Chemical industrial park target ground object identification method based on machine learning

Abstract

The embodiment of the invention discloses a chemical industry park target ground object identification method based on machine learning, and relates to the technical field of data processing. The method comprises the following steps: acquiring satellite image data of a to-be-identified area and chemical industry park interest point vector site data; dividing the to-be-identified region into a plurality of sub-regions by taking a city as a unit; extracting features from the satellite image data of each sub-region, using the feature data and interest point vector site data of each sub-region to construct a chemical industry park classification model of each sub-region, each classification model being based on a machine learning algorithm and using the feature data of the grids in each sub-region as input, and using the feature data of the grids in each sub-region as input; whether the grid belongs to the chemical industry park or not is taken as output; and utilizing the classification model of each sub-region to identify the feature data in the full range of each sub-region to obtain the chemical industry park distribution of each sub-region. According to the embodiment, accurate identification of the chemical industrial park in a large-range area is realized.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data processing, and in particular to a method for identifying target objects in a chemical park based on machine learning. Background Art

[0002] At present, accurate identification of objects in high-resolution remote sensing images based on deep learning has become increasingly popular, but the training process often requires a large number of labeled samples, otherwise the model is easy to overfit on the limited training samples and perform poorly when predicting new unknown data sets. At present, there is a lack of sample data sets of chemical parks that cover general features. Most studies use labeled samples of storage tanks for training and recognition, which cannot be applied to the entire park in practice. In addition, considering the acquisition and processing of high-resolution images, the identification scope of chemical parks is mostly focused on smaller areas such as urban centers, which limits the function of large-scale timely and accurate mapping.

[0003] Relatively speaking, the application of machine learning in remote sensing is relatively mature, and its fine-grained recognition of ground object details is becoming clearer and clearer, and its application objects are becoming more and more extensive. It has become an important method support for research such as land use monitoring and identification. However, there is no accurate method for identifying chemical parks, a specific type of artificial landscape. Summary of the invention

[0004] The embodiment of the present invention provides a method for identifying target objects in a chemical park based on machine learning to solve the above-mentioned technical problems.

[0005] In a first aspect, an embodiment of the present invention provides a method for identifying target objects in a chemical park based on machine learning, comprising:

[0006] Obtain satellite image data of the area to be identified and vector site data of points of interest in the chemical park;

[0007] Taking cities as units, dividing the area to be identified into multiple sub-areas;

[0008] Extract features from satellite image data of each sub-region, and use the feature data and point of interest vector site data of each sub-region to build chemical park classification models for each sub-region, where each classification model is based on a machine learning algorithm, and takes the feature data of the grid in each sub-region as input and whether the grid belongs to a chemical park as output;

[0009] The classification model of each sub-region is used to identify the characteristic data in the entire range of each sub-region to obtain the distribution of chemical parks in each sub-region.

[0010] In a second aspect, an embodiment of the present invention provides an electronic device, the electronic device comprising:

[0011] one or more processors;

[0012] a memory for storing one or more programs,

[0013] When the one or more programs are executed by the one or more processors, the one or more processors implement the chemical park target object identification method based on machine learning as described in any embodiment.

[0014] In a third aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for identifying target objects in a chemical park based on machine learning as described in any embodiment.

[0015] In summary, the embodiments of the present invention provide a method for identifying target objects in chemical parks based on machine learning. Partition modeling is achieved by dividing the area to be identified into multiple sub-areas, reducing intra-class differences, so that the basic framework based on feature data and RF classifier is applicable to the entire large-scale area. At the same time, the differences between sub-areas are reflected through model parameters, taking into account the generalization and accuracy of the model, thereby achieving accurate identification of chemical parks in large-scale areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of a method for identifying target objects in a chemical park based on machine learning provided by an embodiment of the present invention;

[0018] Figure 2 is a flow chart of another method for identifying target objects in a chemical park based on machine learning provided by an embodiment of the present invention;

[0019] Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described clearly and completely below. Obviously, the described embodiments are only 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 belong to the scope of protection of the present invention.

