A natural resource data analysis method and system based on artificial intelligence

Abstract

The present application provides a natural resource data analysis method and system based on artificial intelligence, which obtains the target real-time resource characteristics in the natural resource data set that needs to be processed; when identifying, the recognition ability of the recognition artificial intelligence thread can be used to reduce the errors caused by the analysis ability in the recognition catalog of the general artificial intelligence thread, thereby improving the accuracy of recognition. Based on the natural resource data set that needs to be processed and the target real-time resource characteristics, the general artificial intelligence thread is used to analyze the target real-time resource characteristics to obtain a third possibility distribution, the artificial intelligence analysis thread is used to analyze to obtain a fourth possibility distribution, and the target category is obtained based on the third possibility distribution and the fourth possibility distribution; thus, the analysis ability of the artificial intelligence analysis thread can be further used during analysis, reducing the errors caused by the recognition ability in the analysis catalog of the general artificial intelligence thread, thereby improving the accuracy of analysis.

Description

Technical Field

[0001] The present application relates to the field of data analysis technology, and specifically, to a natural resource data analysis method and system based on artificial intelligence. Background Art

[0002] At present, my country is still facing the basic national conditions of less per capita arable land, low overall quality of arable land, and insufficient reserve resources for arable land. The contradiction and pressure of arable land loss and protection are still prominent. It is necessary not only to identify and analyze resources but also to analyze ecological information. At present, when obtaining natural resource data, manual analysis is required, which wastes a lot of time and manpower. With the continuous development and progress of science and technology, artificial intelligence involves more and more fields. The inventor combines artificial intelligence with the field of natural resources, but how to ensure the accuracy of artificial intelligence analysis of natural resource data is a technical problem that is difficult to solve at present. Summary of the invention

[0003] In order to improve the technical problems existing in the relevant technologies, the present application provides a natural resource data analysis method and system based on artificial intelligence.

[0004] In a first aspect, a natural resource data analysis method based on artificial intelligence is provided, comprising:

[0005] Obtaining the natural resource data sets that require processing;

[0006] Based on the natural resource data set that needs to be processed, a general artificial intelligence thread is used to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a first possibility distribution corresponding to a first natural resource description content, and a recognition artificial intelligence thread is used to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a second possibility distribution corresponding to the first natural resource description content, wherein the first natural resource description content includes a plurality of to-be-determined real-time resource features;

[0007] Based on the first possibility distribution and the second possibility distribution, obtaining target real-time resource characteristics in the natural resource data set that needs to be processed;

[0008] Based on the natural resource data set to be processed and the target real-time resource characteristics, the general artificial intelligence thread is used to analyze the target real-time resource characteristics to obtain a third possibility distribution corresponding to the second natural resource description content, and the artificial intelligence analysis thread is used to analyze the target real-time resource characteristics to obtain a fourth possibility distribution corresponding to the second natural resource description content, wherein the second natural resource description content includes a plurality of to-be-determined categories;

[0009] Based on the third possibility distribution and the fourth possibility distribution, obtaining a target category of the target real-time resource feature;

[0010] Based on the target real-time resource characteristics and the target types of the target real-time resource characteristics, the analysis results of the natural resource data set that needs to be processed are obtained.

[0011] In the present application, the recognition artificial intelligence thread is trained based on the sharing coefficient between the first regression analysis real-time resource feature and the current directory of each first example; the artificial intelligence analysis thread is trained based on the sharing coefficient between the second regression analysis type and the type directory of each second example; the general artificial intelligence thread is trained based on the sharing coefficient between the third analysis result of each third example and the example directory, the third analysis result includes the third regression analysis real-time resource feature and the third type, and the example directory includes the current directory and the type directory of the third example.

[0012] In the present application, the method further comprises:

[0013] Based on the natural resource data set that needs to be processed, an artificial intelligence analysis thread is used to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a fifth possibility distribution corresponding to the first natural resource description content;

[0014] Based on the natural resource data set to be processed and the target real-time resource characteristics, the recognition artificial intelligence thread is used to analyze the target real-time resource characteristics to obtain a sixth possibility distribution corresponding to the second natural resource description content;

[0015] Wherein, obtaining the target real-time resource characteristics in the natural resource data set to be processed based on the first possibility distribution and the second possibility distribution includes:

[0016] Based on the first possibility distribution, the second possibility distribution and the fifth possibility distribution, obtaining the target real-time resource characteristics;

[0017] Wherein, obtaining the target type of the target real-time resource feature based on the third possibility distribution and the fourth possibility distribution includes:

[0018] Based on the third possibility distribution, the fourth possibility distribution and the sixth possibility distribution, a target type corresponding to the target real-time resource feature is obtained.

[0019] In the present application, the method further comprises:

[0020] Get tips for selecting catalogs of natural resource data sets;

[0021] Among them, the prompt natural resource data set includes an identification prompt feature and an analysis prompt feature. The identification prompt feature indicates the identification of real-time resource features of the target field in the natural resource data set that needs to be processed, and the analysis prompt feature indicates the analysis of the identified real-time resource features.

[0022] In the present application, the first possibility distribution is obtained based on the natural resource data set to be processed and the recognition prompt features, and using the general artificial intelligence thread to perform real-time resource feature recognition on the natural resource data set to be processed; the second possibility distribution is obtained based on the natural resource data set to be processed and the recognition prompt features, and using the recognition artificial intelligence thread to perform real-time resource feature recognition on the natural resource data set to be processed;

[0023] Among them, the third possibility distribution is obtained by analyzing the target real-time resource characteristics based on the natural resource data set that needs to be processed, the target real-time resource characteristics and the analysis prompt characteristics, using a general artificial intelligence thread to analyze the target real-time resource characteristics; the fourth possibility distribution is obtained by analyzing the target real-time resource characteristics based on the natural resource data set that needs to be processed, the target real-time resource characteristics and the analysis prompt characteristics, using an artificial intelligence analysis thread.

[0024] In the present application, obtaining the target real-time resource features in the natural resource data set to be processed based on the first possibility distribution and the second possibility distribution includes:

[0025] determining a first identified difference of the second likelihood distribution based on a sharing coefficient between the first likelihood distribution and the second likelihood distribution;

[0026] determining a first debugging possibility for the first possibility distribution based on a comparison result between the second possibility distribution and the first identified difference;

[0027] Debugging the first possibility distribution based on the first debugging possibility to obtain a first target possibility distribution;

[0028] Based on the first target possibility distribution, the target real-time resource characteristics are obtained.

[0029] In the present application, obtaining the target real-time resource feature based on the first possibility distribution, the second possibility distribution and the fifth possibility distribution includes:

[0030] determining a first identified difference of the second likelihood distribution based on a sharing coefficient between the first likelihood distribution and the second likelihood distribution;

[0031] determining a second identified difference of the fifth likelihood distribution based on a sharing coefficient between the first likelihood distribution and a fifth likelihood distribution;

[0032] determining a first debugging possibility for the first possibility distribution based on a comparison result between the second possibility distribution and the first identified difference;

[0033] debugging the first possibility distribution based on the first debugging possibility;

[0034] Determining a second target possibility distribution based on a comparison result between the debugged first possibility distribution and the second identification difference;

[0035] Based on the second target possibility distribution, the target real-time resource characteristics are obtained.

[0036] In the present application, the first possibility distribution and the second possibility distribution are obtained by:

[0037] Based on the natural resource data set that needs to be processed, a general artificial intelligence thread is used to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a first original possibility distribution corresponding to the first natural resource description content, and a recognition artificial intelligence thread is used to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a second original possibility distribution corresponding to the first natural resource description content;

[0038] The first original possibility distribution includes the first original possibility of each pending real-time resource feature, and the second original possibility includes the second original possibility of each pending real-time resource feature;

[0039] Based on the first original possibility distribution and the second original possibility distribution, determining a target undetermined layer from each of the undetermined real-time resource characteristics, in which both the first original possibility and the second original possibility are not lower than a target specified value;

[0040] The possibilities corresponding to the non-target undetermined layers in the first original possibility distribution and the second original possibility distribution are deleted respectively to obtain the first possibility distribution and the second possibility distribution.

[0041] In the present application, the natural resource data set to be processed includes a number of target real-time resource features, and the analysis result includes a number of target real-time resource features and target types of the target real-time resource features; the analysis result is generated by executing x directories, each directory is an identification directory or an analysis directory, and the natural resource data set generated by each directory is the target real-time resource feature or the target type; wherein the execution process of the ath directory includes:

[0042] Determine the type of natural resource data set to be generated based on the generated natural resource data set of the a-1th directory, wherein the type of natural resource data set is an identification type or an analysis type, and the type of natural resource data set corresponding to the first directory is an identification type;

[0043] Based on the natural resource data set to be processed and the natural resource data set generated by the previous a-1 directories, a general artificial intelligence thread is used to obtain a general possibility distribution, an identification artificial intelligence thread is used to obtain a recognition possibility distribution, and an artificial intelligence analysis thread is used to obtain an analysis possibility distribution;

[0044] Wherein, the general possibility distribution is the first possibility distribution or the third possibility distribution, the identification possibility distribution is the second possibility distribution or the sixth possibility distribution, and the analysis possibility distribution is the fourth possibility distribution or the fifth possibility distribution;

[0045] Based on the general possibility distribution, the identification possibility distribution and the analysis possibility distribution, a natural resource data set of the a-th catalog is generated.

