Intelligent extraction system and method of ground objects based on multi-source data fusion

Abstract

The present invention discloses an intelligent extraction system and method of land objects based on multi-source data fusion, which belongs to the field of land object identification technology, and comprises a multi-source data module, a data processing module and a land object extraction module; the multi-source data module is used for processing the multi-source data, determining each target land object, and obtaining the classification data corresponding to each target land object; the data processing module is used for processing the classification data of each target land object, identifying the classification data corresponding to each target land object, identifying each unit data corresponding to the classification data, analyzing each unit data, and identifying conflicting combinations; processing the conflicting combinations, obtaining representative data corresponding to each conflicting combination, and generating corresponding land object analysis data according to the representative data and the classification data; the land object extraction module is used for extracting land objects according to the land object analysis data, and obtaining the land object features of each target land object; and generating corresponding land object information maps according to the land object features of each target land object.

Description

Technical Field

[0001] The present invention belongs to the technical field of land object recognition, and in particular to a land object intelligent extraction system and method based on multi-source data fusion. Background Art

[0002] With the rapid development of remote sensing technology, geographic information system (GIS) and machine learning, object extraction technology has gradually become a key technology in resource survey, environmental monitoring, urban planning and disaster management. Traditional object extraction methods mainly rely on a single data source, such as a single remote sensing image or GIS data. However, this method often has limitations in extraction accuracy and generalization ability. A single data source usually cannot provide complete and accurate object information. For example, although remote sensing images can provide rich spatial information, in some cases they may be affected by weather, lighting and other conditions, resulting in incomplete or distorted information. Although GIS data is highly structured, it may not cover all object types or may not be updated in a timely manner.

[0003] Therefore, in order to overcome the limitation of a single data source, the present invention provides a system and method for intelligent extraction of land features based on multi-source data fusion. Summary of the invention

[0004] In order to solve the problems existing in the above-mentioned solutions, the present invention provides a system and method for intelligent extraction of land features based on multi-source data fusion.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] The intelligent ground feature extraction system based on multi-source data fusion includes a multi-source data module, a data processing module, and a ground feature extraction module;

[0007] The multi-source data module is used to process the multi-source data, determine each target object, and obtain the classification data corresponding to each target object; the classification data is composed of the combination of each unit data.

[0008] Furthermore, the working method of the multi-source data module includes:

[0009] Connect to each data source, obtain the collected data of each data source, analyze the collected data of each data source through a preset object recognition model, identify various objects, and mark them as target objects; classify each collected data according to each target object, obtain the classified data corresponding to each target object, and mark the collected data of each corresponding data source in the classified data as unit data.

[0010] Furthermore, in the process of classifying each collected data according to each target object, a method for processing the collected data as incomplete data includes:

[0011] Mark the incomplete collected data as special data; identify the data sources, obtain the historical collected data corresponding to each data source, and establish an application analysis model based on the historical collected data. The expression of the application analysis model is Where: s is the specific data; output data is the application value;

[0012] Analyze each specific data through application analysis data to obtain corresponding application values; and process the specific data according to the obtained application values.

[0013] The data processing module is used to process the classification data of each target object, identify the classification data corresponding to each target object, identify each unit data corresponding to the classification data, analyze each unit data, and identify conflict combinations, wherein the conflict combinations include each conflict data;

[0014] The conflicting combinations are processed to obtain representative data corresponding to each conflicting combination, and corresponding ground feature analysis data are generated according to the representative data and the classification data.

[0015] Furthermore, the method for processing the conflicting combination includes:

[0016] Identify each conflicting data in the conflicting combination, and set a corresponding decision feature for each conflicting data;

[0017] Establish a judgment model, mark the conflicting data to be compared as reference data, combine the determining features of the reference data with the determining features of the remaining conflicting data, and obtain a number of judgment input data; input the judgment input data into the judgment model to obtain the judgment value corresponding to each judgment input data; generate a comparison vector corresponding to the reference data according to each judgment value; mark each element value in the comparison vector as yi, i = 1, 2, ..., n, and n is a positive integer;

[0018] According to the formula Calculate the corresponding priority comparison value; where: PK is the priority comparison value; exp(*) is an exponential function with constant e as the base;

[0019] The conflicting data with the largest priority comparison value is selected as the representative data.

[0020] Furthermore, the expression of the judgment model is: Where: s is the judgment input data.