[0021] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0022] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0023] Figure 1 This is a flow chart of a method for identifying a target object in a chemical park based on machine learning provided by an embodiment of the present invention. This method is applicable to identifying chemical parks in a large area across a city and is executed by an electronic device. Figure 1 As shown, the method specifically includes:

[0024] S110, obtaining satellite image data of the area to be identified and vector site data of points of interest in the chemical park.

[0025] The area to be identified in this embodiment is a large area across multiple cities, such as a large area along the Yangtze River. This step first obtains satellite image data of the area and POIs (Points of Interest) vector site data related to the chemical park as the data source of the entire method.

[0026] Optionally, the satellite image data may adopt Sentinel-2 remote sensing image data; the POIs vector site data may be acquired from the Internet using crawler technology, such as acquiring the POIs vector site data along the Yangtze River in a certain year, including the POIs vector site data of CIP (Chemical Industrial Parks) and the POIs vector site data of non-CIP.

[0027] S120: Divide the area to be identified into multiple sub-areas based on cities.

[0028] Generally, the appearance of chemical parks in different regions is quite different, so this embodiment divides the large area into multiple sub-areas, and then models each sub-area to reduce intra-class differences. Specifically, cities are selected as the basic units for sub-area division.

[0029] S130: extracting features from the satellite image data of each sub-region.

[0030] After partitioning, specific data features are extracted from satellite image data for each sub-area, which are used as the basis for chemical park identification. In particular, since the plant installations of CIPs will cause the texture features of their remote sensing images to be completely different from other construction land covers, this embodiment uses the texture features of the grid as the key features to improve classification performance. Optionally, the gray level co-occurrence matrix can be used to calculate the texture features.

[0031] In a specific implementation, taking any sub-region as an example, first determine the window size for calculating texture features. The window size largely affects the extraction effect of texture features. It can be determined artificially based on the scale of the ground object and the characteristic scale of the equipment in the chemical park, or it can be calculated based on the clustering characteristics of the texture features and the minimum area occupied by the chemical park. Optionally, the number of grids in the satellite image data is used as the unit of measurement for the window size, and the calculation of the window size can include the following steps:

[0032] S1-1. Extract multiple positive sample grids belonging to the chemical park and multiple negative sample grids not belonging to the chemical park from the vector site data of the point of interest of the chemical park. Optionally, the preliminary positions of multiple chemical parks and non-chemical parks can be extracted from the POIs data of the current sub-area, and then preliminary grid annotations are performed according to each preliminary position, and multiple typical grids belonging to the chemical park are selected as positive samples, and multiple typical grids belonging to the non-chemical park are selected as negative samples. Exemplarily, the grids that are close to the preliminary position of a chemical park and fall within the area of ​​the chemical park can be regarded as typical grids belonging to the chemical park; the grids that are close to the preliminary position of a non-chemical park and fall within the area of ​​the non-chemical park can be regarded as typical grids belonging to the non-chemical park.

[0033] S1-2, setting an initial window size, and using the initial window size as the current window size. Optionally, the initial window size may be set to 1 remote sensing data grid.

[0034] S1-3, extracting texture features of each positive sample grid and negative sample grid, and clustering each texture feature. Specifically, using K center point clustering method, taking K=2, the texture features of all sample grids are clustered into two categories.

[0035] S1-4, taking the positive sample grid set and the negative sample grid set as the true clustering results, calculating the Rand coefficients of the two clusters. Specifically, the larger the Rand coefficient, the more the two clustering results match.

[0036] S1-5. If the Rand coefficient is greater than the coefficient threshold, the current window size is used as the window size determined according to the clustering characteristics of the texture feature; otherwise, the current window size is increased by a certain step size, and S1-3 is returned according to the new current window size until the final Rand coefficient is greater than the coefficient threshold. This embodiment sets a larger coefficient threshold for the Rand coefficient. When the Rand coefficient is greater than the threshold, it indicates that the clustering result in S1-3 can more accurately distinguish between positive and negative samples, that is, the texture feature with a large current window can make the positive and negative samples show significant differences, thereby playing a stronger distinguishing role in chemical park identification.