[0046] In this application, the recognition artificial intelligence thread is trained in the following way:

[0047] Obtaining a first training set, the first training set comprising a plurality of first example natural resource data sets having example directories, wherein the example directory of each first example comprises a current directory of the first example;

[0048] Based on each first example, using the recognition artificial intelligence thread to be trained to perform real-time resource feature recognition on each first example, to obtain a first regression analysis real-time resource feature of each first example;

[0049] Based on each of the first examples and the first regression analysis real-time resource feature, the first regression analysis real-time resource feature is analyzed by using a recognition artificial intelligence thread to be trained to obtain a first regression analysis type of the first regression analysis real-time resource feature, and based on the first regression analysis real-time resource feature and the first regression analysis type of each of the first examples, an analysis result of each of the first examples is obtained;

[0050] A first training evaluation index is determined based on a sharing coefficient between the real-time resource feature and the current directory in a first regression analysis of each first example, and a thread coefficient of the recognition artificial intelligence thread to be trained is optimized based on the first training evaluation index.

[0051] In this application, the artificial intelligence analysis thread is trained in the following way:

[0052] Obtaining a second training set, wherein the second training set includes a plurality of second examples having example directories, wherein the example directory of each second example includes a category directory of each current directory of the second example;

[0053] Based on each second example, using the artificial intelligence analysis thread to be trained to perform real-time resource feature recognition on each second example, to obtain a second regression analysis real-time resource feature of each second example;

[0054] Based on each of the second examples and the second regression analysis real-time resource characteristics, the second regression analysis real-time resource characteristics are analyzed using the artificial intelligence analysis thread to be trained to obtain a second regression analysis type of the second regression analysis real-time resource characteristics, and based on the second regression analysis real-time resource characteristics and the second regression analysis type of each of the second examples, an analysis result of each of the second examples is obtained;

[0055] Based on the sharing coefficient between the second regression analysis category and the category catalog of each second example, a second training evaluation index is determined, and the thread coefficient of the artificial intelligence analysis thread to be trained is optimized based on the second training evaluation index.

[0056] In this application, the general artificial intelligence thread is trained in the following way:

[0057] Obtaining a third training set, the third training set comprising a plurality of third examples having example directories, wherein the example directory of each third example comprises a current directory of the third example and a category directory of each current directory;

[0058] Based on each third example, using the general artificial intelligence thread to be trained to perform real-time resource feature recognition on each third example, to obtain a third regression analysis real-time resource feature of each third example;

[0059] Based on each of the third examples and the third regression analysis real-time resource feature, the third regression analysis real-time resource feature is analyzed by using a general artificial intelligence thread to be trained to obtain a third regression analysis type of the third regression analysis real-time resource feature, and based on the third regression analysis real-time resource feature and the third regression analysis type of each of the third examples, an analysis result of each of the third examples is obtained;

[0060] Based on the third analysis result of each third example and the sharing coefficient between the example directories, a third training evaluation index is determined, and based on the third training evaluation index, the thread coefficient of the general artificial intelligence thread to be trained is optimized.

[0061] In a second aspect, a natural resource data analysis system based on artificial intelligence is provided, comprising a processor and a memory communicating with each other, wherein the processor is used to read a computer program from the memory and execute it to implement the above method.

[0062] The embodiment of the present application provides a natural resource data analysis method and system based on artificial intelligence. Based on the natural resource data set to be processed, a general artificial intelligence thread is used to identify the real-time resource features to obtain a first possibility distribution, and a recognition artificial intelligence thread is used to identify the real-time resource features to obtain a second possibility distribution. Based on the first possibility distribution and the second possibility distribution, the target real-time resource features in the natural resource data set to be processed are obtained; so that during identification, the recognition ability of the recognition artificial intelligence thread can be used to reduce the errors caused by the analysis ability in the recognition catalog of the general artificial intelligence thread, and improve the accuracy of identification. In addition, based on the natural resource data set to be processed and the target real-time resource features, the general artificial intelligence thread is used to analyze the target real-time resource features to obtain a third possibility distribution, and the artificial intelligence analysis thread is used to analyze to obtain a fourth possibility distribution, and the target category is obtained based on the third possibility distribution and the fourth possibility distribution; so that during analysis, the analysis ability of the artificial intelligence analysis thread can be further used to reduce the errors caused by the recognition ability in the analysis catalog of the general artificial intelligence thread, and improve the accuracy of analysis. Thus, natural resource data can be obtained more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0064] Figure 1 A flowchart of a natural resource data analysis method based on artificial intelligence provided in an embodiment of the present application. DETAILED DESCRIPTION

[0065] In order to better understand the above technical scheme, the technical scheme of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0066] See also Figure 1, showing a natural resource data analysis method based on artificial intelligence, which may include the technical solutions described in the following steps 201-206.

[0067] Step 201: The electronic device obtains a natural resource data set that needs to be processed.

[0068] Among them, natural resource data can be obtained through geological exploration equipment, electromagnetic methods, and drones. The natural resource data set includes: one is non-renewable resources, such as various metal and non-metallic minerals, fossil fuels, etc., which require a long geological period to form; the second is renewable resources, which refer to biological, water, land resources, etc., which can be reproduced or recycled in a short period of time; the third is inexhaustible resources, such as wind power, solar energy, etc., which will not lead to a reduction in storage after being used. In this application, the main focus is on land resources and soil composition, so that it can be determined what can be planted in the area, so that land resources can be fully utilized, or when planting some crops, the soil lacks what crops need more resources, and timely artificial intervention can be carried out to ensure the growth of crops. The natural resource data set here is understood as a land resource data set, etc. The present invention is actually applied to cultivated land, but the technology is not limited to this field.

[0069] For example, the natural resource data set that needs to be processed may include real-time resource characteristics of the target area, such as soil composition characteristics (which can represent soil fertility information, etc.), underground mineral characteristics (such as iron ore resource information and copper ore resource information, etc.).

[0070] In a possible implementation example, the target field may be a natural resource field; real-time resource characteristics of the natural resource field (real-time soil information of cultivated land is described in this application).

[0071] Among them, the present application can parse the selection catalog into an identification catalog and an analysis catalog, and identify the real-time resource features in the natural resource data set to be processed through the identification catalog to obtain the target real-time resource features; and analyze the identified target real-time resource features through the analysis catalog to obtain the target category of the target real-time resource features, thereby obtaining the analysis result based on the target real-time resource features and the target category. The catalog in the present application can be understood as a label.

[0072] Step 202: Based on the natural resource data set that needs to be processed, the electronic device uses a general artificial intelligence thread to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a first possibility distribution corresponding to the first natural resource description content, and uses an identification artificial intelligence thread to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a second possibility distribution corresponding to the first natural resource description content.

[0073] Exemplarily, real-time resource features are understood as natural resource information collected directly on site. Artificial intelligence threads can include CNN, decision networks, and the like.

[0074] The first natural resource description content includes several pending real-time resource features. The first possibility distribution includes the first possibility that each pending real-time resource feature is included in the natural resource data set to be processed, and the second possibility distribution includes the second possibility that each pending real-time resource feature is included in the natural resource data set to be processed. In other words, for the identification catalog, the possibility distribution (first possibility distribution, second possibility distribution) includes the possibility that the real-time resource feature analyzed by the thread (general artificial intelligence thread, identification artificial intelligence thread) is the pre-configured real-time resource feature set (pending current natural resource description content).

[0075] It should be understood that the natural resource data set that needs to be processed may include real-time resource features of one or several target fields. When identifying the real-time resource features of different target fields through an artificial intelligence thread, for example, when identifying the real-time resource features of one of the target fields, a regression analysis can be performed on the possibility distribution of each pending real-time resource feature corresponding to the target field, or a regression analysis can be performed on the possibility of each pending real-time resource feature corresponding to each target field. The embodiments of the present application do not limit the fields of real-time resource features to be identified, the number of fields, etc.

[0076] The first possibility distribution and the second possibility distribution are the identification possibility distributions of the general artificial intelligence thread and the identification artificial intelligence thread, respectively. In the first possibility distribution and the second possibility distribution, each of the same pending real-time resource features can correspond. For example, a natural resource description content including a large number of real-time resource features can be pre-configured. When executing the identification catalog, each real-time resource feature in the natural resource description content is determined as a pending real-time resource feature. The general artificial intelligence thread and the identification artificial intelligence thread respectively perform regression analysis on the possibility that each pending real-time resource feature in the natural resource description content is a real-time resource feature in the natural resource data set that needs to be processed.

[0077] Among them, in the first possibility distribution and the second possibility distribution, the possibilities for the same pending real-time resource feature may be different or the same.

