Intelligent detection method and system for soil pollutants

Abstract

The invention provides a soil pollutant intelligent detection method and system. The method comprises the following steps: removing soil matter creation conditions in example soil acquisition component data; obtaining soil component attribute data of the influencing soil matters according to the types of the influencing soil matters, and obtaining the influencing soil matters according to the sample soil collection component data, the multiple soil pollutant information sets and the soil component attribute data of the influencing soil matters through a deep learning-based influencing soil matter pollutant detection network. According to the method, regression analysis information about the range of pollutant data in sample soil acquisition component data is obtained, and the soil matter pollutant detection network is optimized according to the difference between the regression analysis information and annotation information, so that the trained soil matter pollutant detection network can be used for detecting soil matter pollutants in the sample soil acquisition component data. The method has the capability of detecting the matters influencing the soil based on the multi-dimensional information, and the pollutant data can be accurately detected by the pollutant detection network influencing the soil matters.

Description

Technical Field

[0001] The present application relates to the technical field of pollutant detection, and in particular to a method and system for intelligent detection of soil pollutants. Background Art

[0002] The presence of pollutants in the soil has a huge impact on plant growth or human life. Pollutants may affect plant growth or affect human life safety. Therefore, it is necessary to check the pollutants in the soil so that the pollutants can be accurately treated. At present, when checking pollutants, sample soil is used for inspection. This not only has a small inspection range, but also reduces accuracy. Therefore, a technical solution is urgently needed to improve the above technical problems. Summary of the invention

[0003] In order to improve the technical problems existing in the related technologies, the present application provides a method and system for intelligent detection of soil pollutants.

[0004] In a first aspect, a soil pollutant intelligent detection method is provided, the method comprising: Obtaining soil collection component data, wherein the soil collection component data includes matters affecting the soil; Obtaining a soil pollutant information set of the soil collection component data; Classifying the soil pollutant information set to obtain a plurality of soil pollutant information sets, wherein the plurality of soil pollutant information sets include a soil-affecting event category and at least one non-soil-affecting event category; According to the types of soil-affecting matters in the plurality of soil pollutant information sets, obtaining soil component attribute data of the soil-affecting matters; Through a deep learning-based soil matter-affecting pollutant detection network, the range of pollutant data in the soil collection component data is detected according to the soil collection component data, the multiple soil pollutant information sets and the soil component attribute data affecting the soil matters.

[0005] In the present application, the method further comprises: Extracting a data anomaly range from the soil collected component data according to a range where pollutant data in the soil collected component data are located; An abnormal data analysis is performed on the data abnormal range to obtain an abnormal data analysis result.

[0006] In the present application, the method further comprises: Extracting a data anomaly range from the soil collected component data according to a range where pollutant data in the soil collected component data are located; An example of configuring an abnormal data analysis network according to the extracted data abnormality range; The abnormal data analysis network is trained according to the configuration example.

[0007] In the present application, the soil pollutant information set for obtaining the soil collection component data includes: Obtaining a plurality of initial soil pollutant information sets corresponding to the soil collection component data; The multiple initial soil pollutant information sets are spliced ​​together to obtain the soil pollutant information set of the soil collection component data.

[0008] In the present application, the multiple initial soil contaminant information sets are spliced ​​to obtain the soil contaminant information set of the soil collection component data, including: Depolarizing the pollution coefficients corresponding to the same soil element in the plurality of initial soil pollutant information sets to obtain a depolarization result for each soil element; According to the depolarization processing result of each soil element, a soil pollutant information set of the soil collection component data is obtained.

[0009] In the present application, the soil pollutant information set is classified to obtain a plurality of soil pollutant information sets, including: According to the soil pollutant information set, a pollution coefficient corresponding to each soil element in the soil collection component data is obtained; Determining a plurality of pollution coefficient intervals according to each of the pollution coefficients; Dividing each of the pollution coefficients into any one of the multiple pollution coefficient intervals; According to the pollution coefficient interval to which each soil element is divided, the soil pollutant information set is divided into multiple soil pollutant information sets, wherein the pollution coefficient in the type of soil-affecting matters is smaller than the pollution coefficient in the type of non-soil-affecting matters.

[0010] In the present application, the step of obtaining soil component attribute data affecting soil matters according to the types of soil matters affecting soil matters in the plurality of soil pollutant information sets includes: Obtaining the pollution coefficient corresponding to each soil element in the soil-affecting event category in the plurality of soil pollutant information sets; The soil component attribute data affecting the soil matters are determined based on the minimum pollution coefficient and the maximum pollution coefficient among the pollution coefficients corresponding to the soil elements.

[0011] In the present application, the soil pollutant detection network based on deep learning detects the range of pollutant data in the soil collected component data according to the soil collected component data, the multiple soil pollutant information sets and the soil component attribute data affecting the soil matters, including: Inputting the soil collected component data, the plurality of soil pollutant information sets and the soil component attribute data affecting soil matters into a soil matter affecting pollutant detection network based on deep learning; Extracting soil pollutant characteristics of the soil collection component data and soil key characteristics of the plurality of soil pollutant information sets through the soil matter-affecting pollutant detection network, encoding the soil component attribute data affecting soil matters to obtain key characteristics affecting soil matters, and outputting the percentage of each soil element in the soil collection component data belonging to the soil element affecting soil matters according to the soil pollutant characteristics, the soil key characteristics and the key characteristics affecting soil matters; The range of the pollutant data in the soil collected component data is determined according to the percentage of each soil element in the soil collected component data belonging to the soil element affecting soil matters.

[0012] In the present application, the method further comprises: Obtaining a configuration example of a soil-affecting pollutant detection network, the configuration example including sample soil collection component data and annotation information about a range where pollutant data in the sample soil collection component data is located; Obtaining a soil pollutant information set of the sample soil collection component data; Classifying the soil pollutant information set of the example soil collection component data to obtain a plurality of soil pollutant information sets corresponding to the example soil collection component data, wherein the plurality of soil pollutant information sets include a soil-affecting event category and at least one non-soil-affecting event category; According to the types of soil-affecting matters in the plurality of soil pollutant information sets, obtaining soil component attribute data of the soil-affecting matters; Obtaining regression analysis information on the range of pollutant data in the example soil collection component data according to the example soil collection component data, a plurality of soil pollutant information sets corresponding to the example soil collection component data, and the soil component attribute data affecting the soil matter through a soil matter-affecting pollutant detection network; According to the difference between the regression analysis information and the annotation information, the soil pollutant detection network is optimized.

[0013] In a second aspect, a soil pollutant intelligent detection system 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.

