Soil data rapid extraction, integration and cartographic updating method and system based on knowledge enhancement large language model

Abstract

The invention discloses a soil data rapid extraction, integration and cartographic updating method and system based on a knowledge-enhanced large language model, and the method comprises the steps: carrying out the textualization processing of soil survey data and the construction of a knowledge base, obtaining background knowledge assisting the large language model in soil data extraction through RAG retrieval, and fusing the background knowledge with an input literature text; constructing an optimized soil data extraction prompt by utilizing a thinking chain prompt method, and accurately and quickly extracting soil data in literatures; cleaning and treating the extracted heterogeneous soil data to form standard and unified structured soil data, and integrating the structured soil data with historical soil data; and coupling environmental factor data influencing the soil, and drawing and updating the region by using an empirical Bayesian Criging regression method. The method can replace manual work to extract the soil data in the literature, and complete the cleaning and integration of the extracted data and the updating of the regional soil map, thereby greatly improving the efficiency of public source soil data acquisition and map updating.

Description

Technical Field

[0001] The present invention belongs to the technical field of soil data extraction, integration and regional mapping, and in particular to a soil data rapid extraction, integration and mapping update method and system based on a knowledge-enhanced large language model. Background Art

[0002] The rapid acquisition of soil data is of great significance in the fields of agricultural production, ecological environmental protection, and government management of cultivated land occupation and compensation balance. At present, the acquisition of soil data mainly relies on field sampling surveys and laboratory analysis. The data is accurate but inefficient. The acquisition of data through remote sensing technology, sensor technology, and new data processing methods is efficient but the data accuracy is difficult to guarantee.

[0003] In recent years, a large number of publicly published scientific research documents, survey reports and other materials contain a large amount of the latest soil data, which has attracted the attention of soil researchers. However, although the soil data in soil-related literature, especially in published scientific research papers, are scientific data and can withstand verification, more than 90% of them are unstructured data and difficult to use directly. In recent years, natural language processing technology (Natural Language Processing, NLP), especially large language models (Large Language Model, LLM) have made rapid progress, which has promoted the improvement of multimodal information acquisition efficiency and made it easier to complete accurate data extraction tasks from texts efficiently and accurately.

[0004] Therefore, there is an urgent need for a method for accurate soil data extraction, integration and regional mapping based on a large language model, and the development of an application system for rapid and efficient soil data acquisition and mapping updates, thereby forming a technical chain for efficient and intelligent soil data collection, integration and application. Summary of the invention

[0005] The purpose of the present invention is to solve the problems raised in the background technology, automatically and accurately extract soil data from continuously updated soil-related scientific research papers, survey reports and other public documents, clean and integrate the extracted data and update it into the existing soil database, and then use geostatistical analysis methods to realize regional soil mapping updates.

[0006] In order to achieve the purpose of the present invention, the present invention discloses a soil data rapid extraction, integration and mapping update method based on a knowledge-enhanced large language model, comprising the following steps:

[0007] Step S1, text processing of soil survey data, processing paper scans and electronic texts of soil survey knowledge materials to form text data that can be recognized by computers;

[0008] Step S2, segmenting and storing the text of soil survey related knowledge to form a vectorized knowledge base;

[0009] Step S3: Knowledge enhancement fusion: using RAG technology to search from the soil survey knowledge base, generate background knowledge to assist the large language model in soil data extraction, and fuse it with the input literature text;

[0010] Step S4: soil data is extracted quickly and accurately. Based on the integrated background knowledge and literature text, the optimization prompt is constructed using the thinking chain prompt method to clarify the thinking path of extracting soil data with a large model, and to extract soil data from literature quickly and accurately.

[0011] Step S5, soil data integration, performing redundancy removal, transformation, unit conversion, coordinate conversion, and numerical conversion operations on the extracted heterogeneous soil data to form standardized and unified structured soil data, and integrating it with historical soil data;

[0012] Step S6, soil mapping, based on the integration results of extracted soil data and historical soil data, coupled with environmental factor data affecting soil, including but not limited to topography, climate, land use, remote sensing monitoring, using the empirical Bayesian Kriging regression EBKR method to update the mapping of the region.