[0078] In a possible implementation embodiment, the first possibility distribution and the second possibility distribution are obtained after further constraining the possibility distributions directly output by the general artificial intelligence thread and the recognition artificial intelligence thread. Optionally, the implementation method of step 202 may include the following steps 2021-2023: Step 2021, based on the natural resource data set that needs to be processed, use a general artificial intelligence thread to perform real-time resource feature identification on the natural resource data set that needs to be processed, and obtain a first original possibility distribution corresponding to the first natural resource description content, and use an identification artificial intelligence thread to perform real-time resource feature identification on the natural resource data set that needs to be processed, and obtain a second original possibility distribution corresponding to the first natural resource description content; wherein the first original possibility distribution includes the first original possibility of each pending real-time resource feature, and the second original possibility includes the second original possibility of each pending real-time resource feature; Step 2022, based on the first original possibility distribution and the second original possibility distribution, determine from each pending real-time resource feature a target pending layer in which both the first original possibility and the second original possibility are not lower than the target specified value; Step 2023, delete the possibilities corresponding to the non-target pending layers in the first original possibility distribution and the second original possibility distribution, respectively, to obtain the first possibility distribution and the second possibility distribution.

[0079] It should be noted that the first original possibility distribution and the second original possibility distribution are the original possibility distributions directly output by the general artificial intelligence thread and the recognition artificial intelligence thread, respectively, and the original possibility distributions can be constrained by pre-configured constraint requirements. Among them, the original possibility of the pending real-time resource characteristics that do not meet the constraint requirements can be deleted to constrain each possibility value in the first possibility distribution and the second possibility distribution within a certain range. Possibility can be understood as probability.

[0080] Among them, the constraint requirement may be: the original possibility is not less than the target specified value. The target specified value may be configured based on needs, and this application does not limit this. For example, the target specified value corresponding to each original possibility distribution may be configured, and the target specified value may be 80%, 60%, etc. of the maximum original possibility in the corresponding original possibility distribution.

[0081] Among them, the method for determining each target pending layer may include the following steps: the electronic device respectively selects each first pending real-time resource feature whose first original possibility is not lower than the first target specified value from each pending real-time resource feature of the first original possibility distribution to obtain a first current set; and, from each pending real-time resource feature of the second original possibility distribution, selects each second pending real-time resource feature whose second original possibility is not lower than the second target specified value to obtain a second current set; and, performs an intersection operation on the first current set and the second current set to obtain each third pending real-time resource feature commonly included in the first current set and the second current set, and determines each third pending real-time resource feature as the target pending layer.

[0082] Among them, the first target specified value and the second target specified value can be the same or different, and can be configured based on needs, and this application does not limit this.

[0083] The non-target pending layer refers to the pending real-time resource features other than the target pending layer in each original possibility distribution (the first original possibility distribution and the second original possibility distribution).

[0084] The deleting of the possibilities of the non-target undetermined layers refers to deleting or cleaning out the original possibilities of the non-target undetermined layers in the original possibility distributions.

[0085] Step 203: The electronic device obtains target real-time resource features in the natural resource data set that needs to be processed based on the first possibility distribution and the second possibility distribution.

[0086] The electronic device may determine a first target possibility distribution by combining the first possibility distribution and the second possibility distribution, and determine a target real-time resource feature in each of the pending real-time resource features based on the first target possibility distribution.

[0087] For example, for the ath pending real-time resource feature, the target possibility of the ath pending real-time resource feature in the target possibility distribution is Da; based on the target possibility of each pending real-time resource feature covered by the first target possibility distribution, the pending real-time resource feature with the largest target possibility can be selected and determined as the target real-time resource feature.

[0088] In a possible implementation example, the electronic device may sum the first possibility distribution and the second possibility distribution to obtain a first target possibility distribution.

[0089] In one possible implementation, the recognition artificial intelligence thread is trained based on the sharing coefficient between the first regression analysis real-time resource feature of each first example and the current directory; the artificial intelligence analysis thread is trained based on the sharing coefficient between the second regression analysis category of each second example and the category directory; the general artificial intelligence thread is trained based on the third analysis result of each third example and the sharing coefficient between the example directory, the third analysis result includes the third regression analysis real-time resource feature and the third category, and the example directory includes the current directory and the category directory of the third example.

[0090] Therefore, by decoupling the thread recognition capability and analysis capability during the training phase, the capabilities of different artificial intelligence threads in different directories are independently trained, the recognition capability of the artificial intelligence thread is optimized, and the analysis capability of the artificial intelligence analysis thread is optimized, so that the capabilities of different artificial intelligence threads in specific directories can be enhanced.

[0091] It should be noted that since the general artificial intelligence thread jointly trains the recognition ability and the analysis ability in the training stage, there may be mutual interference between the recognition ability and the analysis ability, which may cause errors caused by the analysis ability in the recognition catalog. In the training stage, the thread recognition ability and the analysis ability are decoupled, and the recognition ability of the recognition artificial intelligence thread is optimized. Therefore, the second possibility distribution is the regression analysis possibility that is not interfered by the analysis ability. The first possibility distribution and the second possibility distribution are used for joint regression analysis to obtain the first target possibility distribution, which helps to reduce the error problem caused by the analysis ability of the general artificial intelligence thread in the recognition catalog. The probability of each pending real-time resource feature in the obtained first target possibility distribution is also more accurate, thereby improving the accuracy of the recognition result.

[0092] In a possible implementation, the electronic device may determine the first target possibility distribution based on the similarity between the first possibility distribution and the second possibility distribution. For example, step 203 may include the following steps 2031-2034: step 2031, based on the sharing coefficient between the first possibility distribution and the second possibility distribution, determine the first identification difference of the second possibility distribution; step 2032, based on the comparison result between the second possibility distribution and the first identification difference, determine the first debugging possibility for the first possibility distribution; step 2033, based on the first debugging possibility, debug the first possibility distribution to obtain the first target possibility distribution; step 2034, based on the first target possibility distribution, obtain the target real-time resource characteristics.

[0093] By debugging the first possibility distribution through the first debugging possibility, the parts with large differences in the first possibility distribution and the second possibility distribution can be effectively removed. At the same time, the pending real-time resource features with higher possibilities in the first possibility distribution and the first debugging possibility distribution can be spliced, and the parts with higher sharing coefficients in the two possibility distributions can be more spliced.

[0094] Based on this, the ability of the first target possibility distribution to express the actual target real-time resource characteristics can be enhanced, which helps to more accurately find the target real-time resource characteristics with the highest possibility, thereby improving the accuracy of the recognition results.

[0095] Therefore, by using the first possibility distribution and the second possibility distribution to obtain the target real-time resource characteristics, the recognition ability of the identification artificial intelligence thread can be further used under the premise of the recognition ability of the general artificial intelligence thread to improve the accuracy of the recognition directory. In addition, by subtracting the first recognition difference in the second possibility distribution to obtain the first debugging possibility, the recognition ability of the identification artificial intelligence thread can be effectively corrected, and the first possibility distribution is debugged using the first debugging possibility, so that the recognition ability of the general artificial intelligence thread and the identification artificial intelligence thread is effectively spliced, thereby improving the accuracy of the first target possibility distribution and the accuracy of the recognition result.

[0096] Step 204: Based on the natural resource data set to be processed and the target real-time resource characteristics, the electronic device uses the general artificial intelligence thread to analyze the target real-time resource characteristics to obtain a third possibility distribution corresponding to the second natural resource description content, and uses the artificial intelligence analysis thread to analyze the target real-time resource characteristics to obtain a fourth possibility distribution corresponding to the second natural resource description content.

[0097] Illustratively, the possibility distribution is the probability situation of natural resource types and data.

[0098] The second natural resource description content includes a plurality of pending categories. The third possibility distribution includes the type of the target real-time resource feature as the third possibility of each pending category, and the fourth possibility distribution includes the type of the target real-time resource feature as the fourth possibility of each pending category.

[0099] Among them, for the analysis catalog, a second natural resource description content including several categories may be pre-configured. For example, if the target field is the natural resource field, the categories included in the second natural resource description content may be the current analysis of the natural resource field. The possibility that each undetermined category in the second natural resource description content is the category of the target real-time resource feature may be regressively analyzed through the general artificial intelligence thread and the artificial intelligence analysis thread respectively.

[0100] In this step, the natural resource data set to be processed and the target real-time resource characteristics can be input into the general artificial intelligence thread and the artificial intelligence analysis thread respectively; the general artificial intelligence thread is used to analyze the target real-time resource characteristics to obtain a third possibility distribution; and the artificial intelligence analysis thread is used to analyze the target real-time resource characteristics to obtain a fourth possibility distribution.

[0101] It should be noted that the third possibility distribution and the fourth possibility distribution are the analysis possibility distributions of the general artificial intelligence thread and the artificial intelligence analysis thread respectively; in the two possibility distributions, the possibilities corresponding to the same undetermined category may be different or the same.

[0102] In addition, the identification catalog and the analysis catalog may use the common natural resource description content. The first natural resource description content and the second natural resource description content may be the same or different. For example, the common natural resource description content may include each current word and category word. In the identification catalog, the possibility that all words in the common natural resource description content are real-time resource features in the natural resource data set to be processed can be regressively analyzed; in the analysis catalog, the possibility that all words in the natural resource description content are categories of the target real-time resource features can be regressively analyzed.