[0014] An intelligent soil pollutant detection method and system provided in an embodiment of the present application obtains a configuration example of a pollutant detection network affecting soil matters, wherein the configuration example includes sample soil collection component data and annotation information about the range of pollutant data in the sample soil collection component data; obtains a soil pollutant information set of the sample soil collection component data, classifies the soil pollutant information set, and obtains multiple soil pollutant information sets, wherein the multiple soil pollutant information sets include types of soil affecting matters and at least one type of non-soil affecting matters. The embodiment of the present application classifies the soil pollutant information set, and can accurately distinguish the types of soil affecting matters and at least one type of non-soil affecting matters in the sample soil collection component data, and can extract the location information of the sample soil collection component data, and can accurately distinguish the sample soil affecting matters. The different levels of the example soil collection component data are collected, thereby creating conditions for the subsequent removal of soil-affecting items in the example soil collection component data; according to the types of soil-affecting items, the soil component attribute data of the soil-affecting items are obtained, and through the pollutant detection network affecting soil items based on deep learning, according to the example soil collection component data, multiple soil pollutant information sets and soil component attribute data of the soil-affecting items, the regression analysis information about the range of pollutant data in the example soil collection component data is obtained; according to the difference between the regression analysis information and the annotation information, the pollutant detection network affecting soil items is optimized, so that the trained pollutant detection network affecting soil items has the ability to detect soil-affecting items based on multi-dimensional information, thereby improving the ability of the pollutant detection network affecting soil items to accurately detect pollutant data. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] 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.

[0016] Figure 1 A flow chart of a soil contaminant intelligent detection method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] 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.

[0018] See also Figure 1 , shows a method for intelligent detection of soil pollutants, which may include the technical solutions described in the following steps 202-210.

[0019] Step 202, obtaining soil collection component data, wherein the soil collection component data includes matters affecting the soil.

[0020] Exemplarily, the soil collection component data is obtained by transporting the sample data collected on site to the laboratory for preprocessing. The accuracy of the data is not high, which optimizes the errors in the data. The collection information here is the sample soil collected and analyzed in the prior art. If the number of sample soils collected is not enough, the data of the area may not be prepared. Therefore, this application needs to analyze more sample soils first to reduce errors.

[0021] Among them, matters affecting soil can be understood as objects that affect soil analysis, referred to as interference objects.

[0022] Step 204, obtaining a soil pollutant information set of soil collection component data.

[0023] Exemplarily, the soil pollutant information set represents a data set consisting of all pollutant information in the soil, including metal pollutants, chemical pollutants, etc.

[0024] In the present application, the soil pollutant information set can also be obtained by splicing multiple continuous soil pollutant information sets.

[0025] Step 206, classify the soil pollutant information set to obtain multiple soil pollutant information sets, and the multiple soil pollutant information sets include a category of matters affecting soil and at least one category of matters not affecting soil.

[0026] Among them, classification refers to the process of classifying the corresponding soil elements in the soil pollutant information set according to the soil-affecting items and at least one non-soil-affecting item included in the soil collection component data.

[0027] Specifically, the pollution coefficient of each soil element in the soil pollutant information set can be divided into any one of the multiple pollution coefficient intervals according to the multiple pollution coefficient intervals, and the pollution coefficients of the soil elements divided into the same pollution coefficient interval are classified into one level. For example, this embodiment includes three pollution coefficient intervals, and the pollution coefficient of each soil element in the soil pollutant information set is divided into three pollution coefficient intervals, and the pollution coefficients of the soil elements divided in each pollution coefficient interval are classified into one level, thereby obtaining three soil pollutant information sets.

[0028] In the present application, multiple pollution coefficient intervals can also be obtained through classification network training. For example, a configuration example of a classification network is obtained, each configuration example includes a sample soil pollutant information set of sample soil collection component data, the sample soil pollutant information set includes the pollution coefficient of each soil element in the sample soil collection component data, and the configuration example also includes the pollution coefficient interval corresponding to each soil element in the annotated sample soil collection component data, including the pollution coefficient interval where the soil items are affected and the pollution coefficient interval where the soil items are not affected. When training the classification network, the sample soil pollutant information set is input into the classification network, and the network outputs the pollution coefficient interval corresponding to each soil element in the sample soil collection component data. According to the difference between the pollution coefficient interval corresponding to each soil element output and the pollution coefficient interval corresponding to each soil element annotated, the classification network is optimized. Through the training of a large number of configuration examples, the classification network has the ability to identify the pollution coefficient interval corresponding to each soil element of the soil pollutant information set, thereby dividing the pollution coefficient interval where the soil items are affected and the pollution coefficient interval where the soil items are not affected. After the classification network is trained, the soil pollutant information set is input into the classification network, and multiple soil pollutant information sets can be output.

[0029] In the present application, the soil pollutant information set can be divided into multiple soil pollutant information sets according to the pollution coefficient corresponding to each soil element in the soil pollutant information set. For example, by dividing the pollution coefficients of these soil elements into N sets by mathematical statistics or clustering, each set corresponds to a layer in multiple soil pollutant information sets, and N layers of soil pollutant information sets can be obtained. Which of the N layers of soil pollutant information sets are types of soil-affecting matters and which are types of non-affecting soil matters can be determined according to actual conditions. Specifically, if the pollution coefficient of the soil-affecting matters in the actual conditions is generally smaller than that of the non-affecting soil matters, then the type of soil-affecting matters is the layer where the smaller pollution coefficient is located in the N layers. For example, in the soil collection component data, the soil-affecting matters are eyelashes, and the non-affecting soil matters are irises. The pollution coefficient corresponding to the eyelash soil element is generally smaller than the pollution coefficient corresponding to the iris soil element. Therefore, the layer where the smaller pollution coefficient in the N layers is located is the type of soil-affecting matters, and the layer where the smaller pollution coefficient in the N layers is located is the type of non-affecting soil matters.

[0030] Step 208, obtaining soil component attribute data affecting soil matters according to the types of soil matters affecting soil matters in the plurality of soil pollutant information sets.

[0031] Exemplarily, the types of matters affecting soil are understood as categories of matters affecting soil, and in the present application each category is divided into the same folder.

[0032] The soil component attribute data affecting soil matters refers to the pollution coefficient interval corresponding to the pollution coefficient of the soil element that may correspond to the soil matter in the current soil collection component data, and the pollution coefficient interval affecting soil matters can be determined according to the minimum pollution coefficient and the maximum pollution coefficient of the type of soil matter. After obtaining multiple soil pollutant information sets in step 206, the soil component attribute data affecting soil matters can be determined according to the pollution coefficient of each soil element in the type of soil matter.

[0033] Specifically, the pollution coefficient corresponding to each soil element affecting the type of soil event in multiple soil pollutant information sets is identified, and the soil component attribute data affecting the soil event is determined based on the minimum pollution coefficient and the maximum pollution coefficient in the type of soil event.

[0034] Step 210, through the deep learning-based soil matter pollutant detection network, according to the soil collection component data, multiple soil pollutant information sets and soil component attribute data affecting soil matters, detect the range of pollutant data in the soil collection component data.

[0035] Exemplarily, the scope of pollutant data may be understood as the specific types and properties of pollutants.

[0036] The soil pollutant detection network can be a neural network based on deep learning. The soil pollutant detection network is used to detect the range of pollutant data in the soil collection component data, where the range of pollutant data represents the attributes of each soil element corresponding to the soil impact event in the soil collection component data.

[0037] The embodiment of the present application utilizes a pollutant detection network affecting soil matters to learn and detect the distribution information of soil matters affecting soil matters in multiple soil pollutant information sets from a large number of annotated configuration examples. The pollutant detection network affecting soil matters detects the range of pollutant data in the soil collection component data based on the soil pollutant characteristics of the soil collection component data, the soil key characteristics of multiple soil pollutant information sets, and the key characteristics of soil matters affecting soil matters.