[0013] Furthermore, step S1 includes the following steps:

[0014] Step S11, text processing of soil survey data, which is divided into two processing processes according to the type of input data. If the input data is paper scanned data, it is recognized as text data using OCR technology, and the text data is input into step S12 for processing;

[0015] Step S12: If the input is text data, perform text noise removal processing, which includes but is not limited to special format processing, special character elimination, stop word processing, HTML tag removal, and header and footer removal to form a high-quality soil survey knowledge text.

[0016] Further, step S2 includes the following steps:

[0017] Step S21, constructing a knowledge base for the input high-quality soil survey knowledge text through the FastGPT framework;

[0018] Step S22, the soil survey knowledge text is segmented according to the segmentation + direct block method, the ideal block length is set to 1024, and the segment separator is customized;

[0019] Step S23: Use the LLM text vectorization method, including but not limited to embedding-2 and Baichuan-Text-Embedding, to embed the divided text, and store the vectorized data in a vector database.

[0020] Further, step S4 includes the following steps:

[0021] Step S41: for the extracted soil properties, including but not limited to soil organic matter, soil total nitrogen, soil total phosphorus, soil total potassium, and soil pH, based on the process and purpose of the soil survey experiment, using Chain-of-Thought Prompting, construct optimized soil data extraction prompts;

[0022] Step S42, extracting the content of the prompt word;

[0023] Step S43: Use a large language model for fast and accurate soil data extraction. The large language model includes but is not limited to GLM4, qwen-plus, Baichuan2-Turbo, and moonshot-v1-128k. Set the large model Temperature to 0, fuse the soil survey knowledge obtained by RAG in step S3 with the literature text and prompt and input them into the large language model to extract soil data from the literature in jason format.

[0024] Further, in step S42, the content of the prompt words includes: role and task prompt words, which are used to set the large model roles and tasks; prompt words include but are not limited to "You are a soil survey expert, and you need to extract soil data from scientific research literature" to build a virtual role with soil survey experience; soil sample point feature prompt words, which are used to describe the number of soil sampling points in the literature and their relationship with soil properties; prompt words include but are not limited to "The given literature may contain multiple soil sampling points, and each sampling point may include multiple information such as sample point number, sampling time, sampling depth, longitude, latitude, soil attribute name, attribute value, measurement unit, etc."; special data prompt words, which are used to introduce Different forms of geographical locations and soil property values ​​appearing in the introduction documents; prompt words including but not limited to "soil sample point longitude and latitude, which may be in digital format, in degree-minute format (1°1′), or in a range (1°1′~1°1′); in degree-minute-second format (1°1′1″), or in a range (1°1′1″~1°1′1″)", "soil property values ​​may be in numerical form, or they may be the average value of multiple sample points (1.0±1.0)"; return result prompt words, used to explain the form and composition of the returned results, prompt words including but not limited to "please only return results in jason format, and keywords in the results are displayed in Chinese".

[0025] Further, step S5 includes the following steps:

[0026] Step S51: Perform data cleaning and management on the soil data extracted from the literature, including but not limited to latitude and longitude format conversion, soil attribute name conversion, unit conversion, and data conversion, and align the extracted soil data with the historical soil data;

[0027] Step S52: For multiple documents extracted at one time, the data of step S51 are cleaned in batches, the treatment results are saved in the form of a structured table to the local computer, and the cleaned data are integrated with the historical soil data.

[0028] Further, step S6 includes the following steps:

[0029] Step S61: for a specific geographical region in my country, including but not limited to the Huanghuaihai Plain, the black soil in Northeast China, and the red-yellow soil region in the south, using GIS spatial clipping to obtain environmental factor data affecting soil from national data, the environmental factor data including but not limited to topography, climate, land use, and remote sensing monitoring, to form a soil environmental factor data set;

[0030] Step S62: using GIS spatial overlay analysis to obtain soil data within a specific area from updated soil data;

[0031] Step S63: input regional soil data and influencing environmental factor data into the EBKR model to predict regional soil property distribution and update mapping. EBKR parameter settings include but are not limited to the following: local subset size 100, simulation times 100, maximum and minimum adjacent elements 15 and 10 respectively.