[0021] The feature extraction module is used to extract features according to the feature analysis data, obtain the feature analysis data corresponding to each target feature, and identify each unit data in the feature analysis data;

[0022] Set each subsidiary item, and set a corresponding subsidiary feature template according to each subsidiary item; collect feature data of each unit data according to the subsidiary feature template to obtain feature data of each subsidiary item corresponding to the unit data; analyze the feature data of each subsidiary item to obtain a subsidiary comparison value corresponding to each subsidiary item;

[0023] Generate a corresponding guidance matrix according to each subsidiary comparison value corresponding to each unit data; analyze the feature analysis data and the guidance matrix through a preset feature extraction model to obtain the feature characteristics of each target feature; generate a corresponding feature information map according to the feature characteristics of each target feature.

[0024] Furthermore, the method for analyzing the characteristic data of each auxiliary item includes:

[0025] Obtaining the initial value of each accessory item according to the preset initial value matching table and the accessory item characteristic data; setting the correction value of each accessory item;

[0026] According to the formula CP = CZ + XZ, the corresponding intermediate value of each subsidiary item is calculated; wherein: CP is the intermediate value; CZ is the initial value; XZ is the correction value;

[0027] The corresponding subsidiary comparison value is calculated according to the intermediate value of each subsidiary item and the preset comparison method.

[0028] The intelligent extraction method of ground objects based on multi-source data fusion includes:

[0029] Determine each target object and obtain classification data corresponding to each target object; the classification data includes each unit data;

[0030] Identify each unit data corresponding to the classification data, analyze each unit data, and identify a conflicting combination, wherein the conflicting combination includes each conflicting data;

[0031] Process the conflicting combinations to obtain representative data corresponding to each conflicting combination, and generate corresponding ground feature analysis data based on the representative data and the classified data;

[0032] Identify each unit data in the ground feature analysis data; set each subsidiary item, and set the corresponding subsidiary feature template according to each subsidiary item;

[0033] Collect feature data of each unit data according to the subsidiary feature template to obtain feature data of each subsidiary item corresponding to the unit data; analyze the feature data of each subsidiary item to obtain the subsidiary comparison value corresponding to each subsidiary item;

[0034] Generate a corresponding guidance matrix according to each subsidiary comparison value corresponding to each unit data;

[0035] The feature analysis data and the guidance matrix are analyzed by a preset feature extraction model to obtain the feature characteristics of each target feature; and a corresponding feature information map is generated according to the feature characteristics of each target feature.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] Through the cooperation between the multi-source data module, data processing module and object extraction module, the comprehensive processing of multi-source data is realized, and the feature extraction of the target object is carried out according to the processing results, the extraction accuracy of the object is improved, and more accurate object information is provided for the subsequent application of the object. The accuracy and reliability of object extraction are improved. By integrating multi-source data, the advantages of each can be fully utilized to reduce the errors and uncertainties caused by a single data source. The work efficiency and automation level are improved. Through automated data processing and object recognition algorithms, manual intervention and data processing time can be greatly reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only 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.

[0039] Figure 1 This is a principle block diagram of the present invention. DETAILED DESCRIPTION

[0040] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. 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 are within the scope of protection of the present invention.

[0041] like Figure 1 As shown, the intelligent ground feature extraction system based on multi-source data fusion includes a multi-source data module, a data processing module, and a ground feature extraction module;

[0042] The multi-source data module is used to process multi-source data, connect to various data sources, and obtain the collected data of each data source, such as remote sensing images, geographic information system (GIS) data, lidar data, ground observation data, etc.; establish a corresponding land object recognition model based on a neural network such as a CNN network or a DNN network, and the land object recognition model is used to identify the corresponding land objects, and is established based on the existing technology; analyze the collected data of each data source through the land object recognition model, identify various land objects, and mark them as target land objects; classify each collected data according to the corresponding target land object, obtain the classified data corresponding to each target land object, and mark the data of each corresponding data source in the classified data as unit data, that is, the classified data is composed of the unit data of each data source for the target land object.