[0038] Specifically, soil collection component data, multiple soil pollutant information sets, and soil component attribute data affecting soil events are input into a pollutant detection network affecting soil events based on deep learning. The soil pollutant characteristics of the soil collection component data and the soil key characteristics of multiple soil pollutant information sets are extracted through the pollutant detection network affecting soil events. Based on the soil pollutant characteristics, soil key characteristics, and key characteristics affecting soil events, the range of pollutant data in the soil collection component data is detected.

[0039] In the above-mentioned soil collection component data processing method, a soil pollutant information set of the soil collection component data is obtained, and the soil pollutant information set is classified to obtain multiple soil pollutant information sets, and the multiple soil pollutant information sets include types of soil-affecting matters and at least one type of non-soil-affecting matters. The embodiment of the present application classifies the soil pollutant information set, and can accurately distinguish the types of soil-affecting matters and at least one type of non-soil-affecting matters in the soil collection component data, and can extract the location information of the soil collection component data, and can accurately distinguish the different levels of the soil collection component data, thereby creating conditions for the subsequent removal of soil-affecting matters in the soil collection component data; according to the types of soil-affecting matters, the soil component attribute data affecting the soil matters are obtained. According to the deep learning-based pollutant detection network for soil matters affecting soil, the range of pollutant data in the soil collected component data is detected according to the soil collected component data, multiple soil pollutant information sets and soil component attribute data affecting soil matters. That is to say, the pollutant detection network for soil matters affecting soil matters, in addition to being based on soil collected component data, also detects the range of pollutant data based on multiple soil pollutant information sets carrying the location information of soil collected component data and soil component attribute data affecting soil matters. It detects soil matters affecting soil according to multi-dimensional information, improves the detection accuracy of soil matters affecting soil, can significantly reduce the impact of soil matters affecting soil on soil collected component data, and improves the accuracy of pollutant identification in soil collected component data.

[0040] In one embodiment, after detecting the range of pollutant data in the soil collected component data, in order to improve the accuracy of pollutant identification in the soil collected component data, the range of pollutant data can be eliminated from the soil collected component data to obtain the data anomaly range. The data anomaly range can be applied to various downstream tasks to achieve the integration and application of multiple functions, such as abnormal data analysis, iris network training, etc.

[0041] Optionally, the soil collection component data processing method provided in the embodiment of the present application also includes: extracting a data anomaly range from the soil collection component data according to the range of pollutant data in the soil collection component data; performing abnormal data analysis on the data anomaly range to obtain an abnormal data analysis result.

[0042] Specifically, according to the range of pollutant data in the soil collection component data, the range other than the range of pollutant data is extracted from the soil collection component data as the data abnormal range; or, according to the range of pollutant data, each soil element that affects soil matters is eliminated, and the normal soil data after elimination is the data abnormal range. Then, the abnormal data analysis network is used to perform abnormal data analysis on the data abnormal range to obtain the abnormal data analysis result.

[0043] In this embodiment, the items affecting the soil in the soil collection component data are eliminated according to the range where the detected pollutant data are located, and then abnormal data analysis is performed on the abnormal range of the data from which the items affecting the soil are eliminated, so that a more accurate abnormal data analysis result can be obtained.

[0044] Optionally, the soil collection component data processing method provided in the embodiment of the present application also includes: extracting a data anomaly range from the soil collection component data according to the range of pollutant data in the soil collection component data; building a configuration example of an abnormal data analysis network according to the extracted data anomaly range; and training the abnormal data analysis network according to the configuration example.

[0045] Among them, the soil collection component data mentioned in the embodiment of the present application can be the initial example soil collection component data used to train the abnormal data analysis network. Accordingly, the data anomaly range refers to the data anomaly range extracted from the example soil collection component data by eliminating soil-affecting matters.

[0046] Specifically, according to the range of pollutant data in the sample soil collection component data, the range other than the range of pollutant data is extracted from the sample soil collection component data as the data abnormal range; or, according to the range of pollutant data, the soil elements that affect soil matters are eliminated, and the normal soil data after elimination is the data abnormal range. Then, the extracted data abnormal range can be used as a high-quality normal soil data example to train the abnormal data analysis network.

[0047] An example of configuring an abnormal data analysis network includes the data anomaly range extracted from the example soil collection component data in the above manner, and the attribute annotation information corresponding to the example soil collection component data, wherein the attribute annotation information is used to annotate the attribute information to which the example soil collection component data belongs. An example of configuring an abnormal data analysis network may be to obtain the attribute annotation information corresponding to the example soil collection component data; and to configure an abnormal data analysis network based on the data anomaly range extracted from the example soil collection component data and the attribute annotation information.

[0048] Training an abnormal data analysis network according to a configuration example includes the following steps: obtaining a configuration example; the configuration example includes a data anomaly range and attribute annotation information corresponding to the soil collection component data for annotating a sample; identifying the attribute regression analysis information corresponding to the data anomaly range in the configuration example through an initial abnormal data analysis network; optimizing the initial abnormal data analysis network according to the difference between the attribute annotation information and the attribute regression analysis information, and after multiple iterations, when the training stop condition is met, a trained abnormal data analysis network is obtained, which can be used to perform abnormal data analysis on the soil collection component data to obtain abnormal data analysis results.

[0049] In this embodiment, according to the range of pollutant data in the example soil collected component data, the data anomaly range is extracted from the example soil collected component data, and according to the extracted data anomaly range, a configuration example of the abnormal data analysis network is built, and the abnormal data analysis network is trained according to the configuration example, which can improve the training effect of the abnormal data analysis network.

[0050] In one embodiment, a soil pollutant information set corresponding to the soil collected component data is generated based on the soil collected component data; the soil pollutant information set is classified to obtain multiple soil pollutant information sets, and the multiple soil pollutant information sets include types of soil-affecting matters and at least one type of non-soil-affecting matters; according to the types of soil-affecting matters in the multiple soil pollutant information sets, soil component attribute data affecting soil matters are obtained; a pollutant detection network affecting soil matters is trained through the soil collected component data, multiple soil pollutant information sets and soil component attribute data affecting soil matters to obtain a trained pollutant detection network affecting soil matters; the trained pollutant detection network affecting soil matters is applied to the identification of soil-affecting matters in new soil collected component data, and the range of pollutant data in the soil collected component data is output through the pollutant detection network affecting soil matters; according to the range of pollutant data, a data anomaly range is extracted from the soil collected component data; the data anomaly range can be used for abnormal data analysis, iris database establishment, abnormal data analysis network training, and human-computer interaction.

[0051] In one embodiment, in order to avoid capturing poor quality soil collection component data or soil contaminant information sets when the shooting device is shaken, resulting in low accuracy of abnormal data analysis or iris network training, this embodiment splices multiple soil contaminant information sets into a more accurate soil contaminant information set to improve the resolution and accuracy of the soil contaminant information set. Specifically, obtaining the soil contaminant information set of soil collection component data includes the following steps: A plurality of initial soil pollutant information sets corresponding to the soil collection component data are obtained; and the plurality of initial soil pollutant information sets are spliced ​​to obtain a soil pollutant information set of the soil collection component data.