[0037] In fact, as the window size continues to increase, the Rand coefficient will show a trend of first rising and then falling, or first rising and then basically stabilizing. This is because a window that is too large or too small is not conducive to extracting texture features that are comparable to the size of the object. In order to avoid the subjectivity of manually selecting the window size or manually setting the coefficient threshold, a wider window variation range can also be selected in advance, so that the Rand coefficients obtained under different window sizes within the range show a law of first rising and then falling, or first rising and then stabilizing. Then, the turning point of the Rand coefficient from rising to falling, or from rising to stabilizing, is selected as the window size determined according to the clustering characteristics of the texture feature.

[0038] Finally, after determining the window size according to the clustering characteristics of the texture features, consider whether the window size is smaller than the minimum floor area of ​​the chemical park. If it is smaller than the minimum floor area, or smaller than the set ratio of the minimum floor area, the window size can be used as the final window size.

[0039] After the window size is determined, the texture features of each grid are calculated according to the window size, and the spectral features of each grid are extracted, including the normalized vegetation index, soil-adjusted vegetation index, modified normalized water index and normalized building index. These spectral features and texture features together constitute the feature space of the grid.

[0040] S140. Using the feature data of each sub-region and the vector site data of the point of interest, a chemical park classification model for each sub-region is constructed respectively, wherein each classification model is based on a machine learning algorithm, and takes the feature data of the grid in each sub-region as input, and whether the grid belongs to a chemical park as output.

[0041] In this step, strict visual annotation is performed on the grid samples to construct an accurate sample set, and the machine learning model is partitioned and trained on the sample set to obtain the chemical park classification model for each sub-region.

[0042] In a specific implementation, first, based on the preliminary locations of each chemical park and non-chemical park obtained from the above POIs vector site data, the grids covered by each chemical park and non-chemical park are strictly visually annotated; the feature data and annotation information of each grid constitute a sample set. Optionally, negative samples can cover a variety of land cover types (such as forests, grasslands, farmlands, waters, bare land) and non-CIPs areas that are similar to but different from chemical parks (such as steel mills, cement plants, manufacturing plants, conventional residential areas, etc.) to reduce the occurrence of false positives.

[0043] Then, select a classifier. Optionally, you can choose RF (Random Forest) classifier as the main classification tool. The reasons include: RF has multilinear feature modeling capabilities, can effectively handle complex relationships in multidimensional feature space, and is suitable for feature modeling in chemical parks; RF is robust, and by integrating multiple decision trees, RF shows strong anti-interference ability to noise and outliers, and can better adapt to the uncertainty in remote sensing data; RF has simple parameters. Compared with other classifiers (such as support vector machines), RF's parameter adjustment is relatively simple, and you only need to adjust the number of decision trees, the maximum number of splits for each tree, and the number of features used for training.

[0044] After the classifier is determined, for any sub-region, the sample data in the sub-region can be used to build the RF classifier of the sub-region. Optionally, the RF classifier can be built using the API of Google Earth Engine (GEE) and the Classifier.randomForest() function can be called for model training. During the training process, the grid search method can be used to tune different parameter combinations (including the number of decision trees, the maximum number of splits per tree, and the number of features used for training) to achieve the best classification effect.

[0045] Furthermore, in the initial stage, the labeled training data can be used to train the RF classifier, and then the sub-region dataset can be predicted. Errors in the prediction and unlabeled data will be returned to the experts for further annotation. The new labeled data will be used to retrain the model, focusing on missed CIPs pixels and false positive prediction errors, ensuring that the model can gradually adapt and correct errors. Among them, false positive prediction errors refer to situations where non-CIPs are misclassified as CIPs, and this error is reduced by collecting more representative negative samples. Missing CIPs pixels refer to pixels that are not correctly classified as chemical parks, and re-labeling is required to improve recognition accuracy.