[0103] In a possible implementation, the third possibility distribution and the fourth possibility distribution can also be obtained by further constraining the possibility distribution directly output by the general artificial intelligence thread and the artificial intelligence analysis thread. For example, the implementation method of step 204 may include the following steps 2041-2043: Step 2041: Based on the natural resource data set to be processed and the target real-time resource characteristics, the general artificial intelligence thread is used to analyze the target real-time resource characteristics to obtain a third original possibility distribution corresponding to the second natural resource description content, and the artificial intelligence analysis thread is used to analyze the target object layer to obtain a fourth original possibility distribution corresponding to the second natural resource description content; wherein the third original possibility distribution includes the third original possibility of each pending category, and the fourth original possibility includes the fourth original possibility of each pending category; Step 2042: Based on the third original possibility distribution and the fourth original possibility distribution, from each pending category, determine the target pending category whose third original possibility and fourth original possibility are not lower than the target specified value; Step 2043: Deleting the possibilities corresponding to the non-target pending categories in the third original possibility distribution and the fourth original possibility distribution, respectively, to obtain the third possibility distribution and the fourth possibility distribution.

[0104] It should be noted that the third original possibility distribution and the fourth original possibility distribution are the original possibility distributions directly output by the general artificial intelligence thread and the artificial intelligence analysis thread, respectively. The possibility in each original possibility distribution can be constrained by pre-configured constraint requirements, and the possibility value in each original possibility distribution can be constrained within a certain range. For example, the constraint requirement can be: the original possibility is not lower than the target specified value. For example, the target specified value corresponding to each original possibility distribution can be configured, and the target specified value can be 80%, 60%, etc. of the maximum original possibility in the corresponding original possibility distribution.

[0105] It should be noted that the method for determining each target pending type can be the same as the method for determining each target pending layer in the above steps, and will not be repeated here. The deletion process of the original possibility of the non-target pending type can also be carried out in the same way as the deletion process of the original possibility of each non-target pending layer in the above steps, and will not be repeated here.

[0106] Step 205: The electronic device obtains the target category of the target real-time resource feature based on the third possibility distribution and the fourth possibility distribution.

[0107] In this step, the electronic device can combine the third possibility distribution and the fourth possibility distribution to determine the third target possibility distribution; based on the third target possibility distribution, determine the target category in each pending category; and determine the target real-time resource characteristics and target category as the analysis results of the selection catalog.

[0108] For example, for the ath undetermined category, the target possibility of the ath undetermined category in the third target possibility distribution is Ua; according to the target possibility of each undetermined category in the third target possibility distribution, the undetermined category with the largest target possibility can be selected as the target category.

[0109] It should be noted that since the recognition and analysis capabilities of the general artificial intelligence thread may interfere with each other, errors caused by recognition capabilities may exist in the analysis catalog. The artificial intelligence analysis thread decouples the recognition and analysis capabilities during the training phase, that is, the fourth possibility distribution is the regression analysis possibility that is not interfered by the recognition capability; and the third possibility distribution and the fourth possibility distribution are used for joint regression analysis to obtain the third target possibility distribution, which helps to reduce the error problem caused by the recognition capability of the general artificial intelligence thread in the analysis catalog. The probability of each undetermined category in the obtained third target possibility distribution is also more accurate, thereby improving the accuracy of the analysis results.

[0110] In a possible implementation, the electronic device may determine the third target possibility distribution based on the similarity between the third possibility distribution and the fourth possibility distribution. For example, in step 205, based on the third possibility distribution and the fourth possibility distribution, the implementation method of generating the target type corresponding to the target real-time resource characteristics in the natural resource data set to be processed may include the following steps 2051 to 2054: Step 2051, based on the sharing coefficient between the third possibility distribution and the fourth possibility distribution, determine the first analysis difference of the fourth possibility distribution; Step 2052, based on the comparison result between the fourth possibility distribution and the first analysis difference, determine the second debugging possibility for the third possibility distribution; Step 2053, debug the third possibility distribution based on the second debugging possibility to obtain the third target possibility distribution; Step 2054, based on the third target possibility distribution, obtain the target type.

[0111] It should be noted that, similar to the identification catalog, if the sharing coefficient between the third possibility distribution and the fourth possibility distribution is larger, the analysis difference of the fourth possibility distribution is also smaller, indicating that the analysis ability of the artificial intelligence analysis thread is better. By debugging the third possibility distribution through the second debugging possibility, the parts with large differences between the third possibility distribution and the fourth possibility distribution can be effectively removed. At the same time, the third possibility distribution and the fourth debugging possibility distribution can also be spliced ​​with the higher possibility of the pending categories. The parts with higher sharing coefficients in the two possibility distributions can be spliced ​​more. This enhances the ability of the third target possibility distribution to express the actual target categories, helps to more accurately classify the target real-time resource characteristics into corresponding categories, and improves the accuracy of the analysis results.

[0112] Therefore, by using the third possibility distribution and the fourth possibility distribution to analyze the target real-time resource characteristics, the analysis capability of the artificial intelligence analysis thread can be further utilized under the premise of the analysis capability of the general artificial intelligence thread to improve the accuracy of the analysis results. In addition, by subtracting the first analysis difference in the fourth possibility distribution to obtain the second debugging possibility, the part with a larger difference in the fourth possibility distribution can be removed to effectively correct the analysis capability of the artificial intelligence analysis thread. And using the second debugging possibility to debug the third possibility distribution makes the analysis capability of the general artificial intelligence thread and the analysis-specific artificial intelligence thread more effectively combined, further enhancing the accuracy of the regression analysis of the actual category in the third target possibility distribution, helping to accurately find the target category with the highest probability and improve the accuracy of the analysis results.

[0113] Step 206: The electronic device obtains analysis results of the natural resource data set that needs to be processed based on the target real-time resource characteristics and the target type of the target real-time resource characteristics.

[0114] For example, the analysis result is a natural resource data analysis result obtained through various calculations and processing, and the result can describe the various natural resource conditions in the area, such as the type, content and location of soil components.

[0115] The analysis results of the natural resource data set to be processed include target real-time resource characteristics and types of target real-time resource characteristics.

[0116] If the natural resource data set to be processed includes several real-time resource features in the target field, the identification catalog can be performed multiple times through the above steps 202-203 to obtain several target real-time resource features; and the target category of each target real-time resource feature can be obtained by performing multiple analysis catalogs through the above steps 204-205. The analysis result includes the several target real-time resource features and the target category of each target real-time resource feature.

[0117] In some possible embodiments, for the identification catalog, that is, the identification of real-time resource characteristics of the natural resource data set that needs to be processed, an artificial intelligence analysis thread can also be used for identification; and under the premise of the first possibility distribution and the second possibility distribution, the fifth possibility distribution of the artificial intelligence analysis thread is further combined to determine the final identification result.

[0118] For example, the method further includes the following steps A1: Step A1, based on the natural resource data set to be processed, the artificial intelligence analysis thread is used to perform real-time resource feature recognition on the natural resource data set to be processed, and a fifth possibility distribution corresponding to the first natural resource description content is obtained; for example, the electronic device can input the natural resource data set to be processed into the general artificial intelligence thread, the recognition artificial intelligence thread, and the artificial intelligence analysis thread respectively, and the three artificial intelligence threads respectively execute the recognition directory to respectively obtain the first possibility distribution of the general artificial intelligence thread corresponding to the recognition directory, the second possibility distribution of the recognition artificial intelligence thread corresponding to the recognition directory, and the fifth possibility distribution of the artificial intelligence analysis thread corresponding to the recognition directory. Among them, the fifth possibility distribution may include the possibility of covering each pending real-time resource feature in the natural resource data set to be processed.

[0119] In a possible implementation, the fifth possibility distribution can also be obtained by constraining the possibility distribution directly output by the artificial intelligence analysis thread. For example, in step 202 and step A1, the first possibility distribution, the second possibility distribution and the fifth possibility distribution are obtained by the following steps: based on the natural resource data set to be processed, respectively using the general artificial intelligence thread, the recognition artificial intelligence thread, and the artificial intelligence analysis thread to perform the recognition directory, and obtain the first original possibility distribution output by the general artificial intelligence thread, the second original possibility distribution output by the recognition artificial intelligence thread, and the fifth original possibility distribution output by the artificial intelligence analysis thread, the fifth original possibility distribution including the fifth original possibility of each pending real-time resource feature; based on the first original possibility, the second original possibility and the fifth original possibility of each pending real-time resource feature, from each pending real-time resource feature, determine the target pending layer whose first original possibility, the second original possibility and the fifth original possibility are not lower than the target specified value; respectively delete the possibilities of the non-target pending layer in the first original possibility distribution, the second original possibility distribution and the fifth original possibility distribution, and obtain the first possibility distribution, the second possibility distribution and the fifth possibility distribution.

[0120] It should be noted that the fifth original possibility distribution directly output by the artificial intelligence analysis thread can also be constrained by constraint requirements. The constraint requirements can refer to the constraint requirements in step 202, which will not be described here one by one.