[0052] By splicing multiple initial soil pollutant information sets, the pollution coefficients of the multiple initial soil pollutant information sets can be directly operated. For example, the pollution coefficients of each soil element in the multiple initial soil pollutant information sets can be processed using the maximum value method or the depolarization processing method to obtain the spliced ​​soil pollutant information set.

[0053] Specifically, multiple soil pollutant information sets corresponding to the soil collection component data are obtained, or multiple soil collection component data are obtained, and the multiple soil collection component data are converted into multiple soil pollutant information sets using a relevant algorithm; the pollution coefficient of each soil element in the multiple initial soil pollutant information sets is spliced ​​using the maximum value method or the depolarization processing method to obtain the soil pollutant information set of the soil collection component data.

[0054] In this embodiment, multiple initial soil contaminant information sets corresponding to soil collection component data are obtained; the multiple initial soil contaminant information sets are spliced ​​to obtain a soil contaminant information set of soil collection component data, and the multiple soil contaminant information sets are spliced ​​into a more accurate soil contaminant information set, which can solve the problem of inaccurate soil collection component data or soil contaminant information sets, resulting in low accuracy of abnormal data analysis training, and improve the resolution and accuracy of the soil contaminant information set.

[0055] In one embodiment, multiple initial soil contaminant information sets are spliced ​​to obtain a soil contaminant information set of soil collection component data, including the following steps: The pollution coefficients corresponding to the same soil elements in multiple initial soil pollutant information sets are depolarized to obtain the depolarization results of each soil element; based on the depolarization results of each soil element, a soil pollutant information set of soil collection component data is obtained.

[0056] Specifically, the pollution coefficients of the soil elements at the same position in the multiple initial soil pollutant information sets are depolarized to obtain the depolarization results of each soil element, and the depolarization results of each soil element are used as the pollution coefficients of the soil elements at the corresponding positions in the soil pollutant information set of the soil collection component data, thereby obtaining the soil pollutant information set of the soil collection component data. For example, there are 3 initial soil pollutant information sets, and the pollution coefficients of the soil elements at the same position in the 3 initial soil pollutant information sets are 5.1, 5.3 and 5.5 respectively. The pollution coefficient of the soil element is depolarized to obtain a pollution coefficient of 5.3, and the pollution coefficient of the soil element at the same position in the spliced ​​soil pollutant information set is 5.3.

[0057] In this embodiment, the pollution coefficients corresponding to the same soil elements in multiple initial soil pollutant information sets are depolarized to obtain a more accurate soil pollutant information set.

[0058] In one embodiment, the soil contaminant information set is classified to obtain a plurality of soil contaminant information sets, including the following steps: According to the soil pollutant information set, the pollution coefficient corresponding to each soil element in the soil collection component data is obtained; a plurality of pollution coefficient intervals are determined according to each pollution coefficient; each pollution coefficient is divided into any one of the plurality of pollution coefficient intervals; according to the pollution coefficient interval to which each soil element is divided, the soil pollutant information set is divided into a plurality of soil pollutant information sets, wherein the pollution coefficient in the category of matters affecting soil is smaller than the pollution coefficient in the category of matters not affecting soil.

[0059] In this embodiment, according to the pollution coefficient of each soil element in the soil collection component data, the pollution coefficient of each soil element is divided into any one of the multiple pollution coefficient intervals, and the pollution coefficients of the soil elements divided into the same pollution coefficient interval are classified into one level. For example, this embodiment includes three pollution coefficient intervals, and the pollution coefficients of each soil element in the soil pollutant information set are divided into three pollution coefficient intervals, and the pollution coefficients of the soil elements divided in each pollution coefficient interval are classified into one level, thereby obtaining a three-level soil pollutant information set.

[0060] In one embodiment, obtaining soil component attribute data affecting soil matters according to the types of soil matters affecting soil matters in a plurality of soil pollutant information sets comprises the following steps: The pollution coefficient corresponding to each soil element in the type of soil-affecting event in multiple soil pollutant information sets is obtained; and the soil component attribute data affecting the soil event is determined based on the minimum pollution coefficient and the maximum pollution coefficient among the pollution coefficients corresponding to each soil element.

[0061] Among them, the soil component attribute data affecting soil matters refer to the pollution coefficient range affecting soil matters, and the pollution coefficient range affecting soil matters can be determined based on the minimum pollution coefficient and the maximum pollution coefficient affecting soil matters.

[0062] Specifically, after obtaining the types of matters affecting soil, the pollution coefficient of each soil element in the types of matters affecting soil can be obtained, the minimum pollution coefficient and the maximum pollution coefficient of each soil element in the types of matters affecting soil can be determined, and the pollution coefficient range of matters affecting soil can be determined based on the minimum pollution coefficient and the maximum pollution coefficient of each soil element in the types of matters affecting soil.

[0063] In this embodiment, the soil component attribute data affecting the soil matter is determined based on the minimum pollution coefficient and the maximum pollution coefficient among the pollution coefficients corresponding to each soil element in the type of soil matter affecting the soil matter, providing a data basis for the subsequent pollutant detection network affecting the soil matter to detect the range of the pollutant data.

[0064] In one embodiment, a soil pollutant detection network based on deep learning is used to detect the range of pollutant data in the soil collected component data according to soil collected component data, multiple soil pollutant information sets, and soil component attribute data affecting soil matters, including the following steps: 1. Input soil collection component data, multiple soil pollutant information sets, and soil component attribute data affecting soil events into a deep learning-based pollutant detection network affecting soil events.

[0065] Among them, the soil component attribute data affecting soil matters refers to the pollution coefficient interval corresponding to the pollution coefficient of the soil elements that may correspond to the soil matters affecting soil matters in the current soil collection component data.

[0066] Specifically, soil collection component data, multiple soil pollutant information sets, and soil component attribute data affecting soil matters are input into a soil matter affecting pollutant detection network based on deep learning.

[0067] 2. Through the pollutant detection network affecting soil matters, the soil pollutant characteristics of the soil collection component data are extracted, the soil key characteristics of multiple soil pollutant information sets are extracted, the soil component attribute data affecting soil matters are encoded, and the key characteristics affecting soil matters are obtained. According to the soil pollutant characteristics, soil key characteristics and key characteristics affecting soil matters, the percentage of each soil element in the soil collection component data belonging to the soil element affecting soil matters is output.

[0068] Specifically, through the pollutant detection network affecting soil matters, soil pollutant characteristics of soil collection component data and soil key characteristics of multiple soil pollutant information sets are extracted, and the soil component attribute data affecting soil matters are encoded through the pollutant detection network affecting soil matters to obtain key characteristics affecting soil matters, and the soil pollutant characteristics, soil key characteristics and key characteristics affecting soil matters are input into the pollutant detection network affecting soil matters, and the pollutant detection network affecting soil matters outputs the percentage of each soil element in the soil collection component data belonging to the soil element affecting soil matters based on the soil pollutant characteristics, soil key characteristics and key characteristics affecting soil matters.

[0069] 3. Determine the scope of pollutant data in the soil collection component data based on the percentage of each soil element in the soil collection component data that belongs to the soil elements that affect soil matters.

[0070] Specifically, if the percentage of soil elements in the soil collection component data output by the soil matter pollutant detection network that belong to soil matters affecting soil matters is greater than the preset percentage value, then it is determined that the soil element belongs to the soil matter affecting soil matters, and the scope of the pollutant data is determined based on the description content of all soil elements that belong to the soil matters affecting soil matters.