[0046] In a specific implementation, for the missed CIPs pixels (i.e., the missed positive samples), the positive samples can be re-labeled according to the area and position of the chemical park in the positive samples. Specifically, due to the existence of the window size when calculating the texture features, the chemical park with an area smaller than the window size may be missed. In this case, the positive samples of these chemical parks can be re-labeled and trained again.

[0047] The above-mentioned repeated learning process can also be called active learning. If the test indicators of the model are still not up to standard after multiple active learning, the grids with incorrect classification (referred to as error grids) can be spatially clustered according to their geographical locations to explore whether the error grids show regional clustering characteristics in the geographical space. Among them, spatial clustering refers to clustering the geographical locations of the error grids. Optionally, there are no clustering clusters at the beginning of the clustering. During clustering, each error grid can be traversed once or multiple times, and new clustering clusters are continuously added as the traversal process progresses. Specifically, in each traversal, for the traversed error grid, the following operations are performed: if the error grid does not belong to any existing clustering cluster, for each existing clustering cluster, determine whether the distance between the error grid and the cluster center of the existing clustering cluster is less than a preset threshold, if so, the error grid is divided into the clustering cluster, otherwise, a new cluster is added with the error grid as the clustering center; if the error grid belongs to an existing clustering cluster, for other clustering clusters except the clustering cluster to which the error grid belongs, determine whether the distance between the error grid and the cluster center of the cluster to which it belongs is greater than the distance between the error grid and the cluster center of the other clustering cluster, if so, the error grid is divided into the other clustering cluster. In this way, the error grids with dense distances can be automatically clustered into one category, and the number of clusters is not pre-fixed, but is formed adaptively according to the spatial distribution of the error grids.

[0048] After clustering is completed, select the cluster whose weight is greater than the set threshold, and calculate the density of the incorrectly classified grids and the density of the correctly classified grids in the cluster. Specifically, the weight of each cluster can be determined according to the number of erroneous grids included in each cluster. Optionally, the weight of the cluster is the ratio of the number of erroneous grids in the cluster to the number of all erroneous grids in the current sub-region. If the weight of a cluster is greater than the set threshold, it indicates that the erroneous grids show obvious spatial aggregation in the cluster, and it is necessary to continue to consider whether to separately model the blocks covered by the cluster.

[0049] Optionally, this embodiment calculates the density of the wrong grids and the density of the correct grids in the cluster as a basis for determining whether to model separately. Specifically, the density of the wrong grids is obtained by dividing the number of wrong grids in the cluster by the area of ​​the block covered by the cluster; the density of the correct grids is obtained by dividing the number of correct grids in the cluster by the area of ​​the block covered by the cluster. For the convenience of distinction and description, this embodiment refers to these two densities as the first density and the second density, respectively.

[0050] If the first density is greater than the first threshold and the second density is less than the second threshold (for example, the first threshold is 0.8 and the second threshold is 0.2), that is, the first density is large enough and the second density is small enough, it indicates that the current classification model shows obviously poor performance in this block and is not suitable for most samples in this block. In this case, it is necessary to model this block separately from the current sub-region.

[0051] At this time, continue to detect whether the block covered by the cluster is connected to other sub-regions. If the block is connected to other sub-regions outside the current sub-region, it means that the block is at the border of the adjacent city and may be affected by the adjacent city and have similar chemical park characteristics. Then, the block can be merged into the other sub-regions, and then the merged other sub-regions are modeled. If the block is not connected to any sub-region outside the current sub-region, the block can be modeled separately as an independent sub-region.

[0052] Furthermore, after the blocks covered by the cluster are modeled separately, the test indicators of the existing model can be recalculated for other continuous blocks in the current sub-area except the cluster; if the new test indicators meet the standards, the existing model can be used as the chemical industry park classification model for the other blocks; if the new test indicators do not meet the standards, the other blocks can also be modeled separately as an independent sub-area.

