection algorithm in step 1.2 is a k-means variation detection algorithm.
Further, in step 2.3.1, a k-means clustering algorithm is used for carrying out clustering analysis on the normalized water body index of the pattern spots in the T2 period to obtain threshold values of the water body pattern spots and the non-water body pattern spots; and 2.3.3, carrying out clustering analysis on the normalized vegetation index of the non-water body pattern spots by using a k-means clustering algorithm to obtain the vegetation pattern spots and the threshold values of the non-vegetation pattern spots.
The invention also provides an object-oriented change information automatic detection system based on the multi-temporal remote sensing image and GIS data, which comprises the following modules:
the image segmentation and change information extraction module is used for extracting remote sensing image segmentation and pixel level change information and comprises the following sub-modules;
the grid vector data registration submodule is used for registering the grid data in the T2 period by utilizing the GIS vector data pattern spot type attribute information and the geometric shape information according to the GIS vector data in the T1 period and the grid data in the T2 period to obtain a grid pixel set corresponding to the vector pattern spot in the GIS data in the T1 period, namely a pattern spot in the T2 period;
the registration data re-segmentation submodule is used for performing re-segmentation on each acquired T2-period image spot by using a multi-scale segmentation algorithm to obtain sub image spots of each image spot, namely a T2-period image segmentation result;
the pixel level change detection submodule is used for carrying out change detection on the raster data in the T1 period and the T2 period by utilizing a pixel level change detection algorithm to obtain a pixel level change detection result;
the remote sensing image classification module is used for object-oriented remote sensing image classification and comprises the following sub-modules;
the change pattern spot decision submodule is used for performing superposition analysis on the image segmentation result and the pixel level change detection result in the period T2, counting the number of changed pixels and unchanged pixels in each pattern spot, and when the ratio of the number of the changed pixels to the number of the unchanged pixels in each pattern spot is greater than a threshold value m, the pattern spot is a suspected change pattern spot, otherwise, the pattern spot is an unchanged pattern spot;
the sample selection submodule is used for carrying out feature statistics on each obtained unchanged pattern spot in the T2 period, and selecting different feature combinations according to specific object types to obtain a proper training sample;
the SVM decision tree classification submodule is used for constructing an object-oriented SVM decision tree classification model, training the SVM decision tree classification model through training samples in the sample selection submodule, and classifying all suspected change pattern spots by using the trained model to obtain a classification result of each pattern spot;
and the change detection module is used for carrying out object-oriented classified change detection by utilizing the types of the vector pattern spots in the GIS data in the period T1 and the classified pattern spots of suspected change pattern spots in the period T2 to obtain the pattern spots with inconsistent types in the two periods and the type attribute information thereof, namely the final change detection result.
Furthermore, an object-oriented SVM decision tree classification model constructed in the SVM decision tree classification submodule is used for more accurately dividing 5 types of ground objects, namely water, ploughed garden grass, forest land, artificial structures and bare land, and specifically comprises the following units,
the water body threshold value unit is used for dividing the pattern spots in the T2 period into water body pattern spots and non-water body pattern spots according to a certain threshold value for the normalized water body index characteristic value of each pattern spot;
the water body SVM classification unit is used for selecting the mean value and the variance in the spectral characteristics of the water body image spots obtained in the water body threshold value unit, and reclassifying by using an SVM classification model to obtain more accurate water body image spots and non-water body image spots;
the vegetation threshold unit is used for dividing the normalized vegetation index characteristic value of the non-water body pattern spots obtained in the water body threshold unit and the water body SVM classification unit into vegetation pattern spots and non-vegetation pattern spots according to a certain threshold;
the vegetation SVM classification unit is used for selecting the mean value and the variance in the spectral characteristics of the vegetation patches obtained in the vegetation threshold value unit, and classifying the vegetation patches by using an SVM classification model to obtain garden tillage patches and forest land patches;
and the non-vegetation SVM classification unit is used for selecting the mean value and the variance in the spectral characteristics of the non-vegetation patches obtained in the vegetation threshold value unit, and classifying the non-vegetation patches by using an SVM classification model to obtain artificial structure patches and bare area patches.
Furthermore, a pixel level change detection algorithm in the registration data subdividing sub-module is a k-means change detection algorithm.
Further, a k-means clustering algorithm is used in the water body threshold value unit to perform clustering analysis on the normalized water body index of the pattern spots in the T2 period, so as to obtain the threshold values of the water body pattern spots and the non-water body pattern spots; and in the vegetation threshold value unit, carrying out clustering analysis on the normalized vegetation index of the non-water body pattern spots by using a k-means clustering algorithm to obtain the threshold values of the vegetation pattern spots and the non-vegetation pattern spots.