[0043] Among them, in the process of classifying each collected data according to each target object, the method for processing the collected data that is incomplete is as follows:

[0044] Mark incomplete collected data as special data; identify the available data sources, obtain the historical collected data corresponding to each data source, and establish the corresponding application analysis model based on the historical collected data. The application analysis model is established based on the isolation forest algorithm; it is used to analyze whether each part of the corresponding collected data can be applied to the subsequent object recognition process. If not, it is considered abnormal and the output application value is 1, otherwise the output application value is 0; such as remote sensing images, identify and analyze each part to determine whether the remaining parts of the image can be applied to the recognition of the corresponding object; the expression is Where: s is the input data, i.e. the corresponding specific data;

[0045] The specific data are analyzed by applying the analysis data, and the corresponding specific data are processed according to the obtained application value, that is, the data corresponding to the corresponding features with an application value of 0 are eliminated.

[0046] The data processing module is used to process the classification data of each target object, identify the classification data corresponding to each target object, identify each unit data corresponding to the classification data, analyze each unit data, identify the conflict combination, the conflict combination is composed of corresponding unit data with data conflict, and mark the unit data in the conflict combination as conflict data; determine whether it is conflict data according to whether the unit data corresponding to the same object conflict, such as the same object, but there are differences in multiple unit data (not illumination, clarity, etc.), then mark it as conflict data;

[0047] The conflicting combinations are processed to obtain representative data corresponding to each conflicting combination, and the representative data are combined with the corresponding unit data in the classified data into land feature analysis data. For the convenience of description, the representative data in the land feature analysis data are also regarded as unit data; other conflicting data are eliminated.

[0048] Methods for handling conflicting combinations include:

[0049] Identify each conflicting data in the conflicting combination, and set a corresponding decision feature for each conflicting data. The decision feature is formed by collecting and combining corresponding decision item data according to each preset decision item. The decision item is set according to data that can be compared and judged which conflicting data is real, such as time, corresponding maintenance, construction and other related data, which is used to judge the real data. Specifically, it is set manually;

[0050] According to the setting method of the decision feature, the corresponding simulation feature data is set according to the corresponding historical data, that is, the decision feature of the simulation setting; the corresponding judgment model is established according to the simulation feature data. The judgment model is established based on the isolation forest algorithm. The conflicting data are compared. The untrue conflicting data is regarded as abnormal data, and the judgment value is 1, otherwise it is 0. When the same data is compared, the judgment value is 0; the expression is s is the judgment input data;

[0051] The conflicting data to be compared are marked as reference data, and the determining features of the reference data are combined with the determining features of the remaining conflicting data to obtain a number of judgment input data; the judgment input data are input into the judgment model to obtain the judgment values ​​corresponding to the judgment input data; a comparison vector corresponding to the reference data is generated according to each judgment value; and the comparison vector is marked as For example, Substitute the corresponding judgment value; mark each element value in the comparison vector as yi, i = 1, 2, ..., n, n is a positive integer;

[0052] According to the formula Calculate the corresponding priority comparison value; where: PK is the priority comparison value; exp(*) is an exponential function with a constant e as the base.

[0053] The conflicting data with the largest priority comparison value is selected as the representative data.

[0054] The feature extraction module is used to extract features according to the feature analysis data, obtain the feature analysis data corresponding to each target feature, and identify each unit data in the feature analysis data;

[0055] According to the influence on the identification and extraction of ground objects, various subsidiary items are set, such as data accuracy, resolution and other data items, and corresponding subsidiary feature templates are set according to each subsidiary item for subsequent feature collection; feature data of each unit data is collected according to the subsidiary feature template to obtain the feature data of each subsidiary item corresponding to the unit data; the feature data of each subsidiary item is analyzed to obtain the subsidiary comparison value corresponding to each subsidiary item;

[0056] Generate a corresponding guidance matrix according to each subsidiary comparison value corresponding to each unit data, where each row corresponds to a unit data and each column corresponds to a subsidiary item; substitute the corresponding subsidiary comparison value to obtain the corresponding matrix, which is marked as the guidance matrix; according to the guidance matrix, the quality of each unit data can be quickly understood, which is convenient for subsequent data processing;

[0057] A corresponding feature extraction model is established based on a CNN network or a DNN network, and the feature extraction model is used to extract feature extraction based on feature analysis data and a guidance matrix; a corresponding training set is established manually for training, and the training set includes input data and output data, and the input data includes feature analysis data and a guidance matrix; the output data includes feature characteristics of the target feature; the feature extraction model after successful training is used for analysis to obtain the feature characteristics of each target feature; and a corresponding feature information map is generated based on the feature characteristics of each target feature.

