Large-size image target linkage strict judgment system and method based on identity responsibility

Abstract

The invention belongs to the technical field of intelligent processing and application of remote sensing images, and discloses a large-size image target linkage strict judgment system and method based on identity responsibilities, and an identity responsibility matrix engine module realizes efficient allocation of resources and tasks in the system, and further combines with data analysis, so as to realize the target linkage strict judgment. Accurate user capability evaluation and dynamic identity responsibility adjustment are provided; the intelligent linkage task scheduling center module realizes intelligent matching based on identity responsibility and multi-mode intelligent linkage task distribution; the multi-user linkage real-time interpretation module realizes large-size image display, multi-user linkage target interpretation and real-time synchronization of target interpretation results; the intelligent interpretation auxiliary module utilizes a target historical library, a target knowledge library, a target situation library, an image searching engine and similar target extraction to assist in completing target interpretation. According to the method, the performance of a large-size image target interpretation system is remarkably improved, the accuracy and practicability of remote sensing image analysis are improved, and application of the remote sensing technology in more actual scenes is promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent processing and application of remote sensing images, and in particular relates to a large-size image target linkage strict judgment system and method based on identity responsibility. Background Art

[0002] With the rapid development of remote sensing technology, the size of high-resolution remote sensing images has been increasing. These images contain rich geographic information and can be applied to many fields such as urban planning, disaster monitoring, and resource management. However, the amount of data in large-size remote sensing images is huge, and traditional manual interpretation methods can no longer meet the needs of modern remote sensing image processing. Although machine learning-based methods have made significant progress in target detection and classification in recent years, there are still some problems that need to be solved in practical applications.

[0003] Most current remote sensing image processing systems rely mainly on the physical characteristics of the target (such as shape, texture, color, etc.) for identification and classification, while ignoring the differences in responsibilities of different users and their specific focus on target feature types. For example, in the same remote sensing image, a geologist may be more concerned about the changes in geological structure, while an urban planner may focus on the distribution and changes of buildings. Therefore, how to optimize the extraction and analysis process of target features based on the division of responsibilities of users is a direction worth exploring.

[0004] In addition, when processing large-scale remote sensing images, how to effectively segment the image, quickly locate the target area, and perform efficient feature analysis is also a challenge. Traditional image processing methods often require processing the entire image, which is not only computationally intensive, but also difficult to adapt to dynamically changing scenes. Summary of the invention

[0005] The purpose of the present invention is to provide a large-size image target linkage strict judgment system and method based on identity and responsibility. The system of the present invention not only takes into account the physical characteristics of the target itself, but also combines the division of responsibilities of different users. It can customize the processing of target element types in remote sensing images according to user needs, significantly improves the performance of large-size image target interpretation systems, improves the accuracy and practicality of remote sensing image analysis, and promotes the application of remote sensing technology in more practical scenarios.

[0006] The technical solution adopted by the present invention is a large-size image target linkage strict judgment system based on identity and responsibility, including an identity and responsibility matrix engine module, an intelligent linkage task scheduling center module, a multi-user linkage real-time judgment module, and an intelligent judgment auxiliary module, wherein:

[0007] The identity responsibility matrix engine module realizes the efficient allocation of resources and tasks within the system, and further combines data analysis to provide accurate user capability assessment and dynamic identity responsibility adjustment;

[0008] The intelligent linkage task scheduling hub module realizes intelligent matching based on identity responsibilities and multi-mode intelligent linkage task allocation;

[0009] The multi-user linkage real-time interpretation module realizes large-size image display, multi-user linkage target interpretation, and real-time synchronization of target interpretation results;

[0010] The intelligent interpretation assistance module uses the target history library, target knowledge base, target situation library, image search engine and similar target extraction to assist in target interpretation.