The invention has the advantages and beneficial effects that:
the invention provides an object-oriented change information automatic detection method based on multi-temporal remote sensing images and GIS data, which comprises the contents of remote sensing image segmentation, pixel level change detection, feature extraction and feature statistics, automatic sample selection, classification model construction, change detection after classification and the like. The main technical process comprises three stages of remote sensing image segmentation and pixel-level change information extraction, object-oriented remote sensing image classification and object-oriented change detection after classification. The data processing of the three stages ensures the precision and accuracy of the extraction of the change information in the invention.
The invention relates to an object-oriented change information automatic detection method based on multi-temporal remote sensing images and GIS data, which is characterized in that a computer system can be used for processing the multi-temporal remote sensing images and the GIS data to automatically acquire change information. Compared with the existing method for extracting the change information by using the remote sensing image and the GIS data, the method can realize the automatic change information detection method flow applicable to a large area, can fully mine the correlation between the remote sensing image and the GIS data, and improves the accuracy of the change information extraction by using the attribute information of the GIS data, such as the category and the like.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention.
FIG. 2 is a diagram of an object-oriented classification model according to an embodiment of the present invention.
Detailed Description
The following describes a specific embodiment of the present invention in detail with reference to the accompanying drawings and a specific algorithm, and as shown in fig. 1, an object-oriented change information automatic detection method based on multi-temporal remote sensing images and GIS data includes the following steps:
1. remote sensing image segmentation and pixel level change information extraction
The method mainly comprises the steps of utilizing GIS vector data and a multi-scale segmentation algorithm to segment raster data in a T2 period to obtain image spot information, and utilizing a pixel level change detection algorithm to carry out change detection on the raster data in the T1 period and the T2 period to obtain a pixel level change detection result.
1.1. Remote sensing image segmentation
1.1.1. Grid vector data registration
And registering the grid data in the T2 period by using the GIS vector data pattern attribute information and the geometric shape information according to the GIS vector data in the T1 period and the grid data in the T2 period to obtain a grid pixel set corresponding to the vector pattern in the GIS data in the T1 period.
1.1.2. Re-segmentation based on registration data
And (3) segmenting each obtained grid pixel set by utilizing a multi-scale segmentation algorithm to obtain a T2-period image spot, namely a T2-period image segmentation result.
1.2. Pixel level change detection
And carrying out change detection on the raster data in the T1 period and the T2 period by using a k-means change detection algorithm to obtain a pixel level change detection result.
2. Object-oriented remote sensing image classification
The step is mainly to carry out change pattern spot decision on an image segmentation result in a period T2 and a pixel level change detection result to obtain a suspected change pattern spot in a period T2 and an unchanged pattern spot in a period T2. And then, carrying out feature statistics and selection on the unchanged pattern spots in the T2 period by using a feature extraction algorithm and a sample selection algorithm to obtain a proper training sample, and further obtaining a final object-oriented SVM decision tree classification model through training. And finally, obtaining a segmentation result of the suspected change pattern spots in the T2 period by using the model. The method has the main advantages that unchanged pattern spots can be screened out preliminarily through the decision of the changed pattern spots, errors in post-processing are reduced, meanwhile, feature statistics and sample selection are carried out on the unchanged pattern spots, and more accurate training samples can be obtained. In addition, the object-oriented SVM decision tree classification model provided by the embodiment of the invention fully utilizes the advantages of different characteristic values and different sample combinations to obtain a more accurate classification model.
The categories in this embodiment are divided into five categories, namely water, garden grass, forest land, artificial structures and bare land.
2.1. Change blob decision
And (4) carrying out change pattern spot decision, namely screening the image segmentation result in the T2 period to obtain suspected change pattern spots and unchanged pattern spots in the T2 period. The main method for deciding the change image spots is to perform superposition analysis on the image segmentation result and the pixel-level change detection result in the period T2, and count the number of changed pixels and unchanged pixels in each image spot. And when the ratio of the number of the changed pixels to the number of the unchanged pixels in the image spot is larger than a threshold value m, the image spot is a suspected changed image spot, otherwise, the image spot is an unchanged image spot. The value of the threshold m in this embodiment is 1.
2.2. Automatic sample selection
2.2.1. Feature statistics and selection
Feature statistics including spectral features (such as mean, variance, etc.), texture features (such as contrast, entropy, etc.), shape features (such as aspect ratio), etc. are performed on each of the obtained unchanged patches of the T2 time period. Different feature combinations are selected according to the specific object class. The present embodiment is illustrated with the mean and variance in the statistical spectral features and the normalized water body index and the normalized vegetation index.