[0058] The method for analyzing the characteristic data of each subsidiary item includes:

[0059] According to the actual situation of each subsidiary item, the comparison method of the subsidiary comparison value of each subsidiary item is set. For the case of specific numerical values ​​and real data, the lowest value is used as the benchmark and regarded as 1. Other data are compared with it to calculate the corresponding proportional value, which is marked as a negative comparison value. The lowest value needs to be determined according to the meaning of the data. For example, if the data is larger, it means that the data is better, and the lowest value is the minimum value; the larger the value, the worse it is, and its lowest value is the maximum value, and so on; if the data is non-numerical, it is necessary to set the corresponding numerical conversion method according to the existing method to convert it into the corresponding numerical value; if the directly identified data is distorted, such as being affected by various factors during the collection process, resulting in a difference between the identified value and the actual value, it is necessary to make adjustments and compare them after the adjustments. For example, the data accuracy may be affected by external factors, such as remote sensing data;

[0060] Set the initial value of each subsidiary item. For the subsidiary comparison value identified based on the actual data, its initial value is the data it identifies, such as resolution, and the subsequent identification can be carried out directly; for subsidiary items such as data accuracy, it is set according to the actual accuracy of its historical data, and the preset accuracy value is used as the initial value; integrate the initial values ​​corresponding to each subsidiary item into an initial value matching table; establish a corresponding correction model based on a CNN network or a DNN network, and manually establish a corresponding training set for training. The training set includes input data and output data. The input data includes various related data that affect the data of the subsidiary item, and the output data is the correction value; analyze the correction model after successful training to obtain the corresponding correction value; for subsidiary items that do not need to be corrected, such as resolution, no analysis is required, and corresponding marks are made in the initial value matching mark;

[0061] Get the initial value and correction value of each subsidiary item; if there is no correction value, the correction value is regarded as 0;

[0062] According to the formula CP = CZ + XZ, the corresponding intermediate value of each subsidiary item is calculated; wherein: CP is the intermediate value; CZ is the initial value; XZ is the correction value;

[0063] The corresponding subsidiary comparison value is calculated according to the intermediate value of each subsidiary item, that is, the corresponding proportion calculation is performed.

[0064] Through the cooperation among the multi-source data module, data processing module and object extraction module, the comprehensive processing of multi-source data is realized, and the feature extraction of the target object is carried out according to the processing results to improve the extraction accuracy of the object, and provide more accurate object information for subsequent object applications.

[0065] The intelligent extraction method of ground objects based on multi-source data fusion includes:

[0066] Determine each target object and obtain classification data corresponding to each target object; the classification data includes each unit data;

[0067] Identify each unit data corresponding to the classification data, analyze each unit data, and identify a conflicting combination, wherein the conflicting combination includes each conflicting data;

[0068] Process the conflicting combinations to obtain representative data corresponding to each conflicting combination, and generate corresponding ground feature analysis data based on the representative data and the classified data;

[0069] Identify each unit data in the ground feature analysis data; set each subsidiary item, and set the corresponding subsidiary feature template according to each subsidiary item;

[0070] Collect feature data of each unit data according to the subsidiary feature template to obtain feature data of each subsidiary item corresponding to the unit data; analyze the feature data of each subsidiary item to obtain the subsidiary comparison value corresponding to each subsidiary item;

[0071] Generate a corresponding guidance matrix according to each subsidiary comparison value corresponding to each unit data;

[0072] The feature analysis data and the guidance matrix are analyzed by a preset feature extraction model to obtain the feature characteristics of each target feature; and a corresponding feature information map is generated according to the feature characteristics of each target feature.

[0073] The above formulas are all calculated by removing dimensions and taking numerical values. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained by simulating a large amount of data.

[0074] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. The intelligent ground feature extraction system based on multi-source data fusion is characterized by: It includes multi-source data module, data processing module and ground feature extraction module; The multi-source data module is used to process the multi-source data, determine each target object, and obtain the classification data corresponding to each target object; the classification data is composed of the combination of each unit data; The data processing module is used to process the classification data of each target object, identify the classification data corresponding to each target object, identify each unit data corresponding to the classification data, analyze each unit data, and identify conflict combinations, wherein the conflict combinations include each conflict data; Process the conflicting combinations to obtain representative data corresponding to each conflicting combination, and generate corresponding ground feature analysis data based on the representative data and the classified data; The feature extraction module is used to extract features according to the feature analysis data, obtain the feature analysis data corresponding to each target feature, and identify each unit data in the feature analysis data; Set each subsidiary item, and set a corresponding subsidiary feature template according to each subsidiary item; collect feature data of each unit data according to the subsidiary feature template to obtain feature data of each subsidiary item corresponding to the unit data; Analyze the characteristic data of each subsidiary item to obtain the subsidiary comparison value corresponding to each subsidiary item; Generate a corresponding guidance matrix according to each subsidiary comparison value corresponding to each unit data; Analyze the object analysis data and guidance matrix through the preset object extraction model to obtain the object characteristics of each target object; Generate a corresponding feature information map according to the feature characteristics of each target feature.