[0053] In this way, this embodiment can timely discover regional classification differences within the same sub-region during the modeling process, and model the blocks that are not suitable for the current model separately from the current sub-region; at the same time, it can also break the barriers between sub-regions, so that the boundary blocks of the sub-regions can have the opportunity to participate in the overall modeling of adjacent sub-regions; ultimately, both partitioning and modeling are more accurate.

[0054] S150, using the classification model of each sub-region to identify the characteristic data in the entire range of each sub-region, and obtain the distribution of chemical parks in each sub-region.

[0055] After the modeling is completed, the final sub-area may be different from the sub-area divided in S120, and each final sub-area corresponds to a final classification model. For any final sub-area, the feature data of all grids within its coverage range are input into the corresponding classification model for processing, and the identification results of whether all grids belong to the chemical park distribution can be obtained. Optionally, the grids belonging to the chemical park are identified as 1, and the grids that do not belong to the chemical park are identified as 0, then the chemical park distribution mask of the sub-area can be obtained. After performing the above operations on all sub-areas, the chemical park distribution mask of the entire area to be identified can be obtained.

[0056] In summary, this embodiment provides a method for identifying target objects in a chemical park based on machine learning, which can achieve the following beneficial effects:

[0057] 1. By adjusting the calculation window of texture features, the texture features of the chemical park can be better captured, so that the feature data can reflect the significant characteristics of the chemical park and achieve more accurate ground object recognition;

[0058] 2. By combining active learning with RF classifiers, the recognition performance of chemical parks is gradually improved, which not only enhances the ability to accurately identify chemical parks in complex environments, but also provides a feasible solution for future remote sensing data processing. This process can be continuously iterated to continuously improve classification accuracy;

[0059] 3. By dividing the area to be identified into multiple sub-areas, partition modeling is achieved to reduce intra-class differences, so that the basic framework based on texture features, spectral features, and RF classifiers is applicable to the entire large area. At the same time, the differences between sub-areas are reflected through model parameters, taking into account both the generalization and accuracy of the model;

[0060] 4. Taking the city as the basic zoning unit, the spatial clustering characteristics of the error grid in the model construction can be used to timely identify the regional differences within the basic zoning unit, and the areas that are obviously not suitable for the current model can be modeled separately. At the same time, the barriers between sub-regions can be broken, so that the boundary blocks of the sub-regions can have the opportunity to participate in the overall modeling of the adjacent sub-regions, ultimately improving the accuracy of zoning, modeling and chemical park classification.

[0061] Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention is shown in FIG. Figure 3 As shown, the device includes a processor 60, a memory 61, an input device 62 and an output device 63; the number of processors 60 in the device can be one or more. Figure 3 A processor 60 is taken as an example; the processor 60, the memory 61, the input device 62 and the output device 63 in the device can be connected by a bus or other means. Figure 3The example of connecting through bus is taken in the following.

[0062] The memory 61 is a computer-readable storage medium that can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the method for identifying target objects in a chemical park based on machine learning in the embodiment of the present invention. The processor 60 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 61, that is, realizing the above-mentioned method for identifying target objects in a chemical park based on machine learning.

[0063] The memory 61 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 61 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 61 may further include a memory remotely arranged relative to the processor 60, and these remote memories may be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0064] The input device 62 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the device. The output device 63 may include a display device such as a display screen.

[0065] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for identifying target objects in a chemical park based on machine learning of any embodiment.

[0066] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, a device or a device or used in combination with it.

[0067] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0068] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0069] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as C or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0070] 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 replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying target objects in a chemical park based on machine learning, characterized in that: include: Obtain satellite image data of the area to be identified and vector site data of points of interest in the chemical park; Taking cities as units, dividing the area to be identified into multiple sub-areas; Extract features from satellite image data of each sub-region, and use the feature data and point of interest vector site data of each sub-region to build chemical park classification models for each sub-region, where each classification model is based on a machine learning algorithm, and takes the feature data of the grid in each sub-region as input and whether the grid belongs to a chemical park as output; The classification model of each sub-region is used to identify the characteristic data in the entire range of each sub-region to obtain the distribution of chemical parks in each sub-region.