2.2.2. Training sample selection
According to the statistical principle, for a large number of samples, the characteristic value of the same ground feature should be normally distributed. And screening the mean value and the variance of the pattern spots according to a certain confidence level to obtain pattern spots capable of representing various feature types, namely the pattern spots are training samples.
2.3. Object-oriented SVM decision tree classification model
The method mainly comprises the step of designing and training an object-oriented SVM decision tree classification model. The model mainly comprises five parts, namely a normalized water body index threshold value method, an object-oriented SVM classification model based on the water body, a normalized vegetation index threshold value method, an object-oriented SVM classification model based on the vegetation and an object-oriented SVM classification model based on the non-vegetation.
2.3.1. Threshold method based on normalized water body index
Based on a normalized water body index threshold value method, the normalized water body index characteristic value of each pattern spot is divided into water body pattern spots and non-water body pattern spots according to a certain threshold value in the T2 period. In this embodiment, a k-means clustering algorithm is used to perform clustering analysis on the normalized water body index of the water body pattern at the time T2, so as to obtain the threshold values of the water body pattern and the non-water body pattern.
2.3.2. Object-oriented SVM classification model based on water body
The classification result obtained by the threshold value method in the step 2.3.1 has a certain error, so that the object-oriented SVM classification model is used for reclassification in the step, and higher classification precision is obtained. The object-oriented SVM classification model based on the water body is used for selecting the mean value and the variance in the spectral characteristics of the water body pattern spots obtained in the step 2.3.1, and reclassifying the water body pattern spots and the non-water body pattern spots by using the SVM classification model to obtain more accurate water body pattern spots and non-water body pattern spots.
2.3.3. Based on normalized vegetation index threshold method
The normalized vegetation index threshold based method is a method for dividing the normalized vegetation index characteristic values of the non-water body pattern spots obtained in the steps 2.3.1 and 2.3.2 into vegetation pattern spots and non-vegetation pattern spots according to a certain threshold. In this embodiment, a k-means clustering algorithm is used to perform clustering analysis on the normalized vegetation index of the non-vegetation patches to obtain the threshold values of the vegetation patches and the non-vegetation patches.
2.3.4. Vegetation-based object-oriented SVM classification model
The vegetation pattern spots mainly comprise garden tillage grass pattern spots and forest land pattern spots. The vegetation-based object-oriented SVM classification model is used for selecting the mean value and the variance in the spectral characteristics of the vegetation pattern spots obtained in the step 2.3.3 and classifying the vegetation pattern spots by using the SVM classification model to obtain the tillage garden grass pattern spots and the forest land pattern spots.
2.3.5. Object-oriented SVM classification model based on non-vegetation
The non-vegetation pattern spots mainly comprise artificial structure pattern spots and bare land pattern spots. The non-vegetation-based object-oriented SVM classification model is used for selecting the mean value and the variance in the spectral characteristics of the non-vegetation patches obtained in the step 2.3.3 and classifying the non-vegetation patches by using the SVM classification model to obtain artificial structure patches and bare area patches.
2.3.6. Classification model training
Through the step 2.2, a proper training sample containing a plurality of characteristics can be obtained, and the training sample is used for training the decision tree classification model of the SVM facing the object, so that the classification model suitable for the data can be obtained.
2.4. Object-oriented classification
And classifying all the patches to be classified by using the obtained object-oriented SVM decision tree classification model to obtain the classification result of each patch.
3. Object-oriented post-classification change detection
And carrying out object-oriented classified change detection by using the types of the vector pattern spots in the GIS data in the period T1 and the classified pattern spots of the suspected change pattern spots in the period T2 to obtain the pattern spots with inconsistent types in the two periods and the type attribute information thereof, namely the final change detection result.