[0071] For example, the soil elements in the soil collection component data output by the soil pollutant detection network affecting soil matters, whose percentage of the soil elements belonging to the soil elements affecting soil matters is greater than 0.8, are regarded as soil elements affecting soil matters, and the soil elements in the soil collection component data output by the soil pollutant detection network affecting soil matters, whose percentage of the soil elements belonging to the soil elements affecting soil matters is less than 0.1, are regarded as soil elements not affecting soil matters. In this way, the soil elements affecting soil matters in the soil collection component data are obtained, and the range of the pollutant data is determined according to the description content of the soil elements affecting soil matters.

[0072] In this embodiment, based on the soil pollutant characteristics of the soil collected component data, the soil key characteristics of multiple soil pollutant information sets and the key characteristics of matters affecting the soil, the range of the pollutant data in the soil collected component data is detected, and the matters affecting the soil are detected based on the multi-dimensional information, which improves the accuracy of detecting matters affecting the soil, can greatly reduce the impact of matters affecting the soil on the soil collected component data, and improve the accuracy of pollutant identification in the soil collected component data.

[0073] In this application, this embodiment provides a detailed step of a soil pollutant intelligent detection method, which specifically includes the following steps: 1. Obtain soil collection composition data, which includes matters affecting the soil.

[0074] 2. Obtain multiple initial soil pollutant information sets corresponding to soil collection component data.

[0075] 3. Depolarize the pollution coefficients corresponding to the same soil elements in multiple initial soil pollutant information sets to obtain the depolarization results of each soil element; and obtain the soil pollutant information set of soil collection component data based on the depolarization results of each soil element.

[0076] 4. Based on the soil pollutant information set, obtain the pollution coefficient corresponding to each soil element in the soil collection component data.

[0077] 5. Determine multiple pollution coefficient intervals based on each pollution coefficient.

[0078] 6. Divide each pollution coefficient into any one of a plurality of pollution coefficient intervals.

[0079] 7. According to the pollution coefficient range to which each soil element is divided, the soil pollutant information set is divided into multiple soil pollutant information sets, wherein the multiple soil pollutant information sets include a type of soil-affecting matter and at least one type of non-affecting soil matter; the pollution coefficient in the type of soil-affecting matter is smaller than the pollution coefficient in the type of non-affecting soil matter.

[0080] 8. Obtain the pollution coefficient corresponding to each soil element in the category of soil-affecting matters in multiple soil pollutant information sets.

[0081] 9. Determine the soil component attribute data affecting soil matters based on the minimum and maximum pollution coefficients among the pollution coefficients corresponding to each soil element.

[0082] 10. Input the soil collection component data, multiple soil pollutant information sets and soil component attribute data affecting soil matters into the soil matter affecting pollutant detection network based on deep learning.

[0083] 11. Through the pollutant detection network affecting soil matters, the soil pollutant characteristics of the soil collection component data are extracted, the soil key characteristics of multiple soil pollutant information sets are extracted, the soil component attribute data affecting soil matters are encoded, and the key characteristics affecting soil matters are obtained. According to the soil pollutant characteristics, soil key characteristics and key characteristics affecting soil matters, the percentage of each soil element in the soil collection component data belonging to the soil element affecting soil matters is output.

[0084] 12. Determine the scope of pollutant data in the soil collection component data based on the percentage of each soil element in the soil collection component data that belongs to the soil elements that affect soil matters.

[0085] 13. Based on the range of pollutant data in the soil collected component data, extract the data anomaly range from the soil collected component data.

[0086] 14. Perform abnormal data analysis on the data abnormal range to obtain abnormal data analysis results; or, based on the extracted data abnormal range, build a configuration example of an abnormal data analysis network; and train the abnormal data analysis network based on the configuration example.

[0087] In this embodiment, the soil contaminant information set is classified, and the types of soil-affecting items and at least one non-soil-affecting item type in the soil collection component data can be accurately distinguished, and the location information of the soil collection component data can be extracted, and the different levels of the soil collection component data can be accurately distinguished, thereby creating conditions for the subsequent removal of soil-affecting items in the soil collection component data; the soil-affecting items are detected based on multi-dimensional information, which improves the accuracy of detecting pollutant data, and can greatly reduce the impact of soil-affecting items on soil collection component data, thereby improving the accuracy of pollutant identification in soil collection component data; multiple soil contaminant information sets are spliced ​​into a more accurate soil contaminant information set, which can avoid capturing poor quality soil collection component data or soil contaminant information sets when the shooting equipment is shaken, resulting in abnormal data analysis or low accuracy of iris network training, thereby improving the resolution and accuracy of the soil contaminant information set.

[0088] In one embodiment, after obtaining the sample soil collection component data, a configuration example of a pollutant detection network affecting soil matters is constructed based on the sample soil collection component data and annotation information about the range of pollutant data in the sample soil collection component data, and a configuration example of a pollutant detection network affecting soil matters is obtained, wherein the configuration example includes the sample soil collection component data and annotation information about the range of pollutant data in the sample soil collection component data; a soil pollutant information set of the sample soil collection component data is obtained based on the sample soil collection component data; the soil pollutant information set of the sample soil collection component data is classified to obtain the soil pollutant information set corresponding to the sample soil collection component data. A plurality of soil pollutant information sets, wherein the plurality of soil pollutant information sets include soil-affecting item categories and at least one non-soil-affecting item category; obtaining soil component attribute data affecting soil items according to the soil-affecting item categories in the plurality of soil pollutant information sets; obtaining regression analysis information on the range of pollutant data in the example soil collection component data according to example soil collection component data, a plurality of soil pollutant information sets corresponding to the example soil collection component data, and soil component attribute data affecting soil items through a pollutant detection network affecting soil items; optimizing the pollutant detection network affecting soil items according to the difference between the regression analysis information and the annotation information.

[0089] In one embodiment, a method for training a soil pollutant detection network is provided, which specifically includes the following steps: Step 502: Obtain a configuration example of a soil pollutant detection network, wherein the configuration example includes sample soil collection component data and annotation information about a range of pollutant data in the sample soil collection component data.

[0090] Step 504: Obtain a soil pollutant information set of the sample soil collection component data.

[0091] Step 506: Classify the soil pollutant information set of the example soil collection component data to obtain multiple soil pollutant information sets corresponding to the example soil collection component data, and the multiple soil pollutant information sets include a type of soil-affecting matter and at least one type of non-soil-affecting matter.

[0092] Among them, classification is the process of classifying the corresponding soil elements in the soil pollutant information set according to the soil-affecting items and at least one non-soil-affecting item included in the sample soil collection component data.

[0093] Specifically, the pollution coefficient of each soil element in the soil pollutant information set can be divided into any one of the multiple pollution coefficient intervals according to the multiple pollution coefficient intervals, and the pollution coefficients of the soil elements divided into the same pollution coefficient interval are classified into one level. For example, this embodiment includes three pollution coefficient intervals, and the pollution coefficient of each soil element in the soil pollutant information set is divided into three pollution coefficient intervals, and the pollution coefficients of the soil elements divided in each pollution coefficient interval are classified into one level, thereby obtaining three soil pollutant information sets.