2. The intelligent ground feature extraction system based on multi-source data fusion according to claim 1 is characterized in that: The working methods of the multi-source data module include: Connect to each data source, obtain the collected data of each data source, analyze the collected data of each data source through a preset object recognition model, identify various objects, and mark them as target objects; classify each collected data according to each target object, obtain the classified data corresponding to each target object, and mark the collected data of each corresponding data source in the classified data as unit data.

3. The intelligent ground feature extraction system based on multi-source data fusion according to claim 2 is characterized in that: In the process of classifying each collected data according to each target object, the method for processing the collected data as incomplete data includes: Mark the incomplete collected data as special data; identify the data sources, obtain the historical collected data corresponding to each data source, and establish an application analysis model based on the historical collected data. The expression of the application analysis model is Where: s is the specific data; output data is the application value; Analyze each specific data through application analysis data to obtain corresponding application values; and process the specific data according to the obtained application values.

4. The intelligent ground feature extraction system based on multi-source data fusion according to claim 1 is characterized in that: Methods for handling conflicting combinations include: Identify each conflicting data in the conflicting combination, and set a corresponding decision feature for each conflicting data; Establish a judgment model, mark the conflicting data to be compared as reference data, combine the determining features of the reference data with the determining features of the remaining conflicting data, and obtain a number of judgment input data; input the judgment input data into the judgment model to obtain the judgment value corresponding to each judgment input data; generate a comparison vector corresponding to the reference data according to each judgment value; mark each element value in the comparison vector as yi, i = 1, 2, ..., n, and n is a positive integer; According to the formula Calculate the corresponding priority comparison value; where: PK is the priority comparison value; exp(*) is an exponential function with constant e as the base; The conflicting data with the largest priority comparison value is selected as the representative data.

5. The intelligent ground feature extraction system based on multi-source data fusion according to claim 4 is characterized in that: The expression of the judgment model is: Where: s is the judgment input data.

6. The intelligent ground feature extraction system based on multi-source data fusion according to claim 1 is characterized in that: The method for analyzing the characteristic data of each subsidiary item includes: Obtaining the initial value of each accessory item according to the preset initial value matching table and the accessory item characteristic data; setting the correction value of each accessory item; According to the formula CP = CZ + XZ, the corresponding intermediate value of each subsidiary item is calculated; wherein: CP is the intermediate value; CZ is the initial value; XZ is the correction value; The corresponding subsidiary comparison value is calculated according to the intermediate value of each subsidiary item and the preset comparison method.

7. The intelligent extraction method of ground objects based on multi-source data fusion is characterized by: The method applied to the intelligent ground feature extraction system based on multi-source data fusion as claimed in any one of claims 1 to 6 comprises: Determine each target object and obtain classification data corresponding to each target object; the classification data includes each unit data; Identify each unit data corresponding to the classification data, analyze each unit data, and identify a conflicting combination, wherein the conflicting combination includes each conflicting data; Process the conflicting combinations to obtain representative data corresponding to each conflicting combination, and generate corresponding ground feature analysis data based on the representative data and the classified data; Identify each unit data in the ground feature analysis data; set each subsidiary item, and set the corresponding subsidiary feature template according to each subsidiary item; Collect feature data of each unit data according to the subsidiary feature template to obtain feature data of each subsidiary item corresponding to the unit data; analyze the feature data of each subsidiary item to obtain the subsidiary comparison value corresponding to each subsidiary item; Generate a corresponding guidance matrix according to each subsidiary comparison value corresponding to each unit data; The feature analysis data and the guidance matrix are analyzed by a preset feature extraction model to obtain the feature characteristics of each target feature; and a corresponding feature information map is generated according to the feature characteristics of each target feature.

Images