[0121] Among them, for the determination method of each target pending layer, the electronic device uses constraint requirements to obtain the corresponding first current set based on the first original possibility distribution. Similarly, from each pending real-time resource feature of the fifth original possibility distribution, each pending real-time resource feature that is not less than the target specified value is screened out to obtain the fifth current set; then, the first current set, the second current set, and the fifth current set are subjected to intersection operations to obtain the pending real-time resource features commonly included in the three current sets, and the pending real-time resource features commonly included in the three current sets are determined as the target pending layer.

[0122] Among them, the fifth original possibility value of the non-target undetermined layer in the fifth original possibility distribution can be set to a positive infinite value to obtain the fifth possibility distribution.

[0123] If the fifth possibility distribution is obtained by using an artificial intelligence analysis thread, correspondingly, the above step 203 may include the following steps: obtaining the target real-time resource characteristics based on the first possibility distribution, the second possibility distribution and the fifth possibility distribution.

[0124] For example, the electronic device may determine a second target possibility distribution by combining the first possibility distribution, the second possibility distribution, and the fifth possibility distribution; and determine the target real-time resource feature in each pending real-time resource feature based on the second target possibility distribution.

[0125] In a possible implementation, the electronic device can splice the recognition capabilities of the recognition artificial intelligence thread and the artificial intelligence analysis thread based on the similarity between the first possibility distribution and the second possibility distribution and the fifth possibility distribution. For example, if the artificial intelligence analysis thread is also used to obtain the fifth possibility distribution corresponding to the recognition directory, then in step 203, the target real-time resource feature is generated based on the first possibility distribution, the second possibility distribution and the fifth possibility distribution, including the following steps: based on the sharing coefficient between the first possibility distribution and the second possibility distribution, determine the first recognition difference of the second possibility distribution; based on the sharing coefficient between the first possibility distribution and the fifth possibility distribution, determine the second recognition difference of the fifth possibility distribution; based on the comparison result between the second possibility distribution and the first recognition difference, determine the first debugging possibility of the first possibility distribution; debug the first possibility distribution based on the first debugging possibility; based on the comparison result between the debugged first possibility distribution and the second recognition difference, determine the second target possibility distribution; based on the second target possibility distribution, obtain the target real-time resource feature.

[0126] It should be noted that the artificial intelligence analysis thread may have recognition ability and analysis ability, and the analysis ability performs better than the recognition ability; the accuracy of the fifth possibility distribution may be affected by the analysis ability.

[0127] In another possible embodiment, for the analysis catalog, that is, analyzing the characteristics of the target real-time resources, the recognition artificial intelligence thread can also be used for analysis; and under the premise of the third possibility distribution and the fourth possibility distribution, the sixth possibility distribution is further combined to determine the final analysis result.

[0128] For example, the method also includes the following step A2: Step A2, based on the natural resource data set that needs to be processed and the target real-time resource characteristics, the recognition artificial intelligence thread is used to analyze the target real-time resource characteristics to obtain the sixth possibility distribution corresponding to the second natural resource description content.

[0129] For example, the electronic device can input the natural resource data set and the target real-time resource features to be processed into the general artificial intelligence thread, the recognition artificial intelligence thread, and the artificial intelligence analysis thread, respectively. The three artificial intelligence threads respectively execute the analysis catalog to obtain the third possibility distribution corresponding to the analysis catalog of the general artificial intelligence thread, the fourth possibility distribution corresponding to the analysis catalog of the artificial intelligence analysis thread, and the sixth possibility distribution corresponding to the analysis catalog of the recognition artificial intelligence thread. The sixth possibility distribution may include the possibility that the type of the target real-time resource feature is each to-be-determined type.

[0130] In a possible implementation, the sixth possibility distribution can also be obtained by constraining the possibility distribution directly output by the recognition artificial intelligence thread. The process may include: based on the natural resource data set to be processed and the target real-time resource characteristics, the recognition artificial intelligence thread is used to analyze the target real-time resource characteristics to obtain the sixth original possibility distribution, and the sixth original possibility distribution includes the sixth original possibility of each pending category; based on the third original possibility distribution, the fourth original possibility distribution and the sixth original possibility distribution, from each pending category, determine the target pending category whose third original possibility, fourth original possibility and sixth original possibility are not lower than the target specified value; respectively delete the possibilities of non-target pending categories in the third original possibility distribution, the fourth original possibility distribution and the sixth original possibility distribution to obtain the third possibility distribution, the fourth possibility distribution and the sixth possibility distribution.

[0131] It should be noted that the process of obtaining the third possibility distribution, the fourth possibility distribution, and the sixth possibility distribution for the original possibility distribution directly output by the three artificial intelligence threads is the same as the process of obtaining the first possibility distribution, the second possibility distribution, and the fifth possibility distribution in the above steps, and can also be implemented by adopting the above-mentioned corresponding constraint requirements, corresponding methods of deletion processing, etc., which will not be elaborated here.

[0132] If the sixth possibility distribution is obtained by using the recognition artificial intelligence thread, correspondingly, the above step 205 may include: obtaining the target type corresponding to the target real-time resource feature based on the third possibility distribution, the fourth possibility distribution and the sixth possibility distribution.

[0133] For example, the electronic device may determine a fourth target possibility distribution in combination with the third possibility distribution, the fourth possibility distribution, and the sixth possibility distribution; and determine the target type of the target real-time resource feature based on the fourth target possibility distribution.

[0134] In one possible implementation, the electronic device may combine the analysis capabilities of identifying artificial intelligence threads and artificial intelligence analysis threads based on the similarities between the third possibility distribution and the fourth possibility distribution and the sixth possibility distribution, respectively.

[0135] Among them, similarity can be understood as a sharing coefficient, which indicates the weight of the commonality in the possibility.

[0136] For example, if an identification artificial intelligence thread is also used to obtain the sixth possibility distribution corresponding to the analysis catalog, then in step 205, the target type corresponding to the target real-time resource feature is generated based on the third possibility distribution, the fourth possibility distribution and the sixth possibility distribution, including the following steps: based on the sharing coefficient between the third possibility distribution and the fourth possibility distribution, determining the first analysis difference of the fourth possibility distribution; based on the sharing coefficient between the third possibility distribution and the sixth possibility distribution, determining the second analysis difference of the sixth possibility distribution; based on the comparison result between the fourth possibility distribution and the first analysis difference, determining the second debugging possibility of the third possibility distribution; debugging the third possibility distribution based on the second debugging possibility; determining the fourth target possibility distribution based on the comparison result between the debugged third possibility distribution and the second analysis difference; and obtaining the target real-time resource feature based on the fourth target possibility distribution.

[0137] Since there may be errors caused by recognition capabilities in the analysis catalog of the general artificial intelligence thread, therefore, in the analysis catalog, the analysis capabilities of the general artificial intelligence thread and the artificial intelligence analysis thread are combined; further, by measuring the difference in analysis capabilities between the recognition artificial intelligence thread and the general artificial intelligence thread, the possibility of category words with large differences from the sixth possibility distribution in the third possibility distribution can be weakened, thereby further removing the analysis differences caused by recognition capabilities in the analysis catalog of the general artificial intelligence thread. This will improve the accuracy of the fourth target possibility distribution and the accuracy of the analysis results.

[0138] It should be noted that in the present application, the natural resource data set to be processed is input into three artificial intelligence threads, and each artificial intelligence thread is combined to generate an analysis result by compressing and deriving the natural resource data set to be processed.

[0139] Among them, the specific embodiment of comparative derivation is that in the recognition process, the difference in recognition ability between the general artificial intelligence thread and the recognition artificial intelligence thread and the artificial intelligence analysis thread is used to compare the respective influences of the artificial intelligence analysis thread and the recognition artificial intelligence thread, so as to remove the errors caused by the analysis ability in the general artificial intelligence thread during derivation, and splice the recognition ability of the general artificial intelligence thread and the recognition artificial intelligence thread. In the analysis process, the difference in analysis ability between the general artificial intelligence thread and the artificial intelligence analysis thread and the recognition artificial intelligence thread is used to compare the corresponding influences of the artificial intelligence analysis thread and the recognition artificial intelligence thread on the analysis ability, so as to remove the errors caused by the recognition ability in the general artificial intelligence thread during derivation, and splice the analysis ability of the general artificial intelligence thread and the artificial intelligence analysis thread, thereby using comparative derivation to achieve the effect of removing anomalies and improving accuracy.

[0140] In a possible implementation, the method may also include the following step B1: Step B1, obtaining a prompt natural resource data set for a selection catalog; wherein the prompt natural resource data set includes an identification prompt feature and an analysis prompt feature, the identification prompt feature indicates the identification of real-time resource features of a target area in the natural resource data set that needs to be processed, and the analysis prompt feature indicates the analysis of the identified real-time resource features.

[0141] The selection catalog includes an identification catalog and an analysis catalog, and the prompt natural resource data set includes prompt features corresponding to the identification catalog and the analysis catalog, respectively. The identification catalog identifies the real-time resource features of the target field in the natural resource data set to be processed, and the analysis catalog analyzes the identified real-time resource features.