[0094] In the present application, multiple pollution coefficient intervals can also be obtained through classification network training.

[0095] In the present application, the soil pollutant information set can be divided into multiple soil pollutant information sets according to the pollution coefficient corresponding to each soil element in the soil pollutant information set. For example, by dividing the pollution coefficients of these soil elements into N sets by mathematical statistics or clustering, each set corresponds to a layer in multiple soil pollutant information sets, and N layers of soil pollutant information sets can be obtained. Which of the N layers of soil pollutant information sets are types of soil-affecting matters and which are types of non-affecting soil matters can be determined according to actual conditions. Specifically, if the pollution coefficient of the soil-affecting matters in the actual conditions is generally smaller than that of the non-affecting soil matters, then the type of soil-affecting matters is the layer where the smaller pollution coefficient is located in the N layers. For example, in the sample soil collection component data, the soil-affecting matters are eyelashes, and the non-affecting soil matters are irises. The pollution coefficient corresponding to the eyelash soil element is generally smaller than the pollution coefficient corresponding to the iris soil element. Therefore, the layer where the smaller pollution coefficient is located in the N layers is the type of soil-affecting matters, and the layer where the smaller pollution coefficient is located in the N layers is the type of non-affecting soil matters.

[0096] Step 508: Obtain soil component attribute data affecting soil matters based on the types of soil matters affecting soil matters in multiple soil pollutant information sets.

[0097] Among them, the soil component attribute data affecting soil matters refers to the pollution coefficient interval corresponding to the pollution coefficient of the soil elements that may correspond to the soil matters in the current soil collection component data. The pollution coefficient interval affecting soil matters can be determined based on the minimum pollution coefficient and the maximum pollution coefficient of the type of soil matters. After obtaining multiple soil pollutant information sets, the soil component attribute data affecting soil matters can be determined based on the pollution coefficient of each soil element in the type of soil matters.

[0098] Specifically, the pollution coefficient corresponding to each soil element affecting the type of soil event in multiple soil pollutant information sets is identified, and the soil component attribute data affecting the soil event is determined based on the minimum pollution coefficient and the maximum pollution coefficient in the type of soil event.

[0099] Step 510: Obtain regression analysis information about the range of pollutant data in the example soil collection component data based on the example soil collection component data, multiple soil pollutant information sets corresponding to the example soil collection component data, and soil component attribute data affecting soil matters through the soil matter-affecting pollutant detection network.

[0100] The soil pollutant detection network can be a neural network based on deep learning. The soil pollutant detection network is used to detect the range of pollutant data in the sample soil collection component data, where the range of pollutant data represents the attributes of each soil element corresponding to the soil impact event in the soil collection component data.

[0101] The embodiment of the present application utilizes a pollutant detection network affecting soil matters to learn and detect the distribution information of soil matters affecting soil in multiple soil pollutant information sets from a large number of annotated configuration examples. The pollutant detection network affecting soil matters detects the range of pollutant data in the example soil collection component data based on the soil pollutant characteristics of the example soil collection component data, the soil key characteristics of multiple soil pollutant information sets, and the key characteristics of soil matters affecting soil. The soil matters affecting soil are detected based on multi-dimensional information, thereby improving the accuracy of detecting pollutant data.

[0102] Specifically, the example soil collection component data, multiple soil pollutant information sets, and soil component attribute data affecting soil events are input into a pollutant detection network affecting soil events based on deep learning. The soil pollutant characteristics of the example soil collection component data and the soil key characteristics of multiple soil pollutant information sets are extracted through the pollutant detection network affecting soil events. Based on the soil pollutant characteristics, soil key characteristics, and key characteristics affecting soil events, the regression analysis information of the range where the pollutant data in the example soil collection component data is located is detected.

[0103] Step 512: Optimize the soil pollutant detection network based on the difference between the regression analysis information and the annotation information.

[0104] Among them, optimizing the soil pollutant detection network specifically refers to adjusting the parameters of the soil pollutant detection network, for example, adjusting the network results and learning rate of the soil pollutant detection network.

[0105] Specifically, the configuration example of the soil pollutant detection network is divided into a training set and a test set, and a deep learning soil pollutant detection network is configured. The training set is used to train the soil pollutant detection network so that the soil pollutant detection network has the ability to detect soil pollutants from soil collected component data. The test set is used to evaluate the performance of the soil pollutant detection network, for example, to evaluate the accuracy of the soil pollutant detection network, and to determine whether the performance of the soil pollutant detection network meets the requirements. If the performance of the soil pollutant detection network meets the requirements, the soil pollutant detection network is saved. If the performance of the soil pollutant detection network does not meet the requirements, the parameters of the soil pollutant detection network are optimized according to the difference between the regression analysis information and the annotation information, for example, the network results and learning rate of the soil pollutant detection network are adjusted; until the performance of the soil pollutant detection network meets the requirements, the training of the soil pollutant detection network is stopped to obtain a trained soil pollutant detection network.

[0106] In this embodiment, a configuration example of a soil-affecting pollutant detection network is obtained, the configuration example includes example soil collection component data and annotation information about the range of pollutant data in the example soil collection component data; a soil pollutant information set of the example soil collection component data is obtained, the soil pollutant information set is classified to obtain multiple soil pollutant information sets, the multiple soil pollutant information sets include types of soil-affecting items and at least one type of non-soil-affecting items. The embodiment of the present application classifies the soil pollutant information set, and can accurately distinguish the types of soil-affecting items and at least one type of non-soil-affecting items in the example soil collection component data, and can extract the location information of the example soil collection component data, and can accurately distinguish the types of soil-affecting items in the example soil collection component data. Different levels are created to create conditions for the subsequent removal of soil-affecting items in the sample soil collection component data; according to the types of soil-affecting items, the soil component attribute data of the soil-affecting items are obtained, and through the pollutant detection network affecting soil items based on deep learning, according to the sample soil collection component data, multiple soil pollutant information sets and soil component attribute data of the soil-affecting items, the regression analysis information about the range of pollutant data in the sample soil collection component data is obtained; according to the difference between the regression analysis information and the annotation information, the pollutant detection network affecting soil items is optimized, so that the trained pollutant detection network affecting soil items has the ability to detect soil-affecting items based on multi-dimensional information, thereby improving the ability of the pollutant detection network affecting soil items to accurately detect pollutant data.

[0107] In the present application, obtaining a soil pollutant information set of example soil collection component data includes: obtaining multiple initial soil pollutant information sets corresponding to the example soil collection component data; splicing the multiple initial soil pollutant information sets to obtain the soil pollutant information set of the example soil collection component data. When multiple initial soil pollutant information sets are spliced, the pollution coefficients of the multiple initial soil pollutant information sets can be directly operated, for example, the pollution coefficients of each soil element in the multiple initial soil pollutant information sets are processed using the maximum value method or the depolarization processing method to obtain the spliced ​​soil pollutant information set.

[0108] In one embodiment, multiple initial soil contaminant information sets are spliced ​​to obtain a soil contaminant information set of example soil collection component data, including the following steps: depolarizing the pollution coefficients corresponding to the same soil elements in the multiple initial soil contaminant information sets to obtain the depolarization processing results of each soil element; and obtaining the soil contaminant information set of the example soil collection component data based on the depolarization processing results of each soil element.