The embodiment of the invention also provides an object-oriented change information automatic detection system based on the multi-temporal remote sensing image and GIS data, which comprises the following modules:
the image segmentation and change information extraction module is used for extracting remote sensing image segmentation and pixel level change information and comprises the following sub-modules;
the grid vector data registration submodule is used for registering the grid data in the T2 period by utilizing the GIS vector data pattern spot type attribute information and the geometric shape information according to the GIS vector data in the T1 period and the grid data in the T2 period to obtain a grid pixel set corresponding to the vector pattern spot in the GIS data in the T1 period, namely a pattern spot in the T2 period;
the registration data re-segmentation submodule is used for performing re-segmentation on each acquired T2-period image spot by using a multi-scale segmentation algorithm to obtain sub image spots of each image spot, namely a T2-period image segmentation result;
the pixel level change detection submodule is used for carrying out change detection on the raster data in the T1 period and the T2 period by utilizing a pixel level change detection algorithm to obtain a pixel level change detection result;
the remote sensing image classification module is used for object-oriented remote sensing image classification and comprises the following sub-modules;
the change pattern spot decision submodule is used for performing superposition analysis on the image segmentation result and the pixel level change detection result in the period T2, counting the number of changed pixels and unchanged pixels in each pattern spot, and when the ratio of the number of the changed pixels to the number of the unchanged pixels in each pattern spot is greater than a threshold value m, the pattern spot is a suspected change pattern spot, otherwise, the pattern spot is an unchanged pattern spot;
the sample selection submodule is used for carrying out feature statistics on each obtained unchanged pattern spot in the T2 period, and selecting different feature combinations according to specific object types to obtain a proper training sample;
the SVM decision tree classification submodule is used for constructing an object-oriented SVM decision tree classification model, training the SVM decision tree classification model through training samples in the sample selection submodule, and classifying all suspected change pattern spots by using the trained model to obtain a classification result of each pattern spot;
and the change detection module is used for carrying out object-oriented classified change detection by utilizing the types of the vector pattern spots in the GIS data in the period T1 and the classified pattern spots of suspected change pattern spots in the period T2 to obtain the pattern spots with inconsistent types in the two periods and the type attribute information thereof, namely the final change detection result.
Wherein, the object-oriented SVM decision tree classification model constructed in the SVM decision tree classification submodule is used for more accurately dividing 5 types of ground objects such as water bodies, ploughed garden grass, forest lands, artificial structures and bare lands, and specifically comprises the following units,
the water body threshold value unit is used for dividing the pattern spots in the T2 period into water body pattern spots and non-water body pattern spots according to a certain threshold value for the normalized water body index characteristic value of each pattern spot;
the water body SVM classification unit is used for selecting the mean value and the variance in the spectral characteristics of the water body image spots obtained in the water body threshold value unit, and reclassifying by using an SVM classification model to obtain more accurate water body image spots and non-water body image spots;
the vegetation threshold unit is used for dividing the normalized vegetation index characteristic value of the non-water body pattern spots obtained in the water body threshold unit and the water body SVM classification unit into vegetation pattern spots and non-vegetation pattern spots according to a certain threshold;
the vegetation SVM classification unit is used for selecting the mean value and the variance in the spectral characteristics of the vegetation patches obtained in the vegetation threshold value unit, and classifying the vegetation patches by using an SVM classification model to obtain garden tillage patches and forest land patches;
and the non-vegetation SVM classification unit is used for selecting the mean value and the variance in the spectral characteristics of the non-vegetation patches obtained in the vegetation threshold value unit, and classifying the non-vegetation patches by using an SVM classification model to obtain artificial structure patches and bare area patches.
And the registered data is divided into sub-modules according to the pixel level change detection algorithm, wherein the pixel level change detection algorithm in the sub-modules is a k-means change detection algorithm.
The normalized water body index of the pattern spots in the T2 period is subjected to clustering analysis by using a k-means clustering algorithm in the water body threshold value unit, so that the threshold values of the water body pattern spots and the non-water body pattern spots are obtained; and in the vegetation threshold value unit, carrying out clustering analysis on the normalized vegetation index of the non-water body pattern spots by using a k-means clustering algorithm to obtain the threshold values of the vegetation pattern spots and the non-vegetation pattern spots.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.