[0142] In a possible implementation, the prompt natural resource data set may include an identification prompt feature for identifying a catalog and an analysis prompt feature for analyzing a catalog. In different catalogs, the prompt features corresponding to the catalogs may be used respectively.

[0143] Among them, the first possibility distribution is obtained by using a general artificial intelligence thread to perform real-time resource feature recognition on the natural resource data set that needs to be processed based on the natural resource data set that needs to be processed and the identification prompt features; the second possibility distribution is obtained by using a recognition artificial intelligence thread to perform real-time resource feature recognition on the natural resource data set that needs to be processed based on the natural resource data set that needs to be processed and the identification prompt features; for example, step 202 can be replaced by step T1: step T1, the electronic device uses a general artificial intelligence thread to perform real-time resource feature recognition on the natural resource data set that needs to be processed based on the natural resource data set that needs to be processed and the identification prompt features, and obtains a first possibility distribution corresponding to the first natural resource description content, and uses a recognition artificial intelligence thread to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a second possibility distribution corresponding to the first natural resource description content.

[0144] Among them, the third possibility distribution is obtained by analyzing the target real-time resource characteristics based on the natural resource data set that needs to be processed, the target real-time resource characteristics and the analysis prompt characteristics, using a general artificial intelligence thread to analyze the target real-time resource characteristics; the fourth possibility distribution is obtained by analyzing the target real-time resource characteristics based on the natural resource data set that needs to be processed, the target real-time resource characteristics and the analysis prompt characteristics, using an artificial intelligence analysis thread. For example, step 204 can be replaced by step T2: step T2, the electronic device analyzes the target real-time resource characteristics based on the natural resource data set that needs to be processed, the analysis prompt characteristics and the target real-time resource characteristics using the general artificial intelligence thread to obtain a third possibility distribution corresponding to the second natural resource description content, and analyzes the target real-time resource characteristics using an artificial intelligence analysis thread to obtain a fourth possibility distribution corresponding to the second natural resource description content.

[0145] In another possible implementation, no distinction is made between the identification catalog and the analysis catalog, and the prompt natural resource data set is uniformly used. That is, step 202 can be replaced by: the electronic device uses a general artificial intelligence thread to perform real-time resource feature recognition on the natural resource data set to be processed based on the natural resource data set to be processed and the prompt natural resource data set, and obtains a first possibility distribution corresponding to the first natural resource description content, and uses an identification artificial intelligence thread to perform real-time resource feature recognition on the natural resource data set to be processed, and obtains a second possibility distribution corresponding to the first natural resource description content; the same is true for the analysis catalog, which will not be listed here one by one.

[0146] In one possible implementation, a prompt natural resource data set may also be used for the training process.

[0147] For example, different directories are distinguished, and different directories use prompt features corresponding to the directories. For example, if the recognition artificial intelligence thread to be trained executes the recognition directory, the first example and the recognition prompt feature can be input; if the analysis directory is executed, the first regression analysis real-time resource feature and the analysis prompt feature of the first example can be input. It should be noted that the method of prompting the natural resource data set used in the reasoning stage is also consistent with the training stage. For example, both stages use the directory prompt feature corresponding to the currently executed directory. For example, the above steps T1 and T2.

[0148] Alternatively, both stages do not distinguish between different catalogs, that is, different catalogs all use the complete prompt natural resource data set. For example, if in the training stage, the catalogs are not distinguished, but the complete prompt natural resource data set is uniformly used for the identification catalog and the analysis catalog, then in the inference stage, the complete prompt natural resource data set can also be uniformly used for each catalog.

[0149] It should be noted that by distinguishing different catalogs and using different prompt features, the prompt natural resource data set can be decoupled, the identification prompt features can be used to identify the catalog, and the analysis prompt features can be used to analyze the catalog. Based on this, on the premise of decoupling the recognition ability and the analysis ability, the current catalog can be further prevented from being interfered with by other different catalogs, thereby improving the accuracy.

[0150] In another possible embodiment, the natural resource data set to be processed may include several current monomers, and multiple catalogs need to be executed to obtain a complete analysis result of the natural resource data set to be processed. The result of each identification catalog or analysis catalog can be generated by autoregression.

[0151] For example, the natural resource data set to be processed includes several target real-time resource characteristics, and the analysis result includes several target real-time resource characteristics and target types of the target real-time resource characteristics; the analysis result is generated by executing x directories, each directory is an identification directory or an analysis directory, and the natural resource data set generated in each directory is the target real-time resource characteristics or the target type; wherein x is an integer greater than or equal to 2; wherein the execution process of the a-th directory includes the following steps C1-C3: step C1, based on the generation of the natural resource data set of the a-1-th directory, determine the type of the natural resource data set of the natural resource data set to be generated, the natural resource data set type is the identification type or the analysis type, the natural resource corresponding to the 1st directory The source data set type is an identification type, wherein a is an integer greater than or equal to 2; step C2, based on the natural resource data set to be processed and the generated natural resource data set of the previous a-1 directories, a general artificial intelligence thread is used to obtain a general possibility distribution, an identification artificial intelligence thread is used to obtain an identification possibility distribution, and an artificial intelligence analysis thread is used to obtain an analysis possibility distribution; wherein the general possibility distribution is the first possibility distribution or the third possibility distribution, the identification possibility distribution is the second possibility distribution or the sixth possibility distribution, and the analysis possibility distribution is the fourth possibility distribution or the fifth possibility distribution; step C3, based on the general possibility distribution, the identification possibility distribution and the analysis possibility distribution, the generated natural resource data set of the a-th directory is generated.

[0152] The general possibility distribution is the first possibility distribution, the identification possibility distribution is the second possibility distribution, and the analysis possibility distribution is the fifth possibility distribution. Alternatively, the general possibility distribution is the third possibility distribution, the analysis possibility distribution is the fourth possibility distribution, and the identification possibility distribution is the sixth possibility distribution.

[0153] In a possible implementation, for the first catalog, the first catalog is determined to be an identification catalog based on a preconfigured judgment rule; based on the natural resource data set to be processed, the first catalog is executed by a general artificial intelligence thread and an identification artificial intelligence thread respectively, to obtain a first possibility distribution and a second possibility distribution corresponding to the first catalog; based on the first possibility distribution and the second possibility distribution corresponding to the first catalog, the target real-time resource characteristics are determined, and the determined target real-time resource characteristics are determined as the result of the first catalog. Alternatively, the first catalog can also be executed by an artificial intelligence analysis thread to obtain a fifth possibility distribution; then, based on the first possibility distribution, the second possibility distribution and the fifth possibility distribution, the target real-time resource characteristics of the first catalog are obtained.

[0154] In a possible implementation, the judgment rule is used to judge the type of the natural resource data set of the natural resource data set to be generated, that is, the type of the directory to be executed; the identification type corresponds to the identification directory, and the analysis type corresponds to the analysis directory. For example, the results of each sub-directory generated can be represented in a structured form, and the judgment rule can be based on a predefined structure for judgment.

[0155] The pre-definition may be preset by personnel in the relevant technical field.

[0156] The training process of each artificial intelligence thread is introduced below.

[0157] In this application, the selection catalog can be modeled into a recognition and analysis process, corresponding to the recognition catalog and the analysis catalog. In the training phase, the recognition ability and analysis ability of the artificial intelligence thread can be decoupled.

[0158] In a possible implementation embodiment, the recognition artificial intelligence thread is trained through the following steps S11-S14: Step S11, obtain a first training set, the first training set includes a plurality of first example natural resource data sets with example directories, and the example directory of each first example includes the current directory of the first example; Step S12, based on each first example, use the recognition artificial intelligence thread to be trained to perform real-time resource feature recognition on each first example to obtain a first regression analysis real-time resource feature of each first example; Step S13, based on each first example and the first regression analysis real-time resource feature, use the recognition artificial intelligence thread to be trained to analyze the first regression analysis real-time resource feature to obtain a first regression analysis type of the first regression analysis real-time resource feature, and based on the first regression analysis real-time resource feature and the first regression analysis type of each first example, obtain the analysis result of each first example; Step S14, determine a first training evaluation index based on the sharing coefficient between the first regression analysis real-time resource feature of each first example and the current directory, and optimize the thread coefficient of the recognition artificial intelligence thread to be trained based on the first training evaluation index.

[0159] For example, each first example may be an example natural resource data set, for example, each first example may be a conversation natural resource data set between a doctor and a patient.

[0160] It should be noted that each first example may include one or several real-time resource features, and may also use an autoregressive approach to generate several first regression analysis real-time resource features and the first regression analysis type of each first regression analysis real-time resource feature.

[0161] The difference from the above steps C1-C3 is that in the training phase, the recognition artificial intelligence thread to be trained can be used to independently execute the recognition catalog and the analysis catalog. That is, the recognition artificial intelligence thread to be trained can be used to perform regression analysis on the real-time resource characteristics of the first example to obtain a first regression analysis possibility distribution; based on the first regression analysis possibility distribution, the first regression analysis real-time resource characteristics are generated; then, using the first example and the first regression analysis real-time resource characteristics, the first regression analysis real-time resource characteristics are analyzed by the recognition artificial intelligence thread to be trained to obtain a second regression analysis possibility distribution, and the first regression analysis category is generated based on the second regression analysis possibility distribution. The analysis result is obtained based on the first regression analysis real-time resource characteristics and the first regression analysis category.