[0109] In the present application, a soil pollutant information set of example soil collection component data is classified to obtain multiple soil pollutant information sets corresponding to the example soil collection component data, including the following steps: according to the soil pollutant information set, a pollution coefficient corresponding to each soil element in the example soil collection component data is obtained; according to each pollution coefficient, a plurality of pollution coefficient intervals are determined; each pollution coefficient is divided into any one of the plurality of pollution coefficient intervals; according to the pollution coefficient interval to which each soil element is divided, the soil pollutant information set is divided into a plurality of soil pollutant information sets, wherein the pollution coefficient in the category of affecting soil matters is smaller than the pollution coefficient in the category of not affecting soil matters.

[0110] In this embodiment, according to the pollution coefficient of each soil element in the example soil collection component data, the pollution coefficient of each soil element is divided into any pollution coefficient interval of multiple pollution coefficient intervals, and the pollution coefficients of the soil elements divided into the same pollution coefficient interval are classified into one level.

[0111] In the present application, soil component attribute data affecting soil matters are obtained based on the types of soil affecting matters in multiple soil pollutant information sets, including: obtaining the pollution coefficient corresponding to each soil element in the types of soil affecting matters in multiple soil pollutant information sets; determining the soil component attribute data affecting soil matters based on the minimum pollution coefficient and the maximum pollution coefficient among the pollution coefficients corresponding to each soil element.

[0112] In one embodiment, obtaining regression analysis information about the range of pollutant data in the example soil collection component data according to the example soil collection component data, a plurality of soil pollutant information sets corresponding to the example soil collection component data, and soil component attribute data affecting soil matters through a soil matter-affecting pollutant detection network includes the following steps: 1. Input sample soil collection component data, multiple soil pollutant information sets, and soil component attribute data affecting soil events into a deep learning-based soil event pollutant detection network.

[0113] The soil component attribute data affecting soil events refers to the pollution coefficient interval corresponding to the pollution coefficient of the soil elements that may correspond to the soil events affecting soil events in the current example soil collection component data.

[0114] Specifically, sample soil collection component data, multiple soil pollutant information sets, and soil component attribute data affecting soil matters are input into a soil matter affecting pollutant detection network based on deep learning.

[0115] 2. Through the pollutant detection network affecting soil matters, the soil pollutant characteristics of the sample soil collection component data are extracted, the soil key characteristics of multiple soil pollutant information sets are extracted, the soil component attribute data affecting soil matters are encoded, and the key characteristics affecting soil matters are obtained. According to the soil pollutant characteristics, soil key characteristics and key characteristics affecting soil matters, the percentage of each soil element in the sample soil collection component data belonging to the soil element affecting soil matters is output.

[0116] 3. Determine the regression analysis information of the range of pollutant data in the sample soil collection component data based on the percentage of each soil element in the sample soil collection component data that belongs to the soil element affecting soil events.

[0117] In this embodiment, based on the soil pollutant characteristics of the example soil collected component data, the soil key characteristics of multiple soil pollutant information sets, and the key characteristics of matters affecting soil, the range of pollutant data in the example soil collected component data is detected, and matters affecting soil are detected based on multi-dimensional information, thereby improving the accuracy of detecting pollutant data, significantly reducing the impact of matters affecting soil on soil collected component data, and improving the accuracy of pollutant identification in soil collected component data.

[0118] In one embodiment, this embodiment provides a detailed step of a soil pollutant detection network training method, which specifically includes the following steps: 1. Obtain a configuration example of a soil pollutant detection network, the configuration example including sample soil collection component data and annotation information about the range of pollutant data in the sample soil collection component data.

[0119] 2. Obtain multiple initial soil pollutant information sets corresponding to the sample soil collection component data; depolarize the pollution coefficients corresponding to the same soil elements in the multiple initial soil pollutant information sets to obtain the depolarization results of each soil element; and obtain the soil pollutant information set of the sample soil collection component data based on the depolarization results of each soil element.

[0120] 3. According to the soil pollutant information set, obtain the pollution coefficient corresponding to each soil element in the sample soil collection component data; determine multiple pollution coefficient intervals according to each pollution coefficient; divide each pollution coefficient into any one of the multiple pollution coefficient intervals; according to the pollution coefficient interval to which each soil element is divided, divide the soil pollutant information set into multiple soil pollutant information sets, wherein the multiple soil pollutant information sets include a soil-affecting event category and at least one non-affecting soil event category; the pollution coefficient in the soil-affecting event category is smaller than the pollution coefficient in the non-affecting soil event category.

[0121] 4. Obtain the pollution coefficient corresponding to each soil element in the type of soil-affecting matter in multiple soil pollutant information sets; determine the soil component attribute data affecting the soil matter based on the minimum pollution coefficient and the maximum pollution coefficient among the pollution coefficients corresponding to each soil element.

[0122] 5. Input the sample soil collection component data, multiple soil pollutant information sets and soil component attribute data affecting soil events into the soil event pollutant detection network based on deep learning.

[0123] 6. Through the pollutant detection network affecting soil matters, the soil pollutant characteristics of the sample soil collection component data are extracted, the soil key characteristics of multiple soil pollutant information sets are extracted, the soil component attribute data affecting soil matters are encoded, and the key characteristics affecting soil matters are obtained. According to the soil pollutant characteristics, soil key characteristics and key characteristics affecting soil matters, the percentage of each soil element in the sample soil collection component data belonging to the soil element affecting soil matters is output.

[0124] 7. Determine the regression analysis information of the range of the pollutant data in the sample soil collection component data based on the percentage of each soil element in the sample soil collection component data that belongs to the soil element affecting the soil event.

[0125] 8. Based on the difference between regression analysis information and annotation information, optimize the pollutant detection network affecting soil matters.

[0126] In this embodiment, the soil pollutant information set is classified, and the types of soil-affecting items and at least one non-soil-affecting item type in the example soil collection component data can be accurately distinguished, and the location information of the soil collection component data can be extracted, and the different levels of the soil collection component data can be accurately distinguished, thereby creating conditions for the subsequent removal of soil-affecting items in the example soil collection component data; according to the types of soil-affecting items, soil component attribute data of the soil-affecting items are obtained, and through the soil-affecting item pollutant detection network based on deep learning, according to the example soil collection component data, multiple soil pollutant information sets and soil component attribute data of the soil-affecting items, regression analysis information about the range of pollutant data in the example soil collection component data is obtained; according to the difference between the regression analysis information and the annotation information, the soil-affecting item pollutant detection network is optimized, so that the trained soil-affecting item pollutant detection network has the ability to detect soil-affecting items based on multi-dimensional information, thereby improving the ability of the soil-affecting item pollutant detection network to accurately detect pollutant data.

[0127] On the basis of the above, a soil pollutant intelligent detection device is provided, the device comprising: A data acquisition module, used to obtain soil collection component data, wherein the soil collection component data includes matters affecting the soil; An information acquisition module, used to obtain a soil pollutant information set of the soil collection component data; An information classification module, used to classify the soil pollutant information set to obtain a plurality of soil pollutant information sets, wherein the plurality of soil pollutant information sets include a soil-affecting event category and at least one non-soil-affecting event category; A data acquisition module, used for acquiring soil component attribute data of the soil-affecting items according to the types of soil-affecting items in the plurality of soil pollutant information sets; The pollutant determination module is used to detect the range of pollutant data in the soil collection component data according to the soil collection component data, the multiple soil pollutant information sets and the soil component attribute data affecting the soil matters through a deep learning-based soil matter impact pollutant detection network.