Claims (6)
1. An object-oriented change information automatic detection method based on multi-temporal remote sensing images and GIS data is characterized by comprising the following steps:
step 1, remote sensing image segmentation and pixel level change information extraction, comprising the following substeps;
step 1.1, according to GIS vector data in the T1 period and grid data in the T2 period, utilizing GIS vector data pattern spot type attribute information and geometric shape information to register the grid data in the T2 period to obtain a grid pixel set corresponding to a vector pattern spot in the GIS data in the T1 period;
step 1.2, performing subdivision on each obtained grid pixel set by using a multi-scale division algorithm to obtain a T2-period image spot, namely a T2-period image division result;
step 1.3, carrying out change detection on the raster data in the T1 period and the T2 period by using a pixel-level change detection algorithm to obtain a pixel-level change detection result;
step 2, classifying the remote sensing images facing the object, and comprising the following substeps;
step 2.1, performing superposition analysis on the image segmentation result and the pixel level change detection result in the period T2, counting the number of changed pixels and unchanged pixels in each image spot, and when the ratio of the number of the changed pixels to the number of the unchanged pixels in each image spot is greater than a threshold value m, determining that the image spot is a suspected changed image spot, otherwise, determining that the image spot is an unchanged image spot;
step 2.2, performing feature statistics on each obtained unchanged pattern spot in the T2 period, and selecting different feature combinations according to specific object types to obtain a proper training sample;
step 2.3, constructing an object-oriented SVM decision tree classification model, training the SVM decision tree classification model through the training samples in the step 2.2, and classifying all suspected change patches by using the trained model to obtain the classification result of each patch;
the object-oriented SVM decision tree classification model constructed in the step 2.3 is used for more accurately dividing 5 types of ground objects such as water bodies, garden ploughs, grasses, forest lands, artificial structures and bare lands, and the specific implementation mode is as follows,
step 2.3.1, dividing the pattern spots in the period T2 into water body pattern spots and non-water body pattern spots according to a certain threshold value for the normalized water body index characteristic value of each pattern spot;
step 2.3.2, selecting the mean value and the variance in the spectral characteristics of the water body pattern spots obtained in the step 2.3.1, and performing reclassification by using an SVM classification model to obtain more accurate water body pattern spots and non-water body pattern spots;
step 2.3.3, dividing the normalized vegetation index characteristic value of the non-water body pattern spots obtained in the step 2.3.1 and the step 2.3.2 into vegetation pattern spots and non-vegetation pattern spots according to a certain threshold value;
step 2.3.4, selecting the mean value and the variance in the spectral characteristics of the vegetation pattern spots obtained in the step 2.3.3, and classifying by using an SVM classification model to obtain garden cultivation and grass pattern spots and forest land pattern spots;
step 2.3.5, selecting the mean value and the variance in the spectral characteristics of the non-vegetation patches obtained in the step 2.3.3, and classifying the non-vegetation patches by using an SVM classification model to obtain artificial structure patches and bare area patches;
and 3, carrying out object-oriented classified change detection by using the types of the vector pattern spots in the GIS data in the period T1 and the classified pattern spots of the suspected change pattern spots in the period T2 to obtain the pattern spots with inconsistent types in the two periods and the type attribute information thereof, namely the final change detection result.
2. The automatic detection method for object-oriented change information based on multi-temporal remote sensing images and GIS data according to claim 1, characterized in that: the pixel level change detection algorithm in step 1.3 is a k-means change detection algorithm.
3. The automatic detection method for object-oriented change information based on multi-temporal remote sensing images and GIS data according to claim 1, characterized in that: performing cluster analysis on the normalized water body index of the pattern spots in the period T2 by using a k-means clustering algorithm in the step 2.3.1 to obtain threshold values of the water body pattern spots and the non-water body pattern spots; and 2.3.3, carrying out clustering analysis on the normalized vegetation index of the non-water body pattern spots by using a k-means clustering algorithm to obtain the vegetation pattern spots and the threshold values of the non-vegetation pattern spots.
4. An object-oriented change information automatic detection system based on multi-temporal remote sensing images and GIS data is characterized by comprising the following modules:
the image segmentation and change information extraction module is used for extracting remote sensing image segmentation and pixel level change information and comprises the following sub-modules;