[0032] In order to achieve the purpose of the present invention, the present invention also discloses a soil data rapid extraction, integration, mapping and updating system based on a knowledge-enhanced large language model, including a soil survey data text processing module, a soil survey knowledge base module, a knowledge enhancement fusion module, a soil data rapid and accurate extraction module, a soil data integration module, and a soil mapping module.

[0033] Compared with the prior art, the significant progress of the present invention is as follows: 1) The method provided by the present invention for accurately and quickly extracting soil data from literature using an enhanced soil survey knowledge big language model can automatically batch extract soil survey sample points from the latest literature, as well as physical and chemical property information such as soil organic matter, pH, and mechanical composition, replacing manual operations and greatly improving the extraction speed and efficiency; 2) The method for integrating soil data extracted from literature and updating regional soil mapping provided by the present invention can automatically standardize heterogeneous soil data extracted from different literature that have problems such as different names, inconsistent units, and different expressions, automatically update them to the historical soil database, and use the updated data to update regional soil property mapping using the EBKR method.

[0034] In order to more clearly illustrate the functional characteristics and structural parameters of the present invention, further description is given below in conjunction with the accompanying drawings and specific implementation methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0036] Figure 1 It is a technical solution diagram of the present invention;

[0037] Figure 2 It is a technical flow chart of the present invention;

[0038] Figure 3 It is the soil organic matter intelligent extraction prompt word optimized by the present invention;

[0039] Figure 4 This is the soil organic matter intelligent extraction process based on FastGPT of the present invention;

[0040] Figure 5 It is the result of intelligent extraction of soil organic matter of the present invention;

[0041] Figure 6 The present invention integrates and updates the soil organic matter data of the Huang-Huai-Hai Plain;

[0042] Figure 7 This is the soil organic matter data distribution map of the Huanghuaihai Plain updated by the present invention. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0044] High-precision prediction of the spatial distribution of soil properties requires the support of a large amount of field soil survey sampling data, but soil sampling survey and testing and analysis is a basic professional work with high human and financial resources consumption. Therefore, the sample points generally used for soil spatial distribution prediction and mapping are relatively sparse. However, on the other hand, a large number of papers and reports published in recent years contain the latest soil data, but more than 90% of them are unstructured text data, which is difficult to use directly, limiting the reuse of related data and the integration and mining of their value.

[0045] In response to the above problems, a three-stage solution was developed, from rapid and accurate extraction of soil literature data to standardized integration and mapping updates. It can be roughly divided into four major parts: the construction of a soil survey knowledge base, the batch automatic extraction of soil survey information from literature using a large soil model, the management and integration of extracted data, and the prediction and mapping of the spatial distribution of soil attribute data. The detailed technical solution and process are as follows: Figure 1-2 shown.

[0046] Based on the "Chinese Soil Species Directory" formed by the second national soil census, some provincial and regional soil species directory, some soil series directory formed by the land system survey in China, and soil scientific research survey reports, etc., the scanned paper documents are processed using image and text recognition to form text data. Next, the soil knowledge text is split, vectorized, processed and stored to build a soil survey knowledge base that assists the large model.

[0047] In order to solve the problem that the general large language model does not have a deep understanding of soil survey professional knowledge, RAG technology is used to enhance its soil survey knowledge. At the same time, the content and format of soil sampling survey information appearing in scientific research literature are summarized, and based on the domain large model that enhances soil knowledge, the prompt word engineering is used to automatically extract the relevant information of soil sampling points from the literature in a structured manner.

[0048] The structured soil sampling survey information extracted from the paper will be integrated with the existing structured soil data in the study area, and the existing soil data resources will be updated through data alignment (including sampling depth, sampling time, attribute name, attribute value, unit, etc.) and cleaning.

[0049] Based on the updated soil property data, the empirical Bayesian kriging regression method of coupled environmental factors is used to predict soil organic matter and update the spatial distribution map of soil properties (such as organic matter, pH, mechanical composition, etc.) in the relevant areas.

[0050] Example

[0051] Example 1: Extraction and mapping update of soil organic matter data in the Huanghuaihai region

[0052] 1. Regional introduction

[0053] Soil organic matter is an important component of soil and an important indicator of soil fertility. The Huanghuaihai Plain is an important agricultural area in my country. Taking this area as an example, the application of soil organic matter rapid extraction, integration and mapping update technology based on large language models was carried out.

[0054] The Huang-Huai-Hai Plain is the second largest plain in my country, stretching from Yanshan in the north to the north of Dabie Mountain in the south, from Taihang Mountain in the west to northern Jiangsu and Anhui in the southeast. 2), involving 5 provinces and 2 cities including Beijing, Tianjin, Hebei, Shandong, Henan, Jiangsu and Anhui, and is one of the most densely populated areas in China. The Huang-Huai-Hai Plain has a mild climate, including two climate types: warm temperate continental monsoon climate and subtropical monsoon climate. Rain and heat occur at the same time, with dry and cold winters and hot and rainy summers. The average temperature for many years is 8-15℃, and the average precipitation for many years is 500-1000mm. The seasonal distribution is uneven, and the inter-annual differences are large. The terrain of the region is mainly plains, with a relatively flat terrain and an average altitude of less than 50m.

[0055] The Huanghuaihai Plain has abundant natural resources, among which agricultural resources are dominant, and the land is fertile and suitable for agricultural development. Water resources are also relatively abundant, and the main rivers are the Yellow River, Huaihe River, Haihe River, etc. The area has rich soil types, including tidal soil, paddy soil, brown soil, sandy ginger black soil and coastal saline soil, among which tidal soil and brown soil are the main ones. Crops are generally harvested twice a year or three times in two years. The main food crops are wheat, rice, corn, etc., and the cash crops are cotton, rapeseed, peanuts, etc. It is an important cotton, grain and oil production base in my country.

[0056] 2. Data Overview

[0057] (1) Soil data

[0058] In the study area, there are 630 soil surface organic matter data from the 2010s, mainly from soil scientific research project materials such as the China Soil System Survey and the China Farmland Soil Carbon Sequestration Survey.

[0059] (2) Environmental data

[0060] For environmental data that affect the spatial distribution of soil organic matter, the main data elements used are topography, climate, land use, etc. Among them, the topographic data is the Shuttle Rader Topography Mission and Digital Elevation Model (SRTM DEM) with a spatial resolution of 30m provided by NASA, and the slope, slope aspect, terrain moisture index, terrain roughness and other data generated from this DEM data. The climate data is generated by interpolation of meteorological observation data from 1960s to 2010s, including multi-year average temperature, average rainfall, annual evaporation, etc., with a spatial resolution of 1km. Land use uses the remote sensing images of the US Landsat satellite as the main information source, and the multi-period land use data elements (1980s-2020s) are completed through manual visual interpretation.

[0061] (3) Journal articles

[0062] Acta Pedologica Sinica and Soils are high-level academic journals in the field of soil science in my country. In recent years, they have published a large number of research papers based on soil survey data. Download and obtain nearly 1,500 papers published in the two journals from 2020 to 2024 for intelligent extraction of soil organic matter information.

[0063] 3. Intelligent extraction of soil organic matter data from literature

[0064] (1) Construction of soil survey knowledge base

[0065] For paper scanned materials such as "Chinese Soil Species" and "Soil Taxonomy", OCR technology is used to identify them as text data. For text materials of soil survey knowledge, text noise removal is performed, which includes but is not limited to special format processing, special character elimination, stop word processing, HTML tag removal, header and footer removal, to form soil survey knowledge text.

[0066] Based on the FastGPT framework, the input processed soil survey knowledge text is segmented by segmentation + direct block method, and the segment separator is customized; at the same time, the LLM text vectorization method of embedding-2 is used to embed the segmented text, and the vectorized data is stored in the vector database, thereby forming a soil survey knowledge base.

[0067] (2) Construction of soil intelligent extraction prompt words

[0068] Based on the process and purpose of the soil survey experiment, we use the thinking chain prompts to construct prompts for optimizing soil organic matter data extraction. The prompts for soil organic matter extraction mainly include the following four parts:

[0069] Role and task prompts, used to set large model roles and tasks;

[0070] Soil sampling point feature prompt words are used to describe the number of soil sampling points in the literature and their relationship with soil properties.

[0071] Sexual relations;

[0072] Special data prompt words are used to introduce geographical locations and soil property values ​​that appear in the literature.

[0073] Same form;

[0074] The returned result prompt words are used to explain the form and composition of the returned results.

[0075] Based on the above principles, a soil organic matter intelligent extraction prompt word system suitable for the journals "Acta Pedologica Sinica" and "Soil" was constructed. Figure 3 ).

[0076] (3) Intelligent extraction of soil organic matter

[0077] The GLM4 large language model is used to quickly and accurately extract soil data, using the locally deployed open source FastGPT framework ( Figure 4 ), set the large model Temperature to 0. Use RAG technology to search from the soil survey knowledge base, generate background knowledge to assist the large language model in soil data extraction, and fuse it with the input document text and prompt to input the large language model, obtain the soil data extracted from the document in jason format, and save it on the local computer.

[0078] (4) Soil organic matter extraction capacity

[0079] Using the constructed soil organic matter intelligent extraction framework, soil organic matter information in the literature was extracted, and the results showed that: Figure 5 In A, it can be seen that the proposed extraction scheme can extract the text information and table data of soil sampling description in association; Figure 5 In B, it can be seen that the extraction scheme can also associate the table describing the geographical location of soil sampling with the table of physical and chemical properties for extraction.

[0080] (5) Soil organic matter extraction data cleaning and integration

[0081] More than 1,500 papers published in recent years in Soils and Acta Pedologica Sinica were extracted, and it was found that effective soil organic matter information could be extracted from about 20%. The extracted soil organic matter data was cleaned and managed, including but not limited to latitude and longitude format conversion, soil attribute name conversion, unit conversion, and data conversion. Finally, the standardized extracted soil organic matter data after management was aligned with the historical soil data to achieve the integration of historical and extracted soil organic matter data.

[0082] (6) Soil organic matter extraction data cleaning and integration

[0083] Using GIS spatial overlay analysis, we obtained soil organic matter data in the red-yellow soil area from the updated soil data, and obtained nearly 100 valid soil organic matter data in the Huanghuaihai region. The integrated and updated soil organic matter data in the red-yellow soil area reached more than 700 ( Figure 6 ). At the same time, GIS spatial clipping was used to obtain environmental factor data affecting soil from national data, including topography, climate, land use, etc. Finally, combined with soil organic matter survey data and data on environmental factors affecting soil organic matter, the EBK regression method was used to update the soil organic matter distribution map ( Figure 7 ).

[0084] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0085] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A soil data rapid extraction, integration and mapping update method based on knowledge-enhanced large language model, characterized in that: The following steps are involved: Step S1, text processing of soil survey data, processing paper scans and electronic texts of soil survey knowledge materials to form text data that can be recognized by computers; Step S2, segmenting and storing the text of soil survey related knowledge to form a vectorized knowledge base; Step S3: Knowledge enhancement fusion: using RAG technology to search from the soil survey knowledge base, generate background knowledge to assist the large language model in soil data extraction, and fuse it with the input literature text; Step S4: soil data is extracted quickly and accurately. Based on the integrated background knowledge and literature text, the optimization prompt is constructed using the thinking chain prompt method to clarify the thinking path of extracting soil data with a large model, and to extract soil data from literature quickly and accurately. Step S5, soil data integration, performing redundancy removal, transformation, unit conversion, coordinate conversion, and numerical conversion operations on the extracted heterogeneous soil data to form standardized and unified structured soil data, and integrating it with historical soil data; Step S6, soil mapping, based on the integration results of extracted soil data and historical soil data, coupled with environmental factor data affecting soil, including but not limited to topography, climate, land use, and remote sensing monitoring; using the empirical Bayesian Kriging regression EBKR method to update the mapping of the region.

2. According to claim 1, a soil data rapid extraction, integration and mapping update method based on knowledge-enhanced large language model is characterized in that: Step S1 includes the following steps: Step S11, text processing of soil survey data, which is divided into two processing processes according to the type of input data. If the input data is paper scanned data, it is recognized as text data using OCR technology, and the text data is input into step S12 for processing; Step S12: If the input is text data, perform text noise removal processing, which includes but is not limited to special format processing, special character elimination, stop word processing, HTML tag removal, and header and footer removal to form a soil survey knowledge text.

3. According to claim 1, a soil data rapid extraction, integration, mapping and updating method based on knowledge-enhanced large language model is characterized in that: Step S2 includes the following steps: Step S21, constructing a knowledge base for the input high-quality soil survey knowledge text through the FastGPT framework; Step S22, the soil survey knowledge text is segmented according to the segmentation + direct block method, the ideal block length is set to 1024, and the segment separator is customized; Step S23: Use the LLM text vectorization method, including but not limited to embedding-2 and Baichuan-Text-Embedding, to embed the divided text, and store the vectorized data in a vector database.

4. The soil data rapid extraction, integration and mapping updating method based on knowledge-enhanced large language model according to claim 1 is characterized in that: Step S4 includes the following steps: Step S41: for the extracted soil properties, including but not limited to soil organic matter, soil total nitrogen, soil total phosphorus, soil total potassium, and soil pH, based on the process and purpose of the soil survey experiment, using the thinking chain prompt words, construct optimized soil data extraction prompt words; Step S42, extracting the content of the prompt word; Step S43: Use a large language model for fast and accurate soil data extraction. The large language model includes but is not limited to GLM4, qwen-plus, Baichuan2-Turbo, and moonshot-v1-128k. Set the large model Temperature to 0, fuse the soil survey knowledge obtained by RAG in step S3 with the literature text and prompt and input them into the large language model to extract soil data from the literature in jason format.

5. The soil data rapid extraction, integration and mapping updating method based on knowledge-enhanced large language model according to claim 4 is characterized in that: In step S42, the content of the prompt words includes: role and task prompt words, which are used to set the large model roles and tasks; soil sample point feature prompt words, which are used to describe the number of soil sampling points in the literature and their relationship with soil properties; special data prompt words, which are used to introduce the different forms of geographical locations and soil attribute values ​​that appear in the literature; return result prompt words, which are used to explain the form and composition of the returned results.

6. The soil data rapid extraction, integration and mapping updating method based on knowledge-enhanced large language model according to claim 1 is characterized in that: Step S5 includes the following steps: Step S51: Perform data cleaning and management on the soil data extracted from the literature, including but not limited to latitude and longitude format conversion, soil attribute name conversion, unit conversion, and data conversion, and align the extracted soil data with the historical soil data; Step S52: For multiple documents extracted at one time, the data of step S51 are cleaned in batches, the treatment results are saved in the form of a structured table to the local computer, and the cleaned data are integrated with the historical soil data.

7. The soil data rapid extraction, integration and mapping updating method based on knowledge-enhanced large language model according to claim 1 is characterized in that: Step S6 includes the following steps: Step S61, using GIS spatial clipping to obtain environmental factor data affecting soil from national data, the environmental factor data including but not limited to topography, climate, land use, and remote sensing monitoring, to form a soil environmental factor data set; Step S62: using GIS spatial overlay analysis to obtain soil data within a specific area from updated soil data; Step S63: input regional soil data and influencing environmental factor data into the EBKR model to predict regional soil property distribution and update mapping. EBKR parameter settings include but are not limited to the following: local subset size 100, simulation times 100, maximum and minimum adjacent elements 15 and 10 respectively.

8. A soil data rapid extraction, integration, mapping and updating system based on knowledge-enhanced large language model, the system is based on a soil data rapid extraction, integration, mapping and updating method based on knowledge-enhanced large language model according to any one of claims 1-7, characterized in that: It includes soil survey data text processing module, soil survey knowledge base module, knowledge enhancement fusion module, soil data fast and accurate extraction module, soil data integration module and soil mapping module.

Images