[0128] On the basis of the above, a soil contaminant intelligent detection system is shown, which includes a processor and a memory that communicate with each other, and the processor is used to read a computer program from the memory and execute it to implement the above method.

[0129] Based on the above, a computer-readable storage medium is also provided, on which a computer program stored implements the above method when running.

[0130] In summary, based on the above scheme, a configuration example of a pollutant detection network affecting soil matters is obtained, the configuration example includes example soil collection component data and annotation information about the range of pollutant data in the example soil collection component data; a soil pollutant information set of the example soil collection component data is obtained, and the soil pollutant information set is classified to obtain multiple soil pollutant information sets, and the multiple soil pollutant information sets include types of soil matters affecting and at least one type of non-soil matters affecting. The embodiment of the present application classifies the soil pollutant information set, which can accurately distinguish the types of soil matters affecting and at least one type of non-soil matters affecting in the example soil collection component data, and can extract the location information of the example soil collection component data, which can accurately distinguish the example soil collection component data. Different levels of data are collected, thereby creating conditions for the subsequent removal of soil-affecting items in the sample soil collection component data; according to the types of soil-affecting items, the soil component attribute data of the soil-affecting items are obtained, and through the pollutant detection network affecting soil items based on deep learning, according to the sample soil collection component data, multiple soil pollutant information sets and soil component attribute data affecting soil items, regression analysis information about the range of pollutant data in the sample soil collection component data is obtained; according to the difference between the regression analysis information and the annotation information, the pollutant detection network affecting soil items is optimized, so that the trained pollutant detection network affecting soil items has the ability to detect soil-affecting items based on multi-dimensional information, thereby improving the ability of the pollutant detection network affecting soil items to accurately detect pollutant data.

[0131] 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).

[0132] 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 soil pollutant intelligent detection method, characterized in that: The method comprises: Obtaining soil collection component data, wherein the soil collection component data includes matters affecting the soil; Obtaining a soil pollutant information set of the soil collection component data; Classifying the soil pollutant information set to obtain a plurality of soil pollutant information sets, wherein the plurality of soil pollutant information sets include a soil-affecting event category and at least one non-soil-affecting event category; According to the types of soil-affecting matters in the plurality of soil pollutant information sets, obtaining soil component attribute data of the soil-affecting matters; Through a deep learning-based soil matter-affecting pollutant detection network, the range of pollutant data in the soil collection component data is detected according to the soil collection component data, the multiple soil pollutant information sets and the soil component attribute data affecting the soil matters.

2. The method according to claim 1, characterized in that The method further comprises: Extracting a data anomaly range from the soil collected component data according to a range where pollutant data in the soil collected component data are located; An abnormal data analysis is performed on the data abnormal range to obtain an abnormal data analysis result.

3. The method according to claim 1, characterized in that: The method further comprises: Extracting a data anomaly range from the soil collected component data according to a range where pollutant data in the soil collected component data are located; An example of configuring an abnormal data analysis network according to the extracted data abnormality range; The abnormal data analysis network is trained according to the configuration example.

4. The method according to claim 1, characterized in that: The step of obtaining the soil pollutant information set of the soil collection component data includes: Obtaining a plurality of initial soil pollutant information sets corresponding to the soil collection component data; The multiple initial soil pollutant information sets are spliced ​​together to obtain the soil pollutant information set of the soil collection component data.

5. The method according to claim 4, characterized in that The step of combining the multiple initial soil pollutant information sets to obtain the soil pollutant information set of the soil collection component data comprises: Depolarizing the pollution coefficients corresponding to the same soil element in the plurality of initial soil pollutant information sets to obtain a depolarization result for each soil element; According to the depolarization processing result of each soil element, a soil pollutant information set of the soil collection component data is obtained.

6. The method according to claim 1, characterized in that The soil pollutant information set is classified to obtain a plurality of soil pollutant information sets, including: According to the soil pollutant information set, a pollution coefficient corresponding to each soil element in the soil collection component data is obtained; Determining a plurality of pollution coefficient intervals according to each of the pollution coefficients; Dividing each of the pollution coefficients into any one of the multiple pollution coefficient intervals; According to the pollution coefficient interval to which each soil element is divided, the soil pollutant information set is divided into multiple soil pollutant information sets, wherein the pollution coefficient in the type of soil-affecting matters is smaller than the pollution coefficient in the type of non-soil-affecting matters.

7. The method according to claim 1, characterized in that The step of obtaining soil component attribute data affecting the soil event according to the types of soil event affecting the soil event in the plurality of soil pollutant information sets includes: Obtaining the pollution coefficient corresponding to each soil element in the soil-affecting event category in the plurality of soil pollutant information sets; The soil component attribute data affecting the soil matters are determined based on the minimum pollution coefficient and the maximum pollution coefficient among the pollution coefficients corresponding to the soil elements.

8. The method according to claim 1, characterized in that The deep learning-based soil event-affecting pollutant detection network detects the range of pollutant data in the soil collection component data according to the soil collection component data, the multiple soil pollutant information sets, and the soil component attribute data affecting the soil event, including: Inputting the soil collected component data, the plurality of soil pollutant information sets and the soil component attribute data affecting soil matters into a soil matter affecting pollutant detection network based on deep learning; Extracting soil pollutant characteristics of the soil collection component data and soil key characteristics of the plurality of soil pollutant information sets through the soil matter-affecting pollutant detection network, encoding the soil component attribute data affecting soil matters to obtain key characteristics affecting soil matters, and outputting the percentage of each soil element in the soil collection component data belonging to the soil element affecting soil matters according to the soil pollutant characteristics, the soil key characteristics and the key characteristics affecting soil matters; The range of the pollutant data in the soil collected component data is determined according to the percentage of each soil element in the soil collected component data belonging to the soil element affecting soil matters.

9. The method according to any one of claims 1 to 8, characterized in that: The method further comprises: Obtaining a configuration example of a soil-affecting pollutant detection network, the configuration example including sample soil collection component data and annotation information about a range where pollutant data in the sample soil collection component data is located; Obtaining a soil pollutant information set of the sample soil collection component data; Classifying the soil pollutant information set of the example soil collection component data to obtain a plurality of soil pollutant information sets corresponding to the example soil collection component data, wherein the plurality of soil pollutant information sets include a soil-affecting event category and at least one non-soil-affecting event category; According to the types of soil-affecting matters in the plurality of soil pollutant information sets, obtaining soil component attribute data of the soil-affecting matters; Obtaining regression analysis information about the range of pollutant data in the example soil collection component data according to the example soil collection component data, a plurality of soil pollutant information sets corresponding to the example soil collection component data, and the soil component attribute data affecting the soil matter through a soil matter-affecting pollutant detection network; According to the difference between the regression analysis information and the annotation information, the soil pollutant detection network is optimized.

10. A soil pollutant intelligent detection system, 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 9.

Images