the grid vector data registration submodule is used for registering the grid data in the T2 period by utilizing the GIS vector data pattern spot type attribute information and the geometric shape information according to the GIS vector data in the T1 period and the grid data in the T2 period to obtain a grid pixel set corresponding to the vector pattern spot in the GIS data in the T1 period, namely a pattern spot in the T2 period;
the registration data re-segmentation submodule is used for performing re-segmentation on each acquired T2-period image spot by using a multi-scale segmentation algorithm to obtain sub image spots of each image spot, namely a T2-period image segmentation result;
the pixel level change detection submodule is used for carrying out change detection on the raster data in the T1 period and the T2 period by utilizing a pixel level change detection algorithm to obtain a pixel level change detection result;
the remote sensing image classification module is used for object-oriented remote sensing image classification and comprises the following sub-modules;
the change pattern spot decision submodule is used for performing superposition analysis on the image segmentation result and the pixel level change detection result in the period T2, counting the number of changed pixels and unchanged pixels in each pattern spot, and when the ratio of the number of the changed pixels to the number of the unchanged pixels in each pattern spot is greater than a threshold value m, the pattern spot is a suspected change pattern spot, otherwise, the pattern spot is an unchanged pattern spot;
the sample selection submodule is used for carrying out feature statistics on each obtained unchanged pattern spot in the T2 period, and selecting different feature combinations according to specific object types to obtain a proper training sample;
the SVM decision tree classification submodule is used for constructing an object-oriented SVM decision tree classification model, training the SVM decision tree classification model through training samples in the sample selection submodule, and classifying all suspected change pattern spots by using the trained model to obtain a classification result of each pattern spot;
the object-oriented SVM decision tree classification model constructed in the SVM decision tree classification submodule is used for more accurately dividing 5 types of ground objects such as water bodies, cultivated grasses, forest lands, artificial structures and bare lands, and specifically comprises the following units,
the water body threshold value unit is used for dividing the pattern spots in the T2 period into water body pattern spots and non-water body pattern spots according to a certain threshold value for the normalized water body index characteristic value of each pattern spot;
the water body SVM classification unit is used for selecting the mean value and the variance in the spectral characteristics of the water body image spots obtained in the water body threshold value unit, and reclassifying by using an SVM classification model to obtain more accurate water body image spots and non-water body image spots;
the vegetation threshold unit is used for dividing the normalized vegetation index characteristic value of the non-water body pattern spots obtained in the water body threshold unit and the water body SVM classification unit into vegetation pattern spots and non-vegetation pattern spots according to a certain threshold;
the vegetation SVM classification unit is used for selecting the mean value and the variance in the spectral characteristics of the vegetation patches obtained in the vegetation threshold value unit, and classifying the vegetation patches by using an SVM classification model to obtain garden tillage patches and forest land patches;
the non-vegetation SVM classification unit is used for selecting the mean value and the variance in the spectral characteristics of the non-vegetation patches obtained in the vegetation threshold value unit, and classifying the non-vegetation patches by using an SVM classification model to obtain artificial structure patches and bare area patches;
and the change detection module is used for carrying out object-oriented classified change detection by utilizing the types of the vector pattern spots in the GIS data in the period T1 and the classified pattern spots of suspected change pattern spots in the period T2 to obtain the pattern spots with inconsistent types in the two periods and the type attribute information thereof, namely the final change detection result.
5. The automatic detection system for object-oriented change information based on multi-temporal remote sensing images and GIS data according to claim 4, characterized in that: and the registered data is divided into sub-modules according to the pixel level change detection algorithm which is a k-means change detection algorithm.
6. The automatic detection system for object-oriented change information based on multi-temporal remote sensing images and GIS data according to claim 4, characterized in that: carrying out clustering analysis on the normalized water body index of the pattern spots in the T2 period by using a k-means clustering algorithm in the water body threshold value unit to obtain threshold values of the water body pattern spots and the non-water body pattern spots; and in the vegetation threshold value unit, carrying out clustering analysis on the normalized vegetation index of the non-water body pattern spots by using a k-means clustering algorithm to obtain the threshold values of the vegetation pattern spots and the non-vegetation pattern spots.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810538861.3A CN108846832B (en) | 2018-05-30 | 2018-05-30 | Multi-temporal remote sensing image and GIS data based change detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810538861.3A CN108846832B (en) | 2018-05-30 | 2018-05-30 | Multi-temporal remote sensing image and GIS data based change detection method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108846832A CN108846832A (en) | 2018-11-20 |