[0162] In a possible implementation embodiment, the artificial intelligence analysis thread is trained by the following steps S21-S24: step S21, obtaining a second training set, the second training set including a plurality of second examples having an example directory, the example directory of each second example including a category directory of each current directory of the second example; step S22, based on each second example, using the artificial intelligence analysis thread to be trained to perform real-time resource feature recognition on each second example, and obtaining a second regression analysis real-time resource feature of each second example; step S23, based on each second example and the second regression analysis real-time resource feature, using the artificial intelligence analysis thread to be trained to analyze the second regression analysis real-time resource feature, and obtaining a second regression analysis category of the second regression analysis real-time resource feature, and obtaining an analysis result of each second example based on the second regression analysis real-time resource feature and the second regression analysis category of each second example; step S24, determining a second training evaluation index based on the sharing coefficient between the second regression analysis category and the category directory of each second example, and optimizing the thread coefficient of the artificial intelligence analysis thread to be trained based on the second training evaluation index.

[0163] For example, each second example may be a sample natural resource data set. In which, the real-time resource features of the second example and the type of each real-time resource feature may be generated by an autoregressive method. Similar to the recognition artificial intelligence thread to be trained, the artificial intelligence analysis thread to be trained may be used to independently execute the recognition directory and the analysis directory.

[0164] For example, the artificial intelligence analysis thread to be trained can be used to perform regression analysis on the real-time resource characteristics of the second example to obtain a third regression analysis possibility distribution, and the third regression analysis possibility distribution can be used to generate a second regression analysis real-time resource characteristic; then, using the second example and the second regression analysis real-time resource characteristics, the type of the second regression analysis real-time resource characteristic is analyzed by the artificial intelligence analysis thread to be trained to obtain a fourth regression analysis possibility distribution, and the fourth regression analysis possibility distribution is used to generate the second regression analysis type.

[0165] In a possible implementation embodiment, the general artificial intelligence thread is trained by the following steps S31-S34: step S31, obtaining a third training set, the third training set including a plurality of third examples with an example directory, the example directory of each third example including a current directory of the third example and a category directory of each current directory; step S32, based on each third example, using the general artificial intelligence thread to be trained to perform real-time resource feature recognition on each third example, and obtain a third regression analysis real-time resource feature of each third example; step S33, based on each third example and the third regression analysis real-time resource feature, using the general artificial intelligence thread to be trained to analyze the third regression analysis real-time resource feature, and obtain a third regression analysis category of the third regression analysis real-time resource feature, and based on the third regression analysis real-time resource feature and the third regression analysis category of each third example, obtain an analysis result of each third example; step S34, determining a third training evaluation index based on the third analysis result of each third example and the sharing coefficient between the example directory, and optimizing the thread coefficient of the general artificial intelligence thread to be trained based on the third training evaluation index.

[0166] When training a general artificial intelligence thread, the general artificial intelligence thread can be used to independently generate a third regression analysis real-time resource feature and a third regression analysis category. In addition, a first sharing coefficient between the third regression analysis real-time resource feature of each third example and the current directory, a second sharing coefficient between the third regression analysis category and the category directory can be determined, and an evaluation index between the third analysis result of each third example and the example directory can be determined. The general artificial intelligence thread to be trained is trained using the evaluation index to obtain the general artificial intelligence thread.

[0167] The artificial intelligence-based natural resource data analysis method of the present application uses a general artificial intelligence thread to identify real-time resource features based on a natural resource data set that needs to be processed to obtain a first possibility distribution, uses a recognition artificial intelligence thread to identify real-time resource features to obtain a second possibility distribution, and obtains the target real-time resource features in the natural resource data set that needs to be processed based on the first possibility distribution and the second possibility distribution; so that during identification, the recognition ability of the recognition artificial intelligence thread can be used to reduce the errors caused by the analysis ability in the recognition catalog of the general artificial intelligence thread, and improve the accuracy of identification. In addition, based on the natural resource data set that needs to be processed and the target real-time resource features, the general artificial intelligence thread is used to analyze the target real-time resource features to obtain a third possibility distribution, the artificial intelligence analysis thread is used to analyze to obtain a fourth possibility distribution, and the target category is obtained based on the third possibility distribution and the fourth possibility distribution; so that during analysis, the analysis ability of the artificial intelligence analysis thread can be further used to reduce the errors caused by the recognition ability in the analysis catalog of the general artificial intelligence thread, and improve the accuracy of analysis. Thus, natural resource data can be obtained more accurately.

[0168] On the basis of the above, a natural resource data analysis device based on artificial intelligence is provided, the device comprising:

[0169] A data acquisition module is used to obtain a natural resource data set that needs to be processed;

[0170] A second possibility obtaining module is used to perform real-time resource feature recognition on the natural resource data set that needs to be processed using a general artificial intelligence thread based on the natural resource data set that needs to be processed, to obtain a first possibility distribution corresponding to a first natural resource description content, and to perform real-time resource feature recognition on the natural resource data set that needs to be processed using an identification artificial intelligence thread, to obtain a second possibility distribution corresponding to the first natural resource description content, wherein the first natural resource description content includes a plurality of pending real-time resource features;

[0171] A feature acquisition module, configured to acquire target real-time resource features in the natural resource data set to be processed based on the first possibility distribution and the second possibility distribution;

[0172] A fourth possibility obtaining module is used to analyze the target real-time resource characteristics using the general artificial intelligence thread based on the natural resource data set to be processed and the target real-time resource characteristics to obtain a third possibility distribution corresponding to the second natural resource description content, and to analyze the target real-time resource characteristics using the artificial intelligence analysis thread to obtain a fourth possibility distribution corresponding to the second natural resource description content, wherein the second natural resource description content includes a plurality of to-be-determined categories;

[0173] A category obtaining module, used for obtaining a target category of the target real-time resource feature based on the third possibility distribution and the fourth possibility distribution;

[0174] The result analysis module is used to obtain the analysis result of the natural resource data set that needs to be processed based on the target real-time resource characteristics and the target type of the target real-time resource characteristics.

[0175] On the basis of the above, a natural resource data analysis system based on artificial intelligence is shown, which includes a processor and a memory that communicate with each other, and the processor is used to read and execute a computer program from the memory to implement the above method.

[0176] Based on the above, a computer-readable storage medium is also provided, on which a computer program stored implements the above method when running.

[0177] In summary, based on the above scheme, based on the natural resource data set to be processed, a general artificial intelligence thread is used to identify the real-time resource features to obtain a first possibility distribution, and a recognition artificial intelligence thread is used to identify the real-time resource features to obtain a second possibility distribution. Based on the first possibility distribution and the second possibility distribution, the target real-time resource features in the natural resource data set to be processed are obtained; so that during identification, the recognition ability of the recognition artificial intelligence thread can be used to reduce the errors caused by the analysis ability in the recognition catalog of the general artificial intelligence thread, and improve the accuracy of identification. In addition, based on the natural resource data set to be processed and the target real-time resource features, a general artificial intelligence thread is used to analyze the target real-time resource features to obtain a third possibility distribution, and an artificial intelligence analysis thread is used to analyze to obtain a fourth possibility distribution, and the target category is obtained based on the third possibility distribution and the fourth possibility distribution; so that during analysis, the analysis ability of the artificial intelligence analysis thread can be further used to reduce the errors caused by the recognition ability in the analysis catalog of the general artificial intelligence thread, and improve the accuracy of analysis. Thus, natural resource data can be obtained more accurately.

[0178] It should be understood that the system and its modules shown above can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or a dedicated design hardware. Those skilled in the art will understand that the above methods and systems can be implemented using computer executable instructions and / or included in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The system and its modules of the present application can not only be implemented by hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., but can also be implemented by software such as executed by various types of processors, and can also be implemented by a combination of the above hardware circuits and software (e.g., firmware).

[0179] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or any other beneficial effects that may be obtained.