2. The method according to claim 1, characterized in that The chemical park classification model of each sub-region is constructed by using the feature data of each sub-region and the vector site data of the point of interest, including: Constructing a sample set according to the feature data and the vector site data of the points of interest of any sub-region, and using the sample set to train and test the chemical park classification model of any sub-region; If the model's test indicators do not meet the standards, the incorrectly classified grids will be spatially clustered according to their geographical locations; Select a cluster whose weight is greater than a set threshold, and calculate a first density of misclassified grids and a second density of correctly classified grids in the cluster; If the first density is greater than a first threshold and the second density is less than a second threshold, detecting whether the block covered by the cluster is connected to other sub-areas, wherein the first threshold is greater than the second threshold; If they are connected, merging the block into the other sub-regions and modeling the merged sub-regions; If not connected, the blocks are modeled separately as independent sub-regions.

3. The method according to claim 2, characterized in that After detecting whether the block covered by the cluster is connected to other sub-areas if the first density is greater than a first threshold and the second density is less than a second threshold, the method further includes: For other blocks in any of the sub-regions except the cluster, recalculate the test index of the model; If the new test index meets the standard, the existing model is used as the chemical industry park classification model for the other blocks; If the new test index is not met, the other blocks are modeled separately as independent sub-areas.

4. The method according to claim 2, characterized in that: The step of constructing a sample set according to the feature data of any sub-area and the vector site data of the point of interest comprises: Based on the vector site data of points of interest in any sub-area, the preliminary locations of multiple chemical parks and non-chemical parks are obtained; Based on the preliminary locations of each chemical park and non-chemical park, visually mark the grids covered by each chemical park and non-chemical park; The sample set is composed of the feature data and annotation information of each grid.

5. The method according to claim 1, characterized in that The extracting features from the satellite image data of each sub-region includes: Determine the window size for calculating texture features by taking the grid of the satellite image data as a measurement unit; According to the window size, the texture features of each grid are calculated.

6. The method according to claim 5, characterized in that The step of determining the window size for calculating texture features by taking the grid of the satellite image data as a unit of measurement includes: According to the clustering characteristics of the texture features and the minimum floor area of ​​the chemical park, the window size for calculating the texture features is determined, wherein the window size is measured in units of grids of the satellite image data.

7. The method according to claim 6, characterized in that Determining the window size for calculating the texture features according to the clustering characteristics of the texture features and the minimum floor area of ​​the chemical park includes: S1-1, extracting a plurality of positive sample grids belonging to the chemical park and a plurality of negative sample grids not belonging to the chemical park from the vector site data of the chemical park point of interest; S1-2, setting an initial window size, and using the initial window size as the current window size; S1-3, extracting texture features of each positive sample grid and negative sample grid within the current window size, and clustering each texture feature into two clusters; S1-4, taking the positive sample grid set and the negative sample grid set as the true clustering results, calculating the Rand coefficients of the two clusters; S1-5. If the Rand coefficient is greater than the coefficient threshold, the current window size is used as the window size determined according to the clustering characteristics of the texture features; otherwise, the current window size is increased by a certain step size, and return to S1-3 according to the new current window size until the final Rand coefficient is greater than the coefficient threshold.

8. The method according to claim 1, characterized in that The chemical park classification model of each sub-region is constructed by using the feature data of each sub-region and the vector site data of the point of interest, including: Constructing a sample set based on the feature data and the point of interest vector site data of any sub-region, and training and testing a chemical park classification model for any sub-region; Extracting the missing grids that are not correctly classified as chemical parks, and re-labeling the missing grids according to the area and location of the chemical parks in the missing grids; The chemical park classification model is further trained using the re-labeled missing grids.

9. An electronic device, characterized in that: include: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the chemical park target object identification method based on machine learning as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the method for identifying target objects in a chemical park based on machine learning as described in any one of claims 1-8 is implemented.

Images