Provincial domain level remote sensing AI big data natural resource intelligent monitoring system

Abstract

The invention provides a provincial domain level remote sensing AI big data natural resource intelligent monitoring system. The system comprises a data acquisition module, a data processing module, an intelligent interpretation module, a result verification module, a management application module and a communication interface module. The intelligent interpretation module integrates Segform semantic segmentation and a BIT change detection model, large-scale parallel interpretation is achieved by combining Redis cache management, celery task scheduling and MySQL process control under a GPU cluster architecture, batch image interpretation in a provincial domain range can be completed in a day-level period, linkage of interior work recheck and field work check is established through the achievement verification module, and large-scale parallel interpretation is achieved. And the management application module realizes the dynamic updating, thematic storage and cross-department sharing of the result, can quickly generate a change pattern spot distribution map, a cultivated land non-agrochemical and non-grain statistical table and a law enforcement check result, greatly shortens the interpretation period, and meets the requirement of dynamic supervision.

Description

Technical Field

[0001] This invention relates to the field of remote sensing monitoring technology, and in particular to a provincial-level remote sensing AI big data intelligent monitoring system for natural resources. Background Technology

[0002] With increasingly stringent requirements for natural resource protection and farmland use control, provincial-level natural resource regulatory departments need to conduct routine monitoring of farmland conversion to non-agricultural uses, farmland conversion to non-grain uses, and illegal land occupation for construction. However, existing technologies mainly rely on manual visual interpretation or semi-automatic interpretation of remote sensing satellite imagery, which has the following shortcomings: I. Natural resource monitoring involves multiple sources of data, including domestic public welfare satellites, commercial remote sensing satellites, and data from land change surveys over the years. These data differ in spatial resolution, spectral characteristics, and imaging conditions, making it difficult to directly overlay and compare images acquired by different sensors at different times. This results in insufficient data consistency and comparability, which in turn affects the interpretation accuracy of deep learning models. Second, current provincial monitoring relies heavily on manual visual interpretation or semi-automatic tools, which has a limited processing scale and often takes several weeks to complete a province-wide monitoring. This cannot meet the timeliness requirements of dynamic supervision. At the same time, the interpretation results are mostly stored in isolated files, lacking cross-departmental sharing, making it difficult to provide timely decision support for departments such as natural resources, agriculture, and forestry and grassland. It is also not conducive to the implementation of the business objectives of "early detection, early prevention, and strict investigation and punishment". To address this, a provincial-level remote sensing AI big data intelligent monitoring system for natural resources is proposed. Summary of the Invention

[0003] In view of this, the present invention provides a provincial-level remote sensing AI big data intelligent monitoring system for natural resources to solve or alleviate the technical problems existing in the prior art, and at least provides a beneficial option.

[0004] The technical solution of this invention is implemented as follows: a provincial-level remote sensing AI big data intelligent monitoring system for natural resources, comprising... The data acquisition module is used to receive multi-source remote sensing image data from the national and provincial levels, and communicate with the provincial natural resources big data center to perform geometric correction, tone matching and filtering noise reduction, realize the fusion of data from different resolutions and different sensors, and build a high-resolution sample library covering different regions, seasons and elements based on a combination of manual delineation and automatic annotation. The data acquisition module includes a multi-source data access module and a sample construction module; The multi-source data access module is used to access domestic public satellite imagery, commercial remote sensing imagery, historical land change survey results and geographic national conditions data, and organizes and indexes them through a unified spatial reference system. The sample construction module is communicatively connected to the multi-source data access module. It is used to generate a sample dataset covering typical targets such as construction land, roads, ponds and photovoltaic land by combining manual delineation and deep learning pre-annotation methods. The diversity and representativeness of the samples are ensured through regional division, seasonal coverage and scale control.

[0005] The data processing module is communicatively connected to the data acquisition module and is used to extract features from remote sensing images with the support of the sample library, construct a multi-scale feature pyramid, and perform spectral consistency processing on the fused images to form a high spatiotemporal resolution image database adapted to deep learning interpretation. The data processing module includes an image fusion module and a feature enhancement module; The image fusion module is used to perform geometric correction, spectral normalization and image stitching on remote sensing data from different sensors to ensure the consistency of multi-source heterogeneous images in terms of spatial scale and spectral characteristics. The feature enhancement module is communicatively connected to the image fusion module and is used to hierarchically represent image features using multi-scale pyramid pooling and upsampling strategies to enhance the spectral and boundary features of different ground objects.

[0006] The intelligent interpretation module is connected to the data processing module and is used to call the convolutional neural network and the Transformer network. Based on the Segformer semantic segmentation and BIT change detection model, it realizes the automatic extraction and change detection of construction land, roads, ponds and photovoltaic land targets. It also performs multi-node parallel computing through GPU cluster and Redis+celery+MySQL task scheduling mechanism to output land type patches, change patches and suspected illegal land occupation information. The intelligent interpretation module includes a parallel computing module, a semantic segmentation module, and a change detection module; The parallel computing module relies on a GPU cluster, executes parallel task scheduling through a CPU-GPU collaborative mechanism, and uses Redis cache management, Celery task queues and MySQL database to control the interpretation process, thereby realizing the distributed interpretation of massive remote sensing images. The parallel computing module can dynamically allocate tasks according to the node status, avoid single node overload, improve interpretation throughput and stability, and achieve daily processing cycle in large-scale provincial-level monitoring. The semantic segmentation module is based on the Segformer deep neural network and uses a multi-scale encoder structure to achieve multi-class extraction of construction land, traffic roads, farmland, urban green space and photovoltaic land; The change detection module is communicatively connected to the semantic segmentation module. Based on the BIT model, it performs difference detection on images of different time phases, outputs regional-level change patches, and combines a boundary optimization algorithm to correct the boundaries of ground features, which is used for high-precision identification of "non-agriculturalization" and "non-grainization" of cultivated land.

[0007] The results verification module is interactively connected to the intelligent interpretation module. It is used to conduct internal review of the interpretation results on the interactive interface, and to conduct on-site verification based on the field data collected by the mobile terminal. The verification results are compared and corrected with the business database to generate reliable monitoring results. The results verification module includes an internal review module and an external verification module; The internal review module provides an interactive interface that supports manual drawing, comparison and boundary correction of the interpreted map features; The field verification module is communicatively connected to the office re-inspection module. It is used to verify suspected map features by using on-site GPS positioning and image information collected by mobile terminals, and to send the field-collected data back to the system database for comparison.

[0008] The management application module is connected to the results verification module and is used to classify, store, statistically analyze and visualize the corrected monitoring results. It also shares the results with the natural resources, agriculture and forestry departments through the communication interface to realize the business application of satellite imagery law enforcement and dynamic monitoring of the "non-agricultural" and "non-grain" use of arable land. The management application module includes a dynamic update module and a results sharing module; The dynamic update module is used to automatically trigger the data processing and interpretation process when new remote sensing image data is accessed, so as to realize the dynamic update of monitoring results; The results sharing module is communicatively connected to the dynamic update module and is used to provide monitoring results to the departments of natural resources, agriculture and forestry in the form of a thematic database, supporting cross-departmental collaborative supervision and information sharing. Further includes a results report generation module; The results report generation module can automatically generate report files based on monitoring results. The report includes a distribution map of changed patches, a statistical table of farmland conversion to non-agricultural and non-grain uses, and satellite imagery enforcement verification results, which are presented in the form of charts and visual maps.

[0009] The communication interface module is connected to the management application module and is used to interconnect with the provincial big data center and business platform through the government intranet or data exchange platform to realize online reporting of monitoring results, cross-departmental access, and historical result tracing.

[0010] The embodiments of the present invention have the following advantages due to the adoption of the above technical solutions: I. The data acquisition module of this invention integrates multi-source data access and sample construction, uniformly performs geometric correction, tone matching, and filtering and noise reduction, fuses images from different resolutions and sensors, and constructs a high-resolution sample library covering typical targets such as construction land, roads, ponds, and photovoltaic land. By combining regional division, seasonal coverage, and scale control strategies, the representativeness and diversity of the samples are ensured, enabling the deep learning model to run under uniform scale and spectral conditions during interpretation, thereby improving the consistency of image input and the interpretation accuracy of the model.

[0011] Second, the intelligent interpretation module of this invention integrates the Segformer semantic segmentation and BIT change detection model, and achieves large-scale parallel interpretation under a GPU cluster architecture by combining Redis cache management, Celery task scheduling and MySQL process control. It can complete batch image interpretation of the province within a daily cycle. At the same time, the results verification module establishes a linkage between internal review and external verification to ensure the reliability of interpretation results. The management application module realizes dynamic updating, thematic storage and cross-departmental sharing of results. It can quickly generate change patch distribution maps, statistics tables of farmland non-agriculturalization and non-grain conversion, and law enforcement verification results, which significantly shortens the interpretation cycle and promotes the real-time application of monitoring results in multiple departments such as natural resources, agriculture and forestry and grassland, meeting the needs of dynamic supervision.

[0012] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the invention will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a system module block diagram of the present invention. Detailed Implementation

[0015] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0016] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0017] like Figure 1 As shown, this embodiment of the invention provides a provincial-level remote sensing AI big data intelligent monitoring system for natural resources, the system including a data acquisition module; The data acquisition module communicates with national and provincial natural resources big data centers to receive multi-source remote sensing image data. The data sources include domestic public welfare satellites, commercial remote sensing satellites, annual land change survey results, and geographic national conditions census data. The data acquisition module consists of a multi-source data access module and a sample construction module. The multi-source data access module is responsible for unifying the spatial reference system, performing geometric correction, tone matching, and filtering and noise reduction to ensure the comparability of data from different resolutions and sensors in terms of spatiotemporal consistency. The sample construction module combines manual delineation and deep learning pre-annotation to build sample datasets covering typical targets such as construction land, roads, ponds, and photovoltaic land. Through regional division, seasonal coverage, and scale control, it forms a diverse and representative high-resolution sample library.

[0018] After data acquisition is completed, the received multi-source data is further processed by the data processing module; The data processing module consists of an image fusion module and a feature enhancement module. The image fusion module performs spectral normalization and seamless stitching on the input remote sensing images to form a fused image that is consistent in spatial scale and spectral features, eliminating mismatches caused by sensor differences or imaging conditions. The feature enhancement module adopts a multi-scale feature pyramid structure and combines pyramid pooling and upsampling strategies to achieve multi-level feature enhancement of the images, making the spectral differences, texture details and boundary contours of different ground objects clearer. The feature images output by the data processing module have higher spatiotemporal resolution and stronger deep learning adaptability, and can be directly entered into the subsequent intelligent interpretation module.

[0019] Furthermore, the intelligent interpretation module is communicatively connected to the data processing module, including a parallel computing module, a semantic segmentation module, and a change detection module; The parallel computing module relies on a GPU cluster for operation, employing a CPU-GPU collaborative mechanism for distributed task partitioning and scheduling. Redis is used for cache management, Celery acts as a task queue to control multi-node execution, and a MySQL database is responsible for registering the interpretation process and results, ensuring the traceability and high throughput of the entire interpretation process. The semantic segmentation module, based on the Segformer deep neural network, uses a multi-scale encoder structure to automatically extract features such as construction land, roads, farmland, urban green space, and photovoltaic land, and outputs structured land use patches. The change detection module uses the BIT dual-temporal change detection model to compare and analyze images from different times, generating regional-level change patches. Boundary optimization algorithms are used to refine and smooth the boundaries, achieving high-precision identification of "non-agricultural" and "non-grain" farmland. Through the collaborative operation of the intelligent interpretation module, the module can complete the batch interpretation of massive amounts of remote sensing images within a province within a daily cycle.

[0020] After the interpretation results are generated, the data enters the results verification module to ensure the authenticity and reliability of the data. The results verification module includes an internal review module and a field verification module. The two work together through an interactive interface. The internal review module provides an interactive visual interface, allowing reviewers to manually draw, compare, and correct the boundaries of automatically generated map features. The field verification module completes on-site verification based on mobile terminals, collects GPS positioning and real-time images, and transmits the data back to the system for comparison and joint verification with the internal results, forming the final results after verification or correction. Once the results are verified, the revised results are entered into the management application module. This module communicates with the results verification module and mainly includes a dynamic update module, a results sharing module, and a results report generation module. The dynamic update module automatically triggers the entire data processing and interpretation process when new remote sensing images are accessed, ensuring the results database remains up-to-date and dynamically updated. The results sharing module categorizes and stores the verified results in the form of a thematic database, and displays and statistically analyzes them in a visual manner. It supports departments such as natural resources, agriculture, and forestry and grassland to access the results as needed, enabling cross-departmental collaborative supervision and information sharing. The results report generation module can automatically generate thematic reports based on the verified results, including distribution maps of changed land parcels, statistical tables of farmland conversion to non-agricultural and non-grain uses, and satellite imagery enforcement verification results. These reports are presented in the form of charts and visual maps, facilitating law enforcement supervision and decision analysis by government departments.

[0021] In the final data transfer stage, the communication interface module connects with the management application module and is interconnected with the provincial big data center and related business platforms through the government intranet or provincial data exchange platform. This module enables online reporting of monitoring results, cross-departmental access, and historical result traceability, ensuring that monitoring results can be shared and circulated in multiple business systems, supporting various application scenarios such as farmland protection, satellite imagery enforcement, and resource planning.

[0022] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in the present invention, and these should all be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A provincial level remote sensing AI big data natural resource intelligent monitoring system, characterized in that: Comprising The data acquisition module is used for receiving multi-source remote sensing image data issued from the national and provincial levels, and is in communication connection with the provincial natural resource big data center, executes geometric correction, tone matching and filtering denoising, realizes the fusion of different resolutions and different sensor data, and constructs a high-resolution sample library covering different regions, seasons and elements based on the combination of artificial sketching and automatic labeling; The data processing module is in communication connection with the data acquisition module, and is used for feature extraction of remote sensing images under the support of the sample library, construction of a multi-scale feature pyramid, and spectral consistency processing of fused images to form a high spatio-temporal resolution image database suitable for deep learning interpretation; The intelligent interpretation module is in communication connection with the data processing module, and is used for calling convolutional neural networks and Transformer networks, realizing automatic extraction and change detection of construction land, roads, ponds and photovoltaic land targets based on Segformer semantic segmentation and BIT change detection models, and outputting land class patches, change patches and suspected illegal land occupation information through GPU cluster and Redis+celery+MySQL task scheduling mechanism for multi-node parallel computing; The achievement verification module is in interactive connection with the intelligent interpretation module, and is used for internal review of the interpretation results on the interactive interface, and field verification based on the external data collected by the mobile terminal, comparison and correction of the verification results with the business database, and generation of reliable monitoring achievements; The management application module is in communication connection with the achievement verification module, and is used for classified storage, statistical analysis and visual display of the corrected monitoring achievements, and sharing of the achievements with the natural resources, agriculture and forestry departments through the communication interface to realize the business application of satellite image law enforcement, dynamic monitoring of "non-agriculturalization" and "non-foodization" of cultivated land.

2. The provincial level remote sensing AI big data natural resource intelligent monitoring system according to claim 1, characterized in that: The data acquisition module comprises a multi-source data access module and a sample construction module; The multi-source data access module is used for accessing domestic public satellite images, commercial remote sensing images, historical land change survey results and geographic national conditions data, and organizing and indexing through a unified spatial reference system; The sample construction module is in communication connection with the multi-source data access module, and is used for generating sample data sets of typical targets covering construction land, roads, ponds and photovoltaic land by combining artificial sketching and deep learning pre-labeling methods, and ensuring sample diversity and representativeness through regional division, seasonal coverage and scale control. 3.The natural resource intelligent monitoring system of provincial level remote sensing AI big data according to claim 1, characterized in that: The data processing module comprises an image fusion module and a feature enhancement module; The image fusion module is used for executing geometric correction, spectral normalization and image stitching on remote sensing data of different sensors to ensure the consistency of multi-source heterogeneous images in spatial scale and spectral features; The feature enhancement module is in communication connection with the image fusion module, and is used for hierarchical expression of image features by using multi-scale pyramid pooling and up-sampling strategies to enhance the spectral and boundary features of different ground object targets.

4. The provincial level remote sensing AI big data natural resource intelligent monitoring system according to claim 1, characterized in that: The intelligent interpretation module comprises a parallel computing module; The parallel computing module relies on a GPU cluster, performs parallel task scheduling through a CPU-GPU cooperative mechanism, and utilizes Redis cache management, a celery task queue, and a MySQL database to control the interpretation process, thereby realizing distributed interpretation of massive remote sensing images. The parallel computing module can dynamically allocate tasks according to node states, avoid single-node overload, improve interpretation throughput and stability, and realize daily processing cycles in provincial large-scale monitoring.

5. The provincial level remote sensing AI big data natural resource intelligent monitoring system according to claim 1, characterized in that: The intelligent interpretation module further includes a semantic segmentation module and a change detection module. The semantic segmentation module is based on a Segformer deep neural network and utilizes a multi-scale encoder structure to realize multi-classification extraction of construction land, traffic roads, farmland, urban green land, and photovoltaic land. The change detection module is in communication connection with the semantic segmentation module, performs difference detection on different temporal images based on a BIT model, outputs regional-level change polygons, and corrects feature boundaries by combining a boundary optimization algorithm, thereby achieving high-precision identification of "non-agriculturalization" and "non-foodization" of farmland.

6. The provincial level remote sensing AI big data natural resource intelligent monitoring system according to claim 1, characterized in that: The achievement verification module includes an internal review module and an external verification module. The internal review module provides an interactive interface and supports manual sketching, comparison, and boundary correction of the polygons generated by interpretation. The external verification module is in communication connection with the internal review module, is used for verifying suspected polygons by collecting field GPS positioning and image information through a mobile terminal, and returns the external collection data to the system database for comparison.

7. The provincial level remote sensing AI big data natural resource intelligent monitoring system according to claim 1, characterized in that: The management application module includes a dynamic update module and an achievement sharing module. The dynamic update module is used for automatically triggering data processing and interpretation processes when new remote sensing image data is accessed, thereby realizing dynamic update of monitoring achievements. The achievement sharing module is in communication connection with the dynamic update module, is used for providing monitoring achievements in the form of a thematic database to departments of natural resources, agriculture, and forestry for calling, and supports cross-departmental collaborative supervision and information sharing. 8.The provincial level remote sensing AI big data natural resource intelligent monitoring system according to claim 1, characterized in that: The management application module further includes an achievement report generation module. The achievement report generation module can automatically generate a report file according to monitoring achievements, the report includes a change polygon distribution map, a non-agriculturalization and non-foodization statistical table of farmland, and a satellite image law enforcement verification result, and is displayed in the form of charts and visual maps. 9.The natural resource intelligent monitoring system of provincial level remote sensing AI big data according to claim 1, characterized in that: The system further includes a communication interface module, which is connected with the management application module, is used for interconnecting with a provincial big data center and a business platform through a government affair special network or a data exchange platform, realizes online reporting, cross-departmental calling, and historical achievement tracing of monitoring achievements.

Images