CN113822900B - Method and system for automatically selecting new image sample based on vector constraint object-oriented - Google Patents
Method and system for automatically selecting new image sample based on vector constraint object-oriented Download PDFInfo
- Publication number
- CN113822900B CN113822900B CN202110776117.9A CN202110776117A CN113822900B CN 113822900 B CN113822900 B CN 113822900B CN 202110776117 A CN202110776117 A CN 202110776117A CN 113822900 B CN113822900 B CN 113822900B
- Authority
- CN
- China
- Prior art keywords
- vector
- image
- oriented
- segmentation
- new image
- 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
- 239000013598 vector Substances 0.000 title claims abstract description 120
- 238000000034 method Methods 0.000 title claims abstract description 32
- 230000011218 segmentation Effects 0.000 claims abstract description 39
- 238000003709 image segmentation Methods 0.000 claims abstract description 22
- 238000001228 spectrum Methods 0.000 claims abstract description 16
- 238000001514 detection method Methods 0.000 claims abstract description 12
- 238000004891 communication Methods 0.000 claims description 13
- 238000010586 diagram Methods 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 238000010187 selection method Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 230000015572 biosynthetic process Effects 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 4
- 239000003550 marker Substances 0.000 claims description 3
- 238000003786 synthesis reaction Methods 0.000 claims 1
- 238000000605 extraction Methods 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 238000009825 accumulation Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 241001270131 Agaricus moelleri Species 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
- G06T7/337—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods involving reference images or patches
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method and a system for automatically selecting a new image sample based on vector constraint object-oriented, wherein the method comprises the steps of geometrically registering a vector map and a new image by adopting a vector grid registration mode to obtain a common area of the vector map and the new image; based on the grid boundary of the vector map after rasterization, under the constraint of the grid boundary, carrying out fine segmentation on the image by adopting an object-oriented image segmentation mode, so that the segmentation boundary is consistent with the vector polygon while obtaining the fine segmentation of the interior of the polygon; extracting spectrum and texture features of the segmented image object; corresponding polygons of a ground class in the vector map to obtain image segmentation image spots, and counting the feature distribution corresponding to the image spots; and obtaining the ground object sample object according to the probability distribution and the feature consistency criterion, and using the ground object sample object for classification and change detection of the remote sensing image. The invention can fully utilize the existing basic geographic data and realize the automatic acquisition of the ground object samples of different time phase images based on the vector-guided object-oriented image classification.
Description
Technical Field
The method belongs to the technical field of remote sensing image processing and target recognition, and particularly relates to a novel method and system for automatically selecting different image ground samples.
Background
With the continuous emission of high-resolution remote sensing satellites and the increasing demands of users on investigation of fine homeland resources, the technical demands for classification and change detection of high-spatial-resolution (high-resolution) remote sensing images are also urgent. The high-resolution remote sensing image has rich textures and contextual characteristics, and due to the existence of a large number of foreign matter homography and homography, the traditional pixel-based method is difficult to utilize the geometric and texture information of the ground features in the high-resolution remote sensing image, and a large amount of noise exists in the extraction result, so that the method is difficult to meet the current requirements.
The object-oriented image processing method uses the object generated after the image segmentation as a unit, and uses the characteristics of the object such as spectrum, geometry, texture and the like to extract information such as image classification, change detection and the like, so that the problems of pixel-based processing can be effectively solved. In object-oriented image classification and change detection, image segmentation and feature extraction are key to such methods.
In recent years, the salient expression of the deep learning in terms of image target detection and image semantic segmentation provides a new thought for high-resolution remote sensing image classification and change detection, but because the deep learning requires a large amount of training samples, particularly for the complexity and uncertainty of natural features and images in remote sensing images, the sample construction is more difficult, the problems of incomplete detection, missed detection and the like still exist on extraction results, the research is still in an exploration stage, the object-oriented image segmentation can obtain complete and salient homogeneous feature image spots, is an important mode for obtaining complete features, and is one of important problems that needs to be solved in the current engineering application. Meanwhile, in the image acquisition process, the image difference is large due to the time phase difference, the sensor difference and the imaging environment difference, the selection of samples from different images is one of the main problems of great workload, and how to solve the problem is also an important problem to be solved by the invention.
Currently, along with the accumulation of geographical national census data, the accumulation of remote sensing images, the appearance of remote sensing big data, the object-oriented image classification and the requirement of deep learning of a large number of samples, the effective utilization of the data for acquiring different data source samples has important significance.
In order to solve the problems, the invention designs a sample set acquisition method by combining the existing vector map resources and the continuously accumulated multi-source remote sensing image resources by considering the fact that the difference exists between the acquisition time of the vector map and the acquisition time of the image and the image update speed and the source are far higher than those of the vector map and the situation that the ground feature is possibly changed.
Disclosure of Invention
Aiming at the problem that the workload of acquiring different time-phase image ground object samples of object-oriented image classification is large, the invention provides a novel vector constraint-based object-oriented image segmentation and sample automatic selection scheme, so that sample selection of different time phases and different data source images can be met, and the results can be directly applied to applications such as object-oriented image classification, change detection, deep learning sample library construction and the like.
In order to achieve the above object, the technical scheme provided by the invention is an automatic selection method of a new image sample based on vector constraint object-oriented, comprising the following steps,
step 1, performing geometric registration on a vector map and a new image by adopting a vector grid registration mode to obtain a common area of the vector map and the new image;
step 2, based on the grid boundary of the vector map after rasterization, carrying out fine segmentation on the image by adopting an object-oriented image segmentation mode under the constraint of the grid boundary, so that the segmentation boundary is consistent with the vector polygon while obtaining the fine segmentation of the interior of the polygon;
step 3, extracting spectrum and texture features of the separated image object;
step 4, corresponding polygons of a ground class in the vector map to obtain image segmentation image spots, and counting feature distribution corresponding to the image spots;
and step 5, obtaining the sample object of the ground object according to the probability distribution sum and the characteristic consistency criterion, and using the sample object for classifying and detecting the change of the remote sensing image.
In step 2, the obtained results are differentiated on the basis of the rasterization processing and the image object-oriented segmentation of the vector map, and the connected domain calculation is performed on the differential map, so that the vector constraint object-oriented subdivision result is obtained.
In step 3, multiple spectrum texture feature calculations are performed by taking the object as a unit, and the category to which each object belongs is recorded according to the vector rasterized marker diagram corresponding to the object.
In step 4, feature distribution is analyzed by using feature histogram distribution characteristics.
In step 5, whether the object belongs to a certain type of ground feature is judged according to the feature characteristic obeying normal distribution characteristics to select a sample, and the sample is determined through the combination of multiple feature judgment results.
The invention provides a vector constraint object-oriented new image sample automatic selection system, which is used for realizing the vector constraint object-oriented new image sample automatic selection method.
Furthermore, the device comprises the following modules,
the first module is used for geometrically registering the vector map and the new image by adopting a vector grid registration mode to obtain a common area of the vector map and the new image;
the second module is used for finely dividing the image by adopting an object-oriented image dividing mode under the constraint of the grid boundary after the grid formation of the vector map, so that the dividing boundary is consistent with the vector polygon while the fine division in the polygon is obtained;
the third module is used for extracting spectrum and texture characteristics of the image object obtained by segmentation;
a fourth module, configured to obtain image segmentation patches corresponding to polygons of a ground class in the vector map, and count feature distribution corresponding to the patches;
and the fifth module is used for obtaining the object sample object of the ground object according to the probability distribution sum and the characteristic consistency criterion and is used for classifying and detecting the change of the remote sensing image.
Further, the system comprises a processor and a memory, wherein the memory is used for storing program instructions, and the processor is used for calling the stored instructions in the memory to execute a new image sample automatic selection method based on vector constraint and object oriented.
Further, a readable storage medium having stored thereon a computer program which, when executed, implements a method for automatically selecting new image samples based on vector constraint object-oriented as described above.
Compared with the prior art, the invention has the beneficial effects that:
according to the invention, the existing geographical national condition census ground surface covered land utilization data are utilized, and the ground object samples are automatically selected from the new time phase images through the steps, so that the method is suitable for collecting samples of different data sources, and the workload of manually selecting samples from different images is effectively reduced.
The scheme of the invention is simple and convenient to implement, has strong practicability, solves the problems of low practicability and inconvenient practical application existing in the related technology, can improve user experience, and has important market value.
Drawings
FIG. 1 is a flow chart of a method according to an embodiment of the invention.
FIG. 2 is a diagram of a vector-to-grid result in accordance with an embodiment of the present invention.
FIG. 3 is a graph showing the initial segmentation sub-segmentation results according to an embodiment of the present invention.
Fig. 4 is a schematic diagram of a differential result according to an embodiment of the invention.
Fig. 5 is a detailed view showing the result of vector-guided segmentation according to the present invention.
FIG. 6 is a graph showing a distribution of spectral characteristics of a portion of a field object according to an embodiment of the present invention.
FIG. 7 is a graph showing a partial texture feature distribution of a farm object according to an embodiment of the present invention.
Fig. 8 is a graph showing a spectral characteristic distribution of a partial wave band of a residential object according to an embodiment of the present invention.
Fig. 9 is a partial texture feature distribution diagram of a residential object according to an embodiment of the present invention.
Detailed Description
The technical scheme of the invention is specifically described below with reference to the accompanying drawings and examples.
In order to fully utilize the existing basic geographic data, timely update and extract surface coverage and change information from different time phase images, solve the problem that the requirements of the common information extraction and change detection on the time phase imaging seasons of front and back images are greatly influenced by seasons, and if a post-classification change detection method is adopted, samples are required to be respectively selected for training and classification in the time phase images, and the sample selection workload is large, the invention discloses a sample automatic selection method based on the existing geographic national condition census data.
The embodiment of the invention provides a method for automatically selecting a sample set from different images based on thematic vector constraint, which comprises five steps of vector grid registration, object-oriented image segmentation based on vector constraint, image spectrum and texture object feature extraction, image object spectrum texture feature statistics and sample selection, wherein the implementation flow is shown in figure 1, and the method comprises the following specific steps:
step 1: the geometric registration of the vector map and the grid image comprises the geometric registration of the vector map and the new image by adopting a vector grid registration mode to obtain a common area of the vector map and the new image.
By creating matching point pairs, i.e., control points (feature points), on the vector and grid images, the control points should be uniformly distributed throughout the grid range. Since the accuracy of the correspondence of the feature points to the positions is a key to the accuracy of image registration, the present invention preferably suggests selecting, as the control points, elements that are fixed and do not change with time, such as road intersections or corners of buildings. And establishing a space transformation relation model of the vector map and the grid image according to the feature point pairs, and laying a foundation for image segmentation based on vector constraint by conforming the coordinate space positions of the vector map and the grid image according to the transformation relation model. The registered result can be displayed in a mode of displaying a vector boundary and a superimposed image, so that the effect after the spatial position corresponds is reflected, and the registered result can also be displayed in a mode of displaying a local thematic superimposed image and is used for reflecting image spots corresponding to various categories. And finding out a common area through the upper left corner and the lower right corner of the space geometric coordinates of the registered vector and the image, and obtaining the working range of the following steps and algorithms.
Step 2: the object-oriented image segmentation based on vector constraint can be based on the surface coverage/land utilization polygons of the vector map, and under the constraint of the polygon, the image is finely segmented by adopting an object-oriented image segmentation method, so that the segmentation boundary is consistent with the vector polygon while the internal fine segmentation of the polygons is obtained.
The method comprises the steps of carrying out rasterization processing and image object-oriented segmentation on a vector map, carrying out difference on the vector map and the image object-oriented segmentation, and carrying out connected domain calculation on the difference map to obtain an object-oriented subdivision result of vector constraint. In an embodiment, step 2 is performed by the following steps:
(1) Vector rasterization. The registered vector map data is rasterized by using a rasterization algorithm by using the existing vector rotation rasterization tool, such as a features to raster tool in arcgis software and a gdal_rasterize program in GDAL, and according to the pattern spot numbers. Fig. 2 shows a raster image obtained by vector rasterization and a corresponding vector boundary thereof, white lines are original vector boundaries, gray blocks with different gray scales are image spots obtained by vector rasterization, and it can be seen that, because vectors are connecting lines between points, the raster boundary after vector rasterization is a pixel boundary and has a certain position difference with the vector boundary, and we just need to control the segmentation boundary by using the raster boundary.
(2) And (5) image segmentation. The invention preferably suggests that the watershed segmentation algorithm is adopted for image segmentation, and other object-oriented image segmentation modes can be adopted when the method is implemented. In order to obtain as uniform a pattern as possible. Proper segmentation scale parameters are required to be selected according to the image, the initial segmentation result is affected by different scale parameter settings, and the larger the segmentation scale parameters are, the larger the obtained image spots are, the under segmentation possibly occurs, and the inconsistency occurs in the image spots. Therefore, in actual operation, small initial segmentation scale parameters can be considered to be selected, and the image is finely segmented or over-segmented, so that the image spots obtained by segmentation are not under-segmented as much as possible. The segmentation results are shown in fig. 3.
(3) And (3) differentiating the marked image obtained by the vector rasterization obtained in the step (1) and the automatically segmented image obtained in the step (2) to obtain more areas. The difference result is schematically shown in FIG. 4, wherein 0/1/2/3 represents the difference value. As can be seen from the data analysis of the image obtained by difference, all small areas are marked after difference, for example, the area where the automatic segmentation result falls inside the vector polygon (namely, the vector polygon contains a plurality of objects, and the difference marking value of each object is different), the area where the automatic segmentation result spans the vector boundary (the areas on two sides of the vector boundary can obtain different marking values because the vector IDs are different and the automatic segmentation object IDs are the same), meanwhile, as the difference value between the vector polygon ID and the object ID possibly exists repeatedly, the repetition does not affect the communication area, but the situation that the communication phenomenon appears due to the difference value of just two different objects is not excluded, the situation that the communication phenomenon appears rarely can be eliminated in an algorithm, the communication area judgment is carried out according to the meaning and the requirement of the vector guiding segmentation, at this time, the communication area judgment is only needed according to the numerical value in the difference value diagram, and the communication area is combined into the object which belongs to the same polygon ID with the largest adjacent and similar object size according to the difference value. It has been found that the vector grid pattern can be taken into account when generating the connected region of the difference pattern, so that the generated pattern spots are automatically inside the vector grid pattern and do not cross the boundary.
(4) The difference plot is marked with a plot patch. The link is completed by the following four steps: 1) The progressive image is scanned, successive pixels in each row having the same difference value form a sequence into a cluster, and the start point, the end point, the row number and the difference value are recorded. Thus, the label of each bolus is unique. Groups having the same differential value are classified into one type (in this case, the same differential value is not considered but the groups belong to different features). 2) Construction of a connected region. Judging each group in the similar groups according to the row numbers, the starting point and the ending point of the groups: if the clusters in the previous row are overlapped except the first row, the same communication area can be judged, otherwise, the non-communication area is judged. 3) And traversing each similar group to finish searching the connected region. The initial results may be viewed according to the above procedure, and the small regions may be categorized according to their similarity to the surrounding large regions or exist as separate individuals. At this time, the image value and the differential value obtained by vector rasterization can be combined to solve the communication problem caused by the same differential value of different classes. The search is faster because of the recorded line number of each clique and the starting point coordinates. 4) The panels are merged. When processing small patches, if the area is smaller than the threshold, it is merged into the nearest similar difference patch according to its own vector ID and its original segmentation ID.
The vector-guided segmentation result obtained by the above steps is shown in fig. 5.
Step 3: and (3) calculating the spectrum and texture properties of the object obtained by segmentation based on the result obtained in the step (2). And calculating the spectrum mean value of each wave band of the object and the texture characteristics of the gray level co-occurrence matrix of the object by taking the object as a unit, and recording the category to which each object belongs according to the vector rasterized marker diagram corresponding to the object. Taking farmlands and residents as examples, the calculated partial spectra and texture features are shown in tables 1 and 2.
TABLE 1 results of farmland characteristic calculations
TABLE 2 calculation results of residential properties
Step 4: taking the ground object type in the vector diagram polygon as a research object, carrying out statistical analysis on the characteristics obtained in the step 3 according to the type, and finding out that most of the ground object type characteristics are normally distributed according to the statistical result of the histogram distribution. The mean and standard deviation of each spectrum, texture feature of the class object are calculated. Fig. 6-9 illustrate the statistical distribution of the spectrum and texture features of the farmland and the residential area, respectively.
Step 5: sample selection. In order to ensure the accuracy of sample selection results, selecting the object features calculated in the step 3 according to the types, taking the mean value and standard deviation of each feature calculated in the step 4 as selection basis, and taking the feature number of which the object feature value is within 2 times of the standard deviation of the feature of the type as the basis for judging samples to a certain amount. The specific algorithm implementation details are as follows: setting a sample marking variable, initializing to 0, calculating the difference between each feature of each object and the average value of the class features by taking the object as a unit, adding 1 to the sample marking variable when the absolute value difference is smaller than 1.5-2 times of standard deviation, otherwise, keeping unchanged, and circulating all the features. And after the discrimination of all the object feature marks is finished, marking the object with the sample mark variable being greater than or equal to a certain threshold value as a sample object, so as to obtain a sample selection result, for example, extracting farmland samples and residential samples from the original images.
Thus, the acquisition of the sample object set is completed.
In particular, the method according to the technical solution of the present invention may be implemented by those skilled in the art using computer software technology to implement an automatic operation flow, and a system apparatus for implementing the method, such as a computer readable storage medium storing a corresponding computer program according to the technical solution of the present invention, and a computer device including the operation of the corresponding computer program, should also fall within the protection scope of the present invention.
In some possible embodiments, an automatic selection system for new image samples based on vector constraint object-oriented is provided, comprising the following modules,
the first module is used for geometrically registering the vector map and the new image by adopting a vector grid registration mode to obtain a common area of the vector map and the new image;
the second module is used for finely dividing the image by adopting an object-oriented image dividing mode under the constraint of the grid boundary after the grid formation of the vector map, so that the dividing boundary is consistent with the vector polygon while the fine division in the polygon is obtained;
the third module is used for extracting spectrum and texture characteristics of the image object obtained by segmentation;
a fourth module, configured to obtain image segmentation patches corresponding to polygons of a ground class in the vector map, and count feature distribution corresponding to the patches;
and the fifth module is used for obtaining the sample object of the ground object according to the probability distribution and the characteristic consistency criterion and is used for classifying and detecting the change of the remote sensing image.
In some possible embodiments, a system for automatically selecting new image samples based on vector constraint object-oriented is provided, which includes a processor and a memory, the memory is used for storing program instructions, and the processor is used for calling the storage instructions in the memory to execute a new image sample automatic selection method based on vector constraint object-oriented as described above.
In some possible embodiments, a system for automatically selecting new image samples based on vector constraint object-oriented is provided, comprising a readable storage medium having a computer program stored thereon.
The specific embodiments described herein are offered by way of example only to illustrate the spirit of the invention. Those skilled in the art may make various modifications or additions to the described embodiments or substitutions thereof without departing from the spirit of the invention or exceeding the scope of the invention as defined in the accompanying claims.
Claims (9)
1. An automatic selection method of a new image sample based on vector constraint object-oriented is characterized in that: comprises the steps of,
step 1, performing geometric registration on a vector map and a new image by adopting a vector grid registration mode to obtain a common area of the vector map and the new image;
step 2, based on the grid boundary of the vector map after rasterization, carrying out fine segmentation on the image by adopting an object-oriented image segmentation mode under the constraint of the grid boundary, so that the segmentation boundary is consistent with the vector polygon while obtaining the fine segmentation of the interior of the polygon; the implementation comprises the steps of,
(1) Vector rasterizing, which comprises using a vector rotation rasterizing tool to raster the registered vector map data according to the pattern spot number by adopting a rasterizing algorithm to obtain a raster pattern obtained by vector rasterizing and a corresponding raster boundary thereof;
(2) Image segmentation, including fine segmentation of images using an object-oriented image segmentation algorithm;
(3) Differentiating the marked image obtained by the vector rasterization obtained by the step (1) and the automatically segmented image obtained by the step (2) to obtain more areas;
(4) The differential map spot marking is accomplished by the following four steps:
1) Scanning the image line by line, forming a sequence of consecutive pixels with the same difference value in each line into a group, recording the starting point, the end point, the line number and the difference value of the group, and classifying the groups with the same difference value into one type;
2) The construction of the connected region comprises the steps of judging each group in the similar groups according to the line numbers, the starting point and the end point of each group: if the clusters in the next row and the previous row are overlapped except the first row, judging that the clusters are the same communication area, otherwise, judging that the clusters are non-communication areas;
3) Traversing each similar group to finish searching a communication area, and combining the image values obtained by vector rasterization and the differential values to solve the communication problem caused by the same differential values of different types;
4) Merging the small image spots, namely merging the small image spots into adjacent most similar difference image spots according to the located vector ID and the ID of the original segmentation if the area is smaller than a threshold value when the small image spots are processed;
step 3, extracting spectrum and texture features of the separated image object;
step 4, corresponding polygons of a ground class in the vector map to obtain image segmentation image spots, and counting feature distribution corresponding to the image spots;
step 5, obtaining the sample object of the ground object according to probability distribution and characteristic consistency criteria, wherein the obtained object characteristics are selected according to categories, the mean value and standard deviation of each characteristic are taken as selection basis, the number of the characteristic numbers, of which the object characteristic values fall within a plurality of times of standard deviation of the ground object characteristics, is taken as the basis for judging the sample, and a sample selection result is obtained; all obtained sample objects are used for supporting classification and change detection of the remote sensing images.
2. The method for automatically selecting the new image sample based on the vector constraint object-oriented according to claim 1, wherein the method is characterized by: in step 2, on the basis of carrying out rasterization processing and image object-oriented segmentation on the vector map, the obtained results are differentiated, and connected domain calculation is carried out on the differential map, so that the vector constraint object-oriented subdivision result is obtained.
3. The method for automatically selecting the new image sample based on the vector constraint object-oriented according to claim 1, wherein the method is characterized by: and 3, calculating various spectrum texture features by taking the object as a unit, and recording the category to which each object belongs according to the vector rasterized marker diagram corresponding to the object.
4. The method for automatically selecting the new image sample based on the vector constraint object-oriented according to claim 1, wherein the method is characterized by: and 4, analyzing the feature distribution by utilizing the feature histogram distribution characteristics.
5. The method for automatically selecting the new image sample based on the vector constraint object-oriented according to claim 1, wherein the method is characterized by: in step 5, whether the object belongs to a certain type of ground feature is judged according to the feature characteristic obeying normal distribution characteristics to select a sample, and the sample is determined through the synthesis of multiple feature judgment results.
6. An automatic selection system of new image samples based on vector constraint object-oriented is characterized in that: an automatic selection method for implementing a new image sample based on vector constraint object-oriented as claimed in any one of claims 1-5.
7. The automatic selection system for new image samples based on vector constraint object oriented as claimed in claim 6, wherein: comprising the following modules, wherein the modules are arranged in a row,
the first module is used for geometrically registering the vector map and the new image by adopting a vector grid registration mode to obtain a common area of the vector map and the new image;
the second module is used for finely dividing the image by adopting an object-oriented image dividing mode under the constraint of the grid boundary after the grid formation of the vector map, so that the dividing boundary is consistent with the vector polygon while the fine division in the polygon is obtained;
the third module is used for extracting spectrum and texture characteristics of the image object obtained by segmentation;
a fourth module, configured to obtain image segmentation patches corresponding to polygons of a ground class in the vector map, and count feature distribution corresponding to the patches;
and the fifth module is used for obtaining the object sample object of the ground object according to the probability distribution sum and the characteristic consistency criterion and is used for classifying and detecting the change of the remote sensing image.
8. The automatic selection system for new image samples based on vector constraint object oriented as claimed in claim 6, wherein: comprising a processor and a memory for storing program instructions, the processor being adapted to invoke the stored instructions in the memory to perform a method for automatically selecting new image samples based on vector constraint object-oriented according to any of claims 1-5.
9. The automatic selection system for new image samples based on vector constraint object oriented as claimed in claim 6, wherein: comprising a readable storage medium having stored thereon a computer program which, when executed, implements a method for automatically selecting new image samples based on vector constraint object-oriented as claimed in any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110776117.9A CN113822900B (en) | 2021-07-09 | 2021-07-09 | Method and system for automatically selecting new image sample based on vector constraint object-oriented |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110776117.9A CN113822900B (en) | 2021-07-09 | 2021-07-09 | Method and system for automatically selecting new image sample based on vector constraint object-oriented |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113822900A CN113822900A (en) | 2021-12-21 |
| CN113822900B true CN113822900B (en) | 2023-09-22 |
Family
ID=78912609
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110776117.9A Active CN113822900B (en) | 2021-07-09 | 2021-07-09 | Method and system for automatically selecting new image sample based on vector constraint object-oriented |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113822900B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108846832A (en) * | 2018-05-30 | 2018-11-20 | 理大产学研基地(深圳)有限公司 | A kind of change detecting method and system based on multi-temporal remote sensing image and GIS data |
| CN109389163A (en) * | 2018-09-29 | 2019-02-26 | 华南理工大学 | A kind of unmanned plane image classification system and method based on topographic map |
| CN110363798A (en) * | 2019-07-24 | 2019-10-22 | 宁波市测绘设计研究院 | A kind of generation method of remote sensing image interpretation sample set |
| 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 |
| CN110555857A (en) * | 2019-08-19 | 2019-12-10 | 浙江工业大学 | semantic edge dominant high-resolution remote sensing image segmentation method |
| CN111339947A (en) * | 2020-02-26 | 2020-06-26 | 苏州中科天启遥感科技有限公司 | Method and system for extracting remote sensing image fuzzy boundary ground object, storage medium and equipment |
| CN112396128A (en) * | 2020-12-08 | 2021-02-23 | 中国铁路设计集团有限公司 | Automatic labeling method for railway external environment risk source sample |
| CN112884791A (en) * | 2021-02-02 | 2021-06-01 | 重庆市地理信息和遥感应用中心 | Method for constructing large-scale remote sensing image semantic segmentation model training sample set |
-
2021
- 2021-07-09 CN CN202110776117.9A patent/CN113822900B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108846832A (en) * | 2018-05-30 | 2018-11-20 | 理大产学研基地(深圳)有限公司 | A kind of change detecting method and system based on multi-temporal remote sensing image and GIS data |
| CN109389163A (en) * | 2018-09-29 | 2019-02-26 | 华南理工大学 | A kind of unmanned plane image classification system and method based on topographic map |
| 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 |
| CN110363798A (en) * | 2019-07-24 | 2019-10-22 | 宁波市测绘设计研究院 | A kind of generation method of remote sensing image interpretation sample set |
| CN110555857A (en) * | 2019-08-19 | 2019-12-10 | 浙江工业大学 | semantic edge dominant high-resolution remote sensing image segmentation method |
| CN111339947A (en) * | 2020-02-26 | 2020-06-26 | 苏州中科天启遥感科技有限公司 | Method and system for extracting remote sensing image fuzzy boundary ground object, storage medium and equipment |
| CN112396128A (en) * | 2020-12-08 | 2021-02-23 | 中国铁路设计集团有限公司 | Automatic labeling method for railway external environment risk source sample |
| CN112884791A (en) * | 2021-02-02 | 2021-06-01 | 重庆市地理信息和遥感应用中心 | Method for constructing large-scale remote sensing image semantic segmentation model training sample set |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113822900A (en) | 2021-12-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Huang et al. | A new building extraction postprocessing framework for high-spatial-resolution remote-sensing imagery | |
| Darwish et al. | Image segmentation for the purpose of object-based classification | |
| CN108830870B (en) | Satellite image high-precision farmland boundary extraction method based on multi-scale structure learning | |
| Chen et al. | Building detection in an urban area using lidar data and QuickBird imagery | |
| Neubert et al. | Evaluation of remote sensing image segmentation quality–further results and concepts | |
| Ke et al. | A review of methods for automatic individual tree-crown detection and delineation from passive remote sensing | |
| CN107016403B (en) | A method of completed region of the city threshold value is extracted based on nighttime light data | |
| CN113449594A (en) | Multilayer network combined remote sensing image ground semantic segmentation and area calculation method | |
| CN113223042B (en) | Intelligent acquisition method and equipment for remote sensing image deep learning sample | |
| Abdollahi et al. | Investigation of SVM and level set interactive methods for road extraction from google earth images | |
| US11804025B2 (en) | Methods and systems for identifying topographic features | |
| CN112883900B (en) | Method and device for bare-ground inversion of visible images of remote sensing images | |
| CN107992856A (en) | High score remote sensing building effects detection method under City scenarios | |
| CN105068918A (en) | Page test method and device | |
| CN112329559A (en) | Method for detecting homestead target based on deep convolutional neural network | |
| Zhang et al. | Impervious surface extraction from high-resolution satellite image using pixel-and object-based hybrid analysis | |
| CN114120141A (en) | All-weather remote sensing monitoring automatic analysis method and system thereof | |
| Zhou et al. | Stratified Object‐Oriented Image Classification Based on Remote Sensing Image Scene Division | |
| CN109657728B (en) | Sample production method and model training method | |
| CN111325184A (en) | Intelligent interpretation and change information detection method for remote sensing image | |
| CN114926738A (en) | Deep learning-based landslide identification method and system | |
| CN107392926B (en) | Remote sensing image feature selection method based on early-stage land thematic map | |
| CN107194405B (en) | Interactive semi-automatic high-resolution remote sensing image building extraction method | |
| CN113822900B (en) | Method and system for automatically selecting new image sample based on vector constraint object-oriented | |
| CN110276270B (en) | High-resolution remote sensing image building area extraction method |
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 |