Claims

1. A natural resource data analysis method based on artificial intelligence, characterized in that: The method comprises: Obtaining the natural resource data sets that require processing; Based on the natural resource data set that needs to be processed, a general artificial intelligence thread is used to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a first possibility distribution corresponding to a first natural resource description content, and a recognition artificial intelligence thread is used to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a second possibility distribution corresponding to the first natural resource description content, wherein the first natural resource description content includes a plurality of to-be-determined real-time resource features; Based on the first possibility distribution and the second possibility distribution, obtaining target real-time resource characteristics in the natural resource data set that needs to be processed; Based on the natural resource data set to be processed and the target real-time resource characteristics, the general artificial intelligence thread is used to analyze the target real-time resource characteristics to obtain a third possibility distribution corresponding to the second natural resource description content, and the artificial intelligence analysis thread is used to analyze the target real-time resource characteristics to obtain a fourth possibility distribution corresponding to the second natural resource description content, wherein the second natural resource description content includes a plurality of to-be-determined categories; Based on the third possibility distribution and the fourth possibility distribution, obtaining a target category of the target real-time resource feature; Based on the target real-time resource characteristics and the target type of the target real-time resource characteristics, obtaining analysis results of the natural resource data set that needs to be processed; The recognition artificial intelligence thread is obtained by training based on the sharing coefficient between the first regression analysis real-time resource feature of each first example and the current directory; the artificial intelligence analysis thread is obtained by training based on the sharing coefficient between the second regression analysis category of each second example and the category directory; the general artificial intelligence thread is obtained by training based on the sharing coefficient between the third analysis result of each third example and the example directory, the third analysis result includes the third regression analysis real-time resource feature and the third category, and the example directory includes the current directory and the category directory of the third example; Wherein, the method further comprises: Based on the natural resource data set that needs to be processed, an artificial intelligence analysis thread is used to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a fifth possibility distribution corresponding to the first natural resource description content; Based on the natural resource data set to be processed and the target real-time resource characteristics, the recognition artificial intelligence thread is used to analyze the target real-time resource characteristics to obtain a sixth possibility distribution corresponding to the second natural resource description content; Wherein, obtaining the target real-time resource characteristics in the natural resource data set to be processed based on the first possibility distribution and the second possibility distribution includes: Based on the first possibility distribution, the second possibility distribution and the fifth possibility distribution, obtaining the target real-time resource characteristics; Wherein, obtaining the target type of the target real-time resource feature based on the third possibility distribution and the fourth possibility distribution includes: Based on the third possibility distribution, the fourth possibility distribution and the sixth possibility distribution, obtaining a target type corresponding to the target real-time resource feature; Wherein, obtaining the target real-time resource characteristics in the natural resource data set to be processed based on the first possibility distribution and the second possibility distribution includes: determining a first identified difference of the second likelihood distribution based on a sharing coefficient between the first likelihood distribution and the second likelihood distribution; determining a first debugging possibility for the first possibility distribution based on a comparison result between the second possibility distribution and the first identified difference; Debugging the first possibility distribution based on the first debugging possibility to obtain a first target possibility distribution; Based on the first target possibility distribution, obtaining the target real-time resource characteristics; The identification artificial intelligence thread is trained in the following way: Obtaining a first training set, the first training set comprising a plurality of first example natural resource data sets having example directories, wherein the example directory of each first example comprises a current directory of the first example; Based on each first example, using the recognition artificial intelligence thread to be trained to perform real-time resource feature recognition on each first example, to obtain a first regression analysis real-time resource feature of each first example; Based on each of the first examples and the first regression analysis real-time resource feature, the first regression analysis real-time resource feature is analyzed by using a recognition artificial intelligence thread to be trained to obtain a first regression analysis type of the first regression analysis real-time resource feature, and based on the first regression analysis real-time resource feature and the first regression analysis type of each of the first examples, an analysis result of each of the first examples is obtained; Determine a first training evaluation index based on a first regression analysis of a real-time resource feature of each first example and a sharing coefficient between the current directory, and optimize a thread coefficient of a recognition artificial intelligence thread to be trained based on the first training evaluation index; The artificial intelligence analysis thread is trained in the following way: Obtaining a second training set, wherein the second training set includes a plurality of second examples having example directories, wherein the example directory of each second example includes a category directory of each current directory of the second example; Based on each second example, using the artificial intelligence analysis thread to be trained to perform real-time resource feature recognition on each second example, to obtain a second regression analysis real-time resource feature of each second example; Based on each of the second examples and the second regression analysis real-time resource characteristics, the second regression analysis real-time resource characteristics are analyzed using the artificial intelligence analysis thread to be trained to obtain a second regression analysis type of the second regression analysis real-time resource characteristics, and based on the second regression analysis real-time resource characteristics and the second regression analysis type of each of the second examples, an analysis result of each of the second examples is obtained; Determine a second training evaluation index based on a sharing coefficient between the second regression analysis category and the category catalog of each second example, and optimize a thread coefficient of an artificial intelligence analysis thread to be trained based on the second training evaluation index; The general artificial intelligence thread is trained in the following way: Obtaining a third training set, the third training set comprising a plurality of third examples having example directories, wherein the example directory of each third example comprises a current directory of the third example and a category directory of each current directory; Based on each third example, using the general artificial intelligence thread to be trained to perform real-time resource feature recognition on each third example, to obtain a third regression analysis real-time resource feature of each third example; Based on each of the third examples and the third regression analysis real-time resource feature, the third regression analysis real-time resource feature is analyzed by using a general artificial intelligence thread to be trained to obtain a third regression analysis type of the third regression analysis real-time resource feature, and based on the third regression analysis real-time resource feature and the third regression analysis type of each of the third examples, an analysis result of each of the third examples is obtained; Determine a third training evaluation index based on the third analysis result of each third example and the sharing coefficient between the example directories, and optimize the thread coefficient of the general artificial intelligence thread to be trained based on the third training evaluation index; The obtaining of the target real-time resource characteristics based on the first possibility distribution, the second possibility distribution and the fifth possibility distribution includes: determining a first identified difference of the second likelihood distribution based on a sharing coefficient between the first likelihood distribution and the second likelihood distribution; determining a second identified difference of the fifth likelihood distribution based on a sharing coefficient between the first likelihood distribution and a fifth likelihood distribution; determining a first debugging possibility for the first possibility distribution based on a comparison result between the second possibility distribution and the first identified difference; debugging the first possibility distribution based on the first debugging possibility; Determining a second target possibility distribution based on a comparison result between the debugged first possibility distribution and the second identification difference; Based on the second target possibility distribution, the target real-time resource characteristics are obtained.

2. The method according to claim 1, characterized in that The method further comprises: Get tips for selecting catalogs of natural resource data sets; Among them, the prompt natural resource data set includes an identification prompt feature and an analysis prompt feature. The identification prompt feature indicates the identification of real-time resource features of the target field in the natural resource data set that needs to be processed, and the analysis prompt feature indicates the analysis of the identified real-time resource features.

3. The method according to claim 2, characterized in that The first possibility distribution is obtained by performing real-time resource feature recognition on the natural resource data set to be processed using a general artificial intelligence thread based on the natural resource data set to be processed and the recognition prompt feature; The second possibility distribution is obtained by performing real-time resource feature recognition on the natural resource data set to be processed based on the natural resource data set to be processed and the recognition prompt feature by using the recognition artificial intelligence thread; Among them, the third possibility distribution is obtained by analyzing the target real-time resource characteristics based on the natural resource data set that needs to be processed, the target real-time resource characteristics and the analysis prompt characteristics, using a general artificial intelligence thread to analyze the target real-time resource characteristics; the fourth possibility distribution is obtained by analyzing the target real-time resource characteristics based on the natural resource data set that needs to be processed, the target real-time resource characteristics and the analysis prompt characteristics, using an artificial intelligence analysis thread.

4. The method according to claim 1, characterized in that: The first possibility distribution and the second possibility distribution are obtained by: Based on the natural resource data set that needs to be processed, a general artificial intelligence thread is used to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a first original possibility distribution corresponding to the first natural resource description content, and a recognition artificial intelligence thread is used to perform real-time resource feature recognition on the natural resource data set that needs to be processed to obtain a second original possibility distribution corresponding to the first natural resource description content; The first original possibility distribution includes the first original possibility of each pending real-time resource feature, and the second original possibility includes the second original possibility of each pending real-time resource feature; Based on the first original possibility distribution and the second original possibility distribution, determining a target undetermined layer from each of the undetermined real-time resource characteristics, in which both the first original possibility and the second original possibility are not lower than a target specified value; The possibilities corresponding to the non-target undetermined layers in the first original possibility distribution and the second original possibility distribution are deleted respectively to obtain the first possibility distribution and the second possibility distribution.

5. The method according to claim 1, characterized in that: The natural resource data set to be processed includes a number of target real-time resource features, and the analysis result includes a number of target real-time resource features and target types of the target real-time resource features; the analysis result is generated by executing x directories, each directory is an identification directory or an analysis directory, and the natural resource data set generated by each directory is the target real-time resource feature or the target type; wherein the execution process of the ath directory includes: Determine the type of natural resource data set to be generated based on the generated natural resource data set of the a-1th directory, wherein the type of natural resource data set is an identification type or an analysis type, and the type of natural resource data set corresponding to the first directory is an identification type; Based on the natural resource data set to be processed and the natural resource data set generated by the previous a-1 directories, a general artificial intelligence thread is used to obtain a general possibility distribution, an identification artificial intelligence thread is used to obtain a recognition possibility distribution, and an artificial intelligence analysis thread is used to obtain an analysis possibility distribution; Wherein, the general possibility distribution is the first possibility distribution or the third possibility distribution, the identification possibility distribution is the second possibility distribution or the sixth possibility distribution, and the analysis possibility distribution is the fourth possibility distribution or the fifth possibility distribution; Based on the general possibility distribution, the identification possibility distribution and the analysis possibility distribution, a natural resource data set of the a-th catalog is generated.

6. A natural resource data analysis system based on artificial intelligence, characterized in that: The invention comprises a processor and a memory communicating with each other, wherein the processor is used to read a computer program from the memory and execute the computer program to implement the method according to any one of claims 1 to 5.

Images