[0011] Furthermore, the identity responsibility matrix engine module includes user management, dynamic management of identity responsibilities, a voting mechanism based on identity responsibilities, and a feedback mechanism based on identity responsibilities; user management implements the functions of adding users, deleting users, modifying users, and querying users; dynamic management of identity responsibilities implements flexible allocation and real-time optimization of user identity responsibilities; the voting mechanism based on identity responsibilities means that when a reading user encounters an uncertain reading target, he / she initiates a vote on identity responsibilities of the same level. If the ratio of the number of votes for a certain attribute value of the target to the number of users participating in the vote exceeds 1 / 2, the target attribute value is accepted as the reading value of the target attribute; if a certain attribute value of the target is If the ratio of the voting votes of the attribute value to the number of users participating in the vote does not exceed 1 / 2, feedback will be given to the identity responsibility at a higher level; the feedback mechanism based on identity responsibility means that when the interpretation user still cannot determine the attributes of the interpretation target after voting based on the voting mechanism of identity responsibility, feedback will be given to the identity responsibility at a higher level, and a interpretation user with a higher level identity responsibility will be randomly obtained to interpret the target; when the interpretation user with the identity responsibility at a higher level cannot determine the target attributes, a vote of the identity responsibility at the same level will be initiated; when the voting of the identity responsibility at the same level still cannot determine the target attributes, feedback will continue to be given to the interpretation user with the identity responsibility at a higher level.

[0012] Furthermore, dynamic management of identity responsibilities includes static management and dynamic adjustment of identity responsibilities; static management of identity responsibilities manages identity responsibilities through adding, deleting, modifying, and querying, and supports many-to-many association of identity responsibilities with users; dynamic adjustment of identity responsibilities refers to tracking the completion of target interpretation by users under different identity responsibilities, evaluating the target interpretation ability of users under different identity responsibilities through hierarchical analysis method, generating detailed performance reports, and dynamically suggesting adjustments to identity responsibilities based on analysis results and task requirements.

[0013] Furthermore, the intelligent linkage task scheduling central module includes intelligent matching based on identity responsibilities and multi-mode intelligent linkage task allocation; intelligent matching based on identity responsibilities refers to selecting the judgment target, and arranging the users under a certain judgment target who meet the requirements in ascending order according to the user analysis results based on identity responsibilities and the urgency of the task; multi-mode intelligent linkage task allocation realizes task allocation under three linkage modes: target linkage, regional linkage and intelligent linkage.

[0014] Furthermore, the interpretation target is selected based on the intelligent matching of identity responsibilities. According to the user analysis results based on identity responsibilities and the urgency of the task, the users under a certain interpretation target that meet the requirements are arranged in ascending order, where:

[0015] The urgency of tasks is divided into general, urgent, and urgent;

[0016] Define an array M to represent the average reading time of each user: M = [M 1 ,M 2 ,…,M i ];

[0017] Define an array N to represent the average accuracy of each user: N = [N 1 ,N 2 ,…,N i ];

[0018] Define an array Q to represent the number of unfinished tasks for each user: Q = [Q 1 ,Q 2 ,…,Q i ];

[0019] Define an array P to represent the number of completed interpretation targets for each user: P = [P 1 ,P 2 ,…,P i ];

[0020] Define an array R to represent the number of independent interpretation targets completed by each user: R = [R 1 ,R 2 ,…,R i ];

[0021] Where i is the number of users;

[0022] make represents the index value of the jth element in array X after X is sorted in ascending order. Let represents the index value of the jth element in array X after X is sorted in descending order. The comprehensive score of each user can be calculated by the following formula:

[0023]

[0024] Among them, ω M ,ω N ,ω Q ,ω P ,ω R They are the weights of average interpretation time, average interpretation accuracy, unfinished tasks, number of completed interpretation targets, and number of independently completed interpretation targets;

[0025] According to S j Sort all users in ascending order based on the values ​​of to obtain the final sorted array S:

[0026] S=sort(S 1 ,S 2 ,...,S i );

[0027] Among them, when the urgency of the task is normal:

[0028] ω M =100,ω N =1000,ω Q =10000,ω P =10,ω R =1;

[0029] Among them, when the urgency of the task is urgent:

[0030] ω M =1000,ω N =10000,ω Q =1,ω P =100,ω R =10;

[0031] Among them, when the urgency of the task is extremely urgent:

[0032] ω M =10000,ω N =1000,ω Q =1,ω P =10,ω R =100.

[0033] Furthermore, the multi-mode intelligent linkage task allocation realizes task allocation under three linkage modes: target linkage, regional linkage and intelligent linkage, among which:

[0034] Task allocation in target linkage mode: According to the list of interpretation users output by intelligent matching based on identity responsibilities, the best interpretation user for each interpretation target is selected, and the corresponding target interpretation task is assigned to the interpretation user;

[0035] Task allocation in regional linkage mode: According to the list of users output by intelligent matching based on identity responsibilities, select the number of users n who will participate in target interpretation, divide the entire large-size image into task areas corresponding to the number n, and allocate the task areas to the users who will participate in target interpretation;

[0036] Task allocation under intelligent linkage mode: According to the list of interpretation users output by intelligent matching based on identity responsibilities, the list of interpretation targets that can be used in the target linkage mode is screened out through the pre-set user analysis threshold based on identity responsibilities, and the corresponding optimal target interpretation personnel are calculated. For the remaining interpretation targets and personnel that cannot be allocated through the target linkage mode, task allocation is carried out according to the regional linkage mode, and the task area is evenly allocated to the corresponding interpretation users according to the number of remaining users who can participate in the interpretation.

[0037] Furthermore, the multi-user linkage real-time interpretation module realizes large-size image display, multi-user linkage target interpretation, and real-time synchronization of target interpretation results, among which:

[0038] Large-size image display: By converting large-size images into Cloud optimized GeoTIFF files and storing them in Minio object storage, the standard S3 protocol is provided for loading and displaying them on the browser. It supports free zooming, panning, rolling shutters, gradients, and simultaneous display of multiple image windows. It also switches between global display and task area display according to different interpretation requirements.

[0039] Multi-user linkage target interpretation: supports real-time collaboration among multiple users to annotate and fill in attributes of the same image at the same time; supports users to add marks and annotations to the image, draw the bounding box position of the target, and manually fill in the relevant attribute values ​​of the target. The rule library verifies whether it complies with the relevant rules. If it complies, the target is saved. If it does not, it prompts that the relevant attribute value does not comply with a certain rule. Other users participating in the interpretation task can see these operations in real time;

[0040] Real-time synchronization of target interpretation results: By synchronizing the target interpretation results to the system in real time, all participating users and management users can instantly view the latest interpretation results. At the same time, the historical records of each target interpretation are saved for easy backtracking and review. Version control is supported, allowing users to view and restore previous versions of target interpretation results.

[0041] Furthermore, the rule base is user-defined verification rules.

[0042] Furthermore, the intelligent interpretation auxiliary module uses the target history library, target knowledge library, target situation library, image search engine and similar target extraction to assist in target interpretation, where:

[0043] Target history library: Provides auxiliary interpretation function by querying historically saved target information;

[0044] Target knowledge base: provides information about relevant targets by retrieving knowledge graphs and analyzing text;

[0045] Target situation database: provides auxiliary interpretation function by retrieving the target's historical trajectory and historical activity information;

[0046] Image search engine: By cutting and interpreting the sliced ​​images of the target, it searches for similar targets in the history library, knowledge base, and situation library, displays them on the interpretation interface, and provides auxiliary interpretation functions;

[0047] Similar target extraction: After the user draws a target on a large-size image, the SAM2 model is used to extract the target contours similar to the drawn target on the large-size image, making it easier for users to identify similar targets.

[0048] A method for linking and strictly judging large-size image targets based on identity responsibilities is implemented by the above-mentioned linking and strictly judging large-size image targets based on identity responsibilities, and includes the following steps:

[0049] S1, create users, create identity responsibilities and bind users in the identity responsibility matrix engine module;

[0050] S2, input large size image;

[0051] S3, in the intelligent linkage task scheduling central module, intelligent matching based on identity responsibilities is performed, and then the linkage mode is selected and tