Remote sensing image classification training sample generation method, device and electronic equipment
By processing and equalizing the point of interest data in remote sensing images, high-quality training samples are generated, which solves the problem of long acquisition time for remote sensing image classification training samples in the prior art, and improves classification efficiency and accuracy.
Patent Information
- Application Number
- CN202111327817.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-10
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2041-11-10
AI Technical Summary
The process of obtaining remote sensing image classification training samples in the prior art is time-consuming and labor-intensive, and it is difficult to quickly generate high-quality training samples, which limits the rapid classification application of remote sensing images.
By obtaining the POI data of the point of interest corresponding to the remote sensing image to be classified, the POI surface coverage category is obtained, and it is equalized to generate training samples containing the POI surface coverage category, using geographical location information, attribute fields and preset mappings The relationship analyzes semantic attribute information and meshs to ensure classification accuracy and sample equality.
It realizes the rapid generation of high-quality remote sensing image classification training samples, reduces sample acquisition time, and improves the efficiency and accuracy of remote sensing image classification.
Smart Images

Figure CN114092762B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image processing technology, and in particular to a remote sensing image classification training sample generation method, device and electronic equipment. Background Art
[0002] Remote sensing technology, with its fast and large-scale data acquisition capabilities, is widely used in forest resource planning, crop yield estimation, environmental assessment, disaster monitoring and other fields.
[0003] After acquiring remote sensing images, remote sensing image classification methods are often used to quickly convert remote sensing images from data to information.
[0004] The most commonly used method in remote sensing image classification is supervised classification. Supervised classification methods can ensure relatively good classification accuracy, but obtaining training samples is a time-consuming and labor-intensive process. Rapidly and effectively obtaining training samples has extremely important application value for the rapid classification of remote sensing images. Summary of the invention
[0005] The present invention provides a remote sensing image classification training sample generation method, device and electronic equipment, which are used to solve the defect of slow acquisition of training samples in the prior art and achieve the effect of quickly generating training samples.
[0006] The present invention provides a remote sensing image classification training sample generation method, comprising:
[0007] Obtain the POI data of the point of interest corresponding to the area of the remote sensing image to be classified;
[0008] Processing the POI data to obtain POI surface coverage categories;
[0009] The POI surface coverage category is equalized to obtain a remote sensing image training sample to be classified containing a region corresponding to the remote sensing image to be classified and the POI surface coverage category.
[0010] According to a remote sensing image classification training sample generation method provided by the present invention, the POI data includes geographic location information and attribute fields;
[0011] The step of processing the POI data to obtain the POI surface coverage category includes:
[0012] Projecting the POI onto the remote sensing image to be classified according to the geographical location information to obtain the POI projection coordinates;
[0013] Gridding the POI projection coordinates to obtain a POI grid containing the POI;
[0014] Analyze and process the POI grid according to the attribute field to obtain semantic attribute information of the POI grid;
[0015] According to the semantic attribute information and a preset mapping relationship, the semantic attribute information in the POI grid is mapped to a surface coverage category to obtain a POI surface coverage category.
[0016] According to a remote sensing image classification training sample generation method provided by the present invention, the step of gridding the POI projection coordinates to obtain a POI grid containing the POI includes:
[0017] According to the spatial resolution of the remote sensing image to be classified, the POI projection coordinates are grid-divided using image pixels as grid division units to obtain a POI grid containing the POI, wherein the image pixel is the minimum image unit corresponding to the spatial resolution.
[0018] According to a remote sensing image classification training sample generation method provided by the present invention, the step of analyzing and processing the POI grid according to the attribute field to obtain the semantic attribute information of the POI grid includes:
[0019] Merging the attribute fields of all POIs in the POI grid to obtain text description information;
[0020] The text description information is processed according to a term frequency-inverse document frequency (TF-IDF) method to obtain semantic attribute information of the POI grid.
[0021] According to a remote sensing image classification training sample generation method provided by the present invention, the step of performing equalization processing on the POI surface coverage category includes:
[0022] The remote sensing image to be classified is divided into image blocks with a side length of m, so that the sum of the number of POI surface coverage categories in the image blocks does not exceed a preset number n.
[0023] According to a remote sensing image classification training sample generation method provided by the present invention, the method further includes:
[0024] Detecting the quantity ratio of any two POI surface coverage categories;
[0025] When the quantity ratio exceeds the preset ratio, the following steps are performed:
[0026] Determine a first image block, where the first image block is an image block having the largest sum of POI surface coverage categories among the image blocks;
[0027] A preset number of the POI surface coverage categories are randomly deleted from the first image block.
[0028] The present invention also provides a remote sensing image classification training sample generation device, comprising:
[0029] An acquisition unit, used for acquiring point of interest POI data of a corresponding area of the remote sensing image to be classified;
[0030] A processing unit, used for processing the POI data to obtain a POI surface coverage category;
[0031] The equalization unit is used to perform equalization processing on the POI surface coverage category to obtain a remote sensing image training sample to be classified containing the corresponding area of the remote sensing image to be classified and the POI surface coverage category.
[0032] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the above-mentioned methods for generating remote sensing image classification training samples are implemented.
[0033] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above-mentioned methods for generating remote sensing image classification training samples are implemented.
[0034] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the steps of any of the above-mentioned methods for generating remote sensing image classification training samples are implemented.
[0035] The remote sensing image classification training sample generation method, device and electronic device provided by the present invention obtain POI data, process the POI data to obtain POI surface coverage categories, and perform equalization processing on the POI surface coverage categories to obtain remote sensing image training samples to be classified containing the corresponding areas of the remote sensing images to be classified and the POI surface coverage categories, thereby achieving the effect of quickly generating training samples. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0037] Figure 1 This is one of the flow charts of the remote sensing image classification training sample generation method provided by the present invention;
[0038] Figure 2 It is a schematic diagram of equalization processing of the remote sensing image classification training sample generation method provided by the present invention;
[0039] Figure 3 This is the second flow chart of the remote sensing image classification training sample generation method provided by the present invention;
[0040] Figure 4 It is a schematic diagram of a POI grid of a remote sensing image classification training sample generation method provided by the present invention;
[0041] Figure 5 This is the third flow chart of the remote sensing image classification training sample generation method provided by the present invention;
[0042] Figure 6 This is the fourth flow chart of the remote sensing image classification training sample generation method provided by the present invention;
[0043] Figure 7 It is a structural schematic diagram of a remote sensing image classification training sample generating device provided by the present invention;
[0044] Figure 8 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0045] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0046] Remote sensing technology, with its fast and wide-range data acquisition capabilities, is widely used in forest resource planning, crop yield estimation, environmental assessment, disaster monitoring and other fields. After acquiring remote sensing images, how to quickly convert remote sensing images from data to information has become an important factor restricting the level of remote sensing application. In order to quickly and accurately convert remote sensing images into thematic information, remote sensing image classification methods are often used. Based on whether training samples are needed, remote sensing image classification methods are divided into supervised classification and unsupervised classification. The most commonly used method in remote sensing image classification is supervised classification. Supervised classification methods require manual determination of classification systems, training samples for each category, etc. Supervised classification methods can ensure relatively good classification accuracy, but obtaining training samples is a time-consuming and labor-intensive process. Rapidly obtaining training samples has extremely important application value for the rapid classification of remote sensing images.
[0047] In terms of obtaining training samples, existing studies have proven that active learning technology can play a very important role in this process. The basic idea of active learning is: based on a small number of existing initial training samples, to train a relatively rough classifier model; use the classifier to classify all data, based on a certain method to measure the most informative data set for improving the existing classifier model, and then manually mark the correct category information for the selected data sets to obtain an updated training sample set; then use the updated training sample set to retrain the classifier; iterate the above classification and update training sample set process to obtain a relatively stable and accurate classifier model, and then obtain accurate classification results. The reason why active learning technology is effective is that in each iteration, the active learning strategy selects the most valuable data set for improving the classifier model, avoiding the labeling of meaningless and repeated sample data, thereby greatly saving the time consumption of the sample labeling process.
[0048] Active learning technology can generate enough training samples for remote sensing image classification, but it still requires initial training samples and relies on initial training samples to start the iterative process. The selection of these initial training samples is still a time-consuming and labor-intensive process.
[0049] Therefore, quickly acquiring training samples has extremely important application value for the rapid classification of remote sensing images.
[0050] The present invention provides a remote sensing image classification training sample generation method, such as Figure 1 As shown, the method comprises the following steps:
[0051] S11, obtaining point of interest POI data of the corresponding area of the remote sensing image to be classified;
[0052] For ease of understanding, the technical solution of the present invention will be described below with the electronic device as the execution subject, which does not serve as a limitation.
[0053] Specifically, the electronic device can obtain the POI data of the point of interest corresponding to the area of the remote sensing image to be classified, wherein the remote sensing image to be classified can be input into the electronic device by the user, or can be downloaded by the electronic device through the network. The POI data may include: the name, geographic location information, attribute fields and purpose of the POI, etc. For example, if the POI is a building, the POI information corresponding to the building may include the name of the building, the coordinates representing the geographic location information of the building, the text representing the attribute fields of the building, and the text representing the purpose of the building.
[0054] S12, processing the POI data to obtain the POI surface coverage category;
[0055] Specifically, the electronic device may process the obtained POI data to obtain the POI surface coverage category. For example, according to the POI data of the building in the above example, through processing, it can be obtained that the POI surface coverage category of the building is a built-up area.
[0056] S13, performing equalization processing on the POI surface coverage category to obtain a remote sensing image training sample to be classified containing the corresponding area of the remote sensing image to be classified and the POI surface coverage category;
[0057] Specifically, the electronic device can perform equalization processing on the POI surface coverage categories to obtain a remote sensing image training sample to be classified containing a corresponding area of the remote sensing image to be classified and the POI surface coverage category. After obtaining multiple POI surface coverage categories, the electronic device can perform equalization processing on the multiple POI surface coverage categories, such as Figure 2 As shown, it is possible to avoid an extreme number of POI surface coverage categories in one category, so that the surface coverage categories of POIs of different categories maintain a reasonable ratio.
[0058] In an embodiment of the present invention, POI data of a POI is acquired, the POI data is processed to obtain a POI surface coverage category, and the POI surface coverage category is equalized to obtain a remote sensing image training sample to be classified containing a corresponding area of the remote sensing image to be classified and a POI surface coverage category, thereby achieving the effect of quickly generating training samples.
[0059] According to a remote sensing image classification training sample generation method provided by an embodiment of the present invention, the POI data includes geographic location information and attribute fields; Figure 3 As shown, step S12 may include the following steps:
[0060] S31, projecting the POI onto the remote sensing image to be classified according to the geographical location information to obtain POI projection coordinates;
[0061] Specifically, the POI data includes geographic location information and attribute fields. The electronic device can project the POI onto the remote sensing image to be classified according to the geographic location information in the POI data to obtain the POI projection coordinates. The geographic location information may be longitude and latitude coordinates. The projection coordinates may be obtained by selecting a map projection of the remote sensing image to be classified, such as the Universal Transverse Mercator Projection (UTM) commonly used by satellites. The projection tool may be an open source software package.
[0062] S32, gridding the POI projection coordinates to obtain a POI grid containing the POI;
[0063] Specifically, the electronic device can set the grid size according to a preset size, divide the POI projection coordinates into grids, and obtain a POI grid containing the POI, or set the grid size according to the image pixels corresponding to the spatial resolution of the remote sensing image to be classified, divide the POI projection coordinates into grids, and obtain a POI grid containing the POI. One POI projection coordinate corresponds to one POI.
[0064] S33, analyzing and processing the POI grid according to the attribute field to obtain semantic attribute information of the POI grid;
[0065] Specifically, the electronic device can analyze and process the POI grid according to the attribute fields in the POI data to obtain the semantic attribute information of the POI grid. In an example, a POI grid contains attribute fields of two POIs, and the attribute fields of the two POIs are both buildings. The electronic device performs analysis and processing to obtain semantic attribute information describing that the two POI attribute fields in the POI grid are both buildings.
[0066] S34, mapping the semantic attribute information in the POI grid to a surface coverage category according to the semantic attribute information and a preset mapping relationship, to obtain a POI surface coverage category;
[0067] Specifically, the electronic device can map the POIs to surface coverage categories one by one according to the semantic attribute information based on the preset mapping relationship to obtain the POI surface coverage category. The POI surface coverage category can be set according to actual needs and is not limited to this. In an example, the POI surface coverage category can be farmland, road, forest, grassland, water body and built-up area, etc.
[0068] Using the semantic attribute information in the example of step S33, the electronic device can map POIs whose two attribute fields are both buildings to POI surface coverage categories of the built-up class one by one according to the semantic attribute information and the preset mapping relationship. The preset mapping relationship can be set according to actual needs and is not limited to this.
[0069] In the embodiment of the present invention, the geographical location information in the POI data is projected to obtain POI projection coordinates, the POI projection coordinates are gridded to obtain the POI grid, the POI grid is analyzed and processed according to the attribute field in the POI data to obtain the semantic attribute information of the POI grid, and the POI in the POI grid is mapped to the surface coverage category according to the semantic attribute information of the POI grid to obtain the POI surface coverage category. The technical solution in the embodiment of the present invention can efficiently and quickly complete the step of processing the POI data to obtain the POI surface coverage category, thereby achieving the effect of quickly generating training samples.
[0070] According to a remote sensing image classification training sample generation method provided by an embodiment of the present invention, step S32 may specifically be:
[0071] S321, according to the spatial resolution of the remote sensing image to be classified, using image pixels as grid division units, gridding the POI projection coordinates to obtain a POI grid containing the POI, wherein the image pixel is the minimum image unit corresponding to the spatial resolution;
[0072] Specifically, Figure 4 As shown, the remote sensing image to be classified has a spatial resolution attribute, and the image pixel is the smallest image unit corresponding to the spatial resolution. The electronic device can use the image pixel as a grid division unit according to the spatial resolution of the remote sensing image to be classified, divide the POI projection coordinates into grids according to the image pixels, and obtain a POI grid containing the POI.
[0073] In the embodiment of the present invention, the image pixels corresponding to the spatial resolution of the remote sensing image to be classified are used as grid division units, and the POI projection coordinates are grid divided to obtain a POI grid containing the POI, thereby achieving accurate POI grid division and thus achieving rapid generation of reliable training samples.
[0074] According to an embodiment of the present invention, a remote sensing image classification training sample generation method is provided. Figure 5 As shown, step S33 may include the following steps:
[0075] S51, merging the attribute fields of all POIs in the POI grid to obtain text description information;
[0076] Specifically, the electronic device may merge the attribute fields of all POIs in the POI grid as text description information.
[0077] S52, processing the text description information according to the term frequency-inverse document frequency TF-IDF method to obtain the semantic attribute information of the POI grid;
[0078] Specifically, the electronic device can extract the semantic attribute information of the POI grid with the most classification characteristics from the text description information according to the term frequency-inverse document frequency TF-IDF method. The classification characteristics are text information whose number of occurrences in the text description information increases proportionally, but whose frequency of occurrence in the corpus decreases inversely. The corpus can be set according to actual needs and is not limited at this time.
[0079] In the embodiment of the present invention, the process of merging the attribute fields of POIs in the POI grid and extracting the semantic attribute information of the POI grid with the most classification characteristics according to the IF-IDF method is described in detail, so as to accurately and efficiently obtain the semantic attribute information of the POI grid, and then achieve the effect of quickly generating training samples.
[0080] According to a remote sensing image classification training sample generation method provided by an embodiment of the present invention, step S13 is specifically as follows:
[0081] S131, dividing the remote sensing image to be classified into image blocks with a side length of m, so that the sum of the number of POI surface cover categories in the image blocks does not exceed a preset number n, and obtaining training samples containing the POI surface cover categories.
[0082] Specifically, the electronic device can divide the remote sensing image to be classified into image blocks with a side length of m, so that the sum of the number of POI surface coverage categories in the image blocks does not exceed a preset number n, and obtain training samples containing POI surface coverage categories. Among them, m is an integer greater than 1 and much smaller than the minimum value of the length and width of the remote sensing image to be classified. The preset number n can be set according to actual needs and is not limited to this.
[0083] In the embodiment of the present invention, by dividing the remote sensing image to be classified into image blocks with a side length of m, the sum of the number of POI surface coverage categories in the image blocks does not exceed the preset number n, and the POI surface coverage categories are equalized to obtain training samples containing POI surface coverage categories. The POI surface coverage categories are relatively uniform in spatial distribution, thereby realizing the rapid generation of training samples with uniform spatial distribution.
[0084] According to an embodiment of the present invention, a remote sensing image classification training sample generation method is provided. Figure 6 , the method further comprises:
[0085] S61, detecting the quantity ratio of any two POI surface coverage categories;
[0086] Specifically, the electronic device may detect the quantity ratio of any two POI surface coverage categories.
[0087] When the quantity ratio exceeds the preset ratio, the following steps are performed:
[0088] S62: determining a first image block, where the first image block is an image block having the largest sum of the number of POI surface coverage categories among the image blocks;
[0089] Specifically, when the number ratio exceeds the preset ratio, the electronic device can determine that the image block containing the largest sum of the number of POI surface coverage categories is the first image block. The preset ratio can be set according to actual needs and is not limited thereto.
[0090] S63: Randomly delete a preset number of the POI surface coverage categories from the first image block.
[0091] Specifically, the electronic device may randomly delete a preset number of POI surface coverage categories from the first image block. The preset number may be set according to actual needs and is not limited thereto.
[0092] In an embodiment of the present invention, the quantity ratio of any two POI surface cover categories is detected. When the quantity ratio exceeds a preset ratio, a preset number of POI surface cover categories are deleted from the first image block containing the largest sum of the number of POI surface cover categories, thereby completing the equalization processing of the number of categories, achieving a relatively reasonable ratio of different types of POI surface cover categories, avoiding the occurrence of an extreme number of POI surface cover categories, and thereby achieving the rapid generation of training samples with a balanced number of POI surface cover categories.
[0093] The remote sensing image classification training sample generation device provided by the present invention is described below. The remote sensing image classification training sample generation device described below and the remote sensing image classification training sample generation method described above can be referred to each other. Figure 7 As shown, the device comprises the following units:
[0094] An acquisition unit 71 is used to acquire point of interest POI data of a corresponding area of the remote sensing image to be classified;
[0095] A processing unit 72, configured to process the POI data to obtain a POI surface coverage category;
[0096] The equalization unit 73 is used to perform equalization processing on the POI surface coverage category to obtain a remote sensing image training sample to be classified containing the corresponding area of the remote sensing image to be classified and the POI surface coverage category.
[0097] In an embodiment of the present invention, POI data of a POI is acquired, the POI data is processed to obtain a POI surface coverage category, and the POI surface coverage category is equalized to obtain a remote sensing image training sample to be classified containing a corresponding area of the remote sensing image to be classified and a POI surface coverage category, thereby achieving the effect of quickly generating training samples.
[0098] According to a remote sensing image classification training sample generation device provided by an embodiment of the present invention, the POI data includes geographic location information and attribute fields;
[0099] The processing unit 72 is specifically configured to project the POI onto the remote sensing image to be classified according to the geographical location information to obtain the POI projection coordinates;
[0100] Gridding the POI projection coordinates to obtain a POI grid containing the POI;
[0101] Analyze and process the POI grid according to the attribute field to obtain semantic attribute information of the POI grid;
[0102] According to the semantic attribute information and a preset mapping relationship, the semantic attribute information in the POI grid is mapped to a surface coverage category to obtain a POI surface coverage category.
[0103] According to a remote sensing image classification training sample generation device provided by an embodiment of the present invention, the processing unit 72 is specifically used to grid the POI projection coordinates according to the spatial resolution of the remote sensing image to be classified, using image pixels as grid division units, to obtain a POI grid containing the POI, wherein the image pixel is the smallest image unit corresponding to the spatial resolution.
[0104] According to a remote sensing image classification training sample generation device provided by an embodiment of the present invention, the processing unit 72 is specifically used to merge the attribute fields of all POIs in the POI grid to obtain text description information;
[0105] The text description information is processed according to a term frequency-inverse document frequency (TF-IDF) method to obtain semantic attribute information of the POI grid.
[0106] According to a remote sensing image classification training sample generation device provided by an embodiment of the present invention, the equalization unit 73 is specifically used to divide the remote sensing image to be classified into image blocks with a side length of m, so that the sum of the number of POI surface coverage categories in the image blocks does not exceed a preset number n.
[0107] According to a remote sensing image classification training sample generation device provided by an embodiment of the present invention, the equalization unit 73 is further used to detect the quantity ratio of any two POI surface coverage categories;
[0108] When the quantity ratio exceeds a preset ratio, the balancing unit 73 is further configured to:
[0109] Determine a first image block, where the first image block is an image block having the largest sum of POI surface coverage categories among the image blocks;
[0110] A preset number of the POI surface coverage categories are randomly deleted from the first image block.
[0111] Figure 8 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 8 As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communication interface 820 and the memory 830 communicate with each other through the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute the remote sensing image classification training sample generation method, the method comprising: obtaining the point of interest POI data of the area corresponding to the remote sensing image to be classified; processing the POI data to obtain the POI surface coverage category; performing equalization processing on the POI surface coverage category to obtain the remote sensing image training sample to be classified containing the area corresponding to the remote sensing image to be classified and the POI surface coverage category.
[0112] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0113] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the remote sensing image classification training sample generation method provided by the above-mentioned methods, and the method includes: obtaining point of interest POI data of the area corresponding to the remote sensing image to be classified; processing the POI data to obtain the POI surface coverage category; and performing equalization processing on the POI surface coverage category to obtain a remote sensing image training sample to be classified containing the area corresponding to the remote sensing image to be classified and the POI surface coverage category.
[0114] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the remote sensing image classification training sample generation method provided by the above-mentioned methods, the method comprising: obtaining point of interest (POI) data of the area corresponding to the remote sensing image to be classified; processing the POI data to obtain the POI surface coverage category; performing equalization processing on the POI surface coverage category to obtain a remote sensing image training sample to be classified containing the area corresponding to the remote sensing image to be classified and the POI surface coverage category.
[0115] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0116] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for generating remote sensing image classification training samples, characterized in that: include: Obtain the POI data of the point of interest corresponding to the area of the remote sensing image to be classified; Processing the POI data to obtain POI surface coverage categories; Performing equalization processing on the POI surface coverage category to obtain a remote sensing image training sample to be classified containing the area corresponding to the remote sensing image to be classified and the POI surface coverage category; The POI data includes geographic location information and attribute fields; The step of processing the POI data to obtain the POI surface coverage category includes: Projecting the POI onto the remote sensing image to be classified according to the geographical location information to obtain the POI projection coordinates; Gridding the POI projection coordinates to obtain a POI grid containing the POI; Analyze and process the POI grid according to the attribute field to obtain semantic attribute information of the POI grid; According to the semantic attribute information and a preset mapping relationship, the semantic attribute information in the POI grid is mapped to a surface coverage category to obtain a POI surface coverage category; The step of performing equalization processing on the POI surface coverage category comprises: Divide the remote sensing image to be classified into image blocks with a side length of m, so that the sum of the number of POI surface coverage categories in the image blocks does not exceed a preset number n; The method further comprises: Detecting the quantity ratio of any two POI surface coverage categories; When the quantity ratio exceeds the preset ratio, the following steps are performed: Determine a first image block, where the first image block is an image block having the largest sum of POI surface coverage categories among the image blocks; A preset number of the POI surface coverage categories are randomly deleted from the first image block.
2. The remote sensing image classification training sample generation method according to claim 1, characterized in that: The step of gridding the POI projection coordinates to obtain a POI grid containing the POI includes: According to the spatial resolution of the remote sensing image to be classified, the POI projection coordinates are grid-divided using image pixels as grid division units to obtain a POI grid containing the POI, wherein the image pixel is the minimum image unit corresponding to the spatial resolution.
3. The remote sensing image classification training sample generation method according to claim 1, characterized in that: The step of analyzing and processing the POI grid according to the attribute field to obtain the semantic attribute information of the POI grid includes: Merging the attribute fields of all POIs in the POI grid to obtain text description information; The text description information is processed according to a term frequency-inverse document frequency (TF-IDF) method to obtain semantic attribute information of the POI grid.
4. A remote sensing image classification training sample generation device, characterized in that: include: An acquisition unit, used for acquiring point of interest POI data of a corresponding area of the remote sensing image to be classified; A processing unit, used for processing the POI data to obtain a POI surface coverage category; An equalization unit, configured to perform equalization processing on the POI surface coverage category to obtain a remote sensing image training sample to be classified containing a corresponding area of the remote sensing image to be classified and the POI surface coverage category; The POI data includes geographic location information and attribute fields, and the processing unit is specifically used to: project the POI onto the remote sensing image to be classified according to the geographic location information to obtain the POI projection coordinates; Gridding the POI projection coordinates to obtain a POI grid containing the POI; Analyze and process the POI grid according to the attribute field to obtain semantic attribute information of the POI grid; According to the semantic attribute information and a preset mapping relationship, the semantic attribute information in the POI grid is mapped to a surface coverage category to obtain a POI surface coverage category; The equalization unit is specifically used to divide the remote sensing image to be classified into image blocks with a side length of m, so that the sum of the number of POI surface coverage categories in the image blocks does not exceed a preset number n; The balancing unit is further used to detect the quantity ratio of any two POI surface coverage categories; When the quantity ratio exceeds a preset ratio, the balancing unit is further configured to: Determine a first image block, where the first image block is an image block having the largest sum of POI surface coverage categories among the image blocks; A preset number of the POI surface coverage categories are randomly deleted from the first image block.
5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the remote sensing image classification training sample generation method as described in any one of claims 1 to 3 are implemented.
6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for generating remote sensing image classification training samples as claimed in any one of claims 1 to 3 are implemented.
7. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method for generating remote sensing image classification training samples as claimed in any one of claims 1 to 3 are implemented.
Citation Information
Patent Citations
Vehicle detecting and early-warning method based on machine vision and cascade classifier
CN106295459A
Acupoint seeding bud ratio monitoring method and device based on remote control of unmanned aerial vehicle
CN108537164A
High spatial resolution remote sensing image transfer learning classification method based on OpenStreetMap
CN110598564A
Ground surface coverage image classification method and device and electronic equipment
CN112329852A
Method, apparatus, device and storage medium for determining point of interest area
US20210108941A1
Cited By
Remote sensing image surface feature classification intelligent labeling method based on improved watershed algorithm
CN120783096A