| CN108846832B true CN108846832B (en) | 2021-06-15 |
Family
ID=64210020
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810538861.3A Active CN108846832B (en) | 2018-05-30 | 2018-05-30 | Multi-temporal remote sensing image and GIS data based change detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108846832B (en) |
Families Citing this family (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109655583A (en) * | 2018-12-14 | 2019-04-19 | 天津珞雍空间信息研究院有限公司 | A kind of atmospheric environment ground monitoring website based on satellite remote sensing is deployed to ensure effective monitoring and control of illegal activities network-building method |
| CN109657610A (en) * | 2018-12-18 | 2019-04-19 | 北京航天泰坦科技股份有限公司 | A kind of land use change survey detection method of high-resolution multi-source Remote Sensing Images |
| CN109829425B (en) * | 2019-01-31 | 2020-12-22 | 沈阳农业大学 | Farmland landscape small-scale ground feature classification method and system |
| CN109816749B (en) * | 2019-02-14 | 2023-04-28 | 中国测绘科学研究院 | Map symbol filling method and device, computer equipment and storage medium |
| CN109934154B (en) * | 2019-03-08 | 2021-06-01 | 北京科技大学 | Remote sensing image change detection method and detection device |
| CN109901240A (en) * | 2019-04-03 | 2019-06-18 | 中国矿业大学 | A method of detecting residual coal spontaneous combustion region of appearing |
| CN110263627A (en) * | 2019-05-09 | 2019-09-20 | 生态环境部卫星环境应用中心 | A kind of mankind's activity variation clue monitoring method towards protecting field supervision |
| CN110348483B (en) * | 2019-06-06 | 2021-09-21 | 重庆工商大学融智学院 | Spatial information change detection and classification method based on spatial information and statistical learning |
| CN110473205A (en) * | 2019-07-10 | 2019-11-19 | 北京吉威数源信息技术有限公司 | Remote sensing image information extracting method and system based on arrow bar phantom |
| CN110738679A (en) * | 2019-10-21 | 2020-01-31 | 山东省国土测绘院 | Geographical provincial monitoring method and system based on automatic extraction of remote sensing image change |
| CN110826454B (en) * | 2019-10-30 | 2022-06-28 | 北京科技大学 | Remote sensing image change detection method and device |
| CN111739067A (en) * | 2020-05-28 | 2020-10-02 | 中国国土勘测规划院 | Remote sensing image change detection method and device |
| CN111897907B (en) * | 2020-08-12 | 2024-03-01 | 国家基础地理信息中心 | Method, device and storage medium for filtering pseudo-change information of earth surface coverage data |
| CN112101159B (en) * | 2020-09-04 | 2022-05-24 | 国家林业和草原局中南调查规划设计院 | Multi-temporal forest remote sensing image change monitoring method |
| CN112215135B (en) * | 2020-10-10 | 2022-07-08 | 东华理工大学 | Mining area mining and treatment effect monitoring method and device |
| CN112580510B (en) * | 2020-12-18 | 2024-05-21 | 广州市城市规划设计有限公司 | Method, device, equipment and storage medium for estimating permeable ground area rate |
| CN112801170A (en) * | 2021-01-25 | 2021-05-14 | 河南省国土资源调查规划院(河南省国土资源厅不动产登记中心) | Multi-temporal change detection and ground feature identification and classification method based on remote sensing image |
| CN113298755B (en) * | 2021-04-13 | 2021-11-26 | 生态环境部卫星环境应用中心 | Method and device for rapidly detecting ecological environment change patch based on time sequence image |
| CN113484245B (en) * | 2021-07-05 | 2022-11-22 | 重庆市规划和自然资源调查监测院 | Remote sensing rapid monitoring method and system for paddy field planting pattern in hilly and mountainous areas and computer readable storage medium |
| CN113822900B (en) * | 2021-07-09 | 2023-09-22 | 武汉大学 | Method and system for automatically selecting new image sample based on vector constraint object-oriented |
| CN113989657B (en) * | 2021-10-11 | 2024-05-10 | 中国测绘科学研究院 | Cultivated land range change detection method and device based on invariable information sample screening |
| CN114119575B (en) * | 2021-11-30 | 2022-07-19 | 二十一世纪空间技术应用股份有限公司 | Spatial information change detection method and system |
| CN113989652B (en) * | 2021-12-27 | 2022-04-26 | 中国测绘科学研究院 | Method and system for detecting farmland change under layered multiple judgment rules |
| CN114882084B (en) * | 2022-05-07 | 2024-04-05 | 安徽农业大学 | Land utilization change pattern automatic identification method based on artificial intelligence |
| CN117612020B (en) * | 2024-01-24 | 2024-07-05 | 西安宇速防务集团有限公司 | SGAN-based detection method for resisting change of remote sensing image element of neural network |
| CN118230186B (en) * | 2024-04-09 | 2024-07-26 | 长江水利委员会长江科学院 | Visible light remote sensing image water body change detection method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20050041277A (en) * | 2003-10-30 | 2005-05-04 | 이준환 | Automatic analysing method for high precision satellite image |
| CN102496034A (en) * | 2011-11-29 | 2012-06-13 | 南京师范大学 | High-spatial resolution remote-sensing image bag-of-word classification method based on linear words |
| CN102622345A (en) * | 2011-01-26 | 2012-08-01 | 杨存建 | High-precision land-utilization remote sensing updating technology with synergistic multisource spatio-temporal data |
| CN103400151A (en) * | 2013-08-16 | 2013-11-20 | 武汉大学 | Optical remote-sensing image, GIS automatic registration and water body extraction integrated method |
| CN103729653A (en) * | 2014-01-21 | 2014-04-16 | 武汉大学 | Supervision change detection method of high-resolution remote sensing images |
| CN103971115A (en) * | 2014-05-09 | 2014-08-06 | 中国科学院遥感与数字地球研究所 | Automatic extraction method for newly-increased construction land image spots in high-resolution remote sensing images based on NDVI and PanTex index |
| CN106803097A (en) * | 2016-12-28 | 2017-06-06 | 中国科学院深圳先进技术研究院 | A kind of utilization remote sensing image extracts method and device, the user equipment of vegetation sunshade net classification |
-
2018
- 2018-05-30 CN CN201810538861.3A patent/CN108846832B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20050041277A (en) * | 2003-10-30 | 2005-05-04 | 이준환 | Automatic analysing method for high precision satellite image |
| CN102622345A (en) * | 2011-01-26 | 2012-08-01 | 杨存建 | High-precision land-utilization remote sensing updating technology with synergistic multisource spatio-temporal data |
| CN102496034A (en) * | 2011-11-29 | 2012-06-13 | 南京师范大学 | High-spatial resolution remote-sensing image bag-of-word classification method based on linear words |
| CN103400151A (en) * | 2013-08-16 | 2013-11-20 | 武汉大学 | Optical remote-sensing image, GIS automatic registration and water body extraction integrated method |
| CN103729653A (en) * | 2014-01-21 | 2014-04-16 | 武汉大学 | Supervision change detection method of high-resolution remote sensing images |
| CN103971115A (en) * | 2014-05-09 | 2014-08-06 | 中国科学院遥感与数字地球研究所 | Automatic extraction method for newly-increased construction land image spots in high-resolution remote sensing images based on NDVI and PanTex index |
| CN106803097A (en) * | 2016-12-28 | 2017-06-06 | 中国科学院深圳先进技术研究院 | A kind of utilization remote sensing image extracts method and device, the user equipment of vegetation sunshade net classification |
Non-Patent Citations (5)
| Title |
|---|
| Change Detection Based on Gabor Wavelet Features for Very High Resolution Remote Sensing Images;Zhenxuan Li et al;《IEEE GEOSCIENCE AND REMOTE SENSING LETTERS》;20170531;第14卷(第5期);第783-787页 * |
| LAND COVER CHANGE AND EFFECT ON AGRICULTURAL IRRIGATION BASED ON RS AND GIS FOR 20 YEARS IN YANQING COUNTY, BEIJING;Linrui Song et al;《2011 IEEE international Geoscience and Remote Sensing Symposium 》;20111020;第700-703页 * |
| 基于特征的时空数据模型研究及在土地利用变化动态监测中的应用;崔伟宏 等;《测绘学报》;20040531;第33卷(第2期);第138-145页 * |
| 遥感影像与GIS分析相结合的变化检测方法;张晓东 等;《武汉大学学报 信息科学版》;20060331;第31卷(第3期);第266-269页 * |
| 高分辨率遥感影像与矢量数据结合的变化检测方法研究;赵珍珍;《中国优秀硕士学位论文全文数据库 基础科学辑》;20150115;第2015年卷(第01期);第A008-51页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108846832A (en) | 2018-11-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108846832B (en) | Multi-temporal remote sensing image and GIS data based change detection method and system | |
| CN112164062A (en) | Wasteland information extraction method and device based on remote sensing time sequence analysis | |
| Wang et al. | Optimal segmentation of high-resolution remote sensing image by combining superpixels with the minimum spanning tree | |
| CN106778605B (en) | Automatic remote sensing image road network extraction method under assistance of navigation data | |
| Zhang et al. | Automated delineation of agricultural field boundaries from Sentinel-2 images using recurrent residual U-Net | |
| Vidović et al. | Crop row detection by global energy minimization | |
| Zhou et al. | On detecting road regions in a single UAV image | |
| CN108830870A (en) | Satellite image high-precision field boundary extracting method based on Multi-scale model study | |
| Zhou et al. | Individual tree parameters estimation for plantation forests based on UAV oblique photography | |
| CN109389062B (en) | Method for extracting lake water and land dividing line by using high-resolution satellite-borne SAR image | |
| Jo et al. | Deep learning applications on multitemporal SAR (Sentinel-1) image classification using confined labeled data: The case of detecting rice paddy in South Korea | |
| CN111028255A (en) | Farmland area pre-screening method and device based on prior information and deep learning | |
| CN103034863A (en) | Remote-sensing image road acquisition method combined with kernel Fisher and multi-scale extraction | |
| Liang et al. | Maximum likelihood classification of soil remote sensing image based on deep learning | |
| CN107301649B (en) | Regional merged SAR image coastline detection algorithm based on superpixels | |
| CN112906627B (en) | Green pricklyash peel identification method based on semantic segmentation | |
| Xu et al. | The comparative study of three methods of remote sensing image change detection | |
| CN109543729A (en) | Time series data land cover classification method based on characteristic parameter cluster | |
| Lobo et al. | Fine-scale mapping of a grassland from digitized aerial photography: An approach using image segmentation and discriminant analysis | |
| Zhao et al. | Improving object-oriented land use/cover classification from high resolution imagery by spectral similarity-based post-classification | |
| CN109344837B (en) | SAR image semantic segmentation method based on deep convolutional network and weak supervised learning | |
| CN116310864A (en) | Automatic identification method, system, electronic equipment and medium for crop lodging | |
| Shumilo | Automatic anomaly detection methodology for crop classification data using morphological features | |
| Turker et al. | Agricultural field detection from satellite imagery using the combined otsu’s thresholding algorithm and marker-controlled watershed-based transform | |
| CN111768101B (en) | Remote sensing cultivated land change detection method and system taking account of physical characteristics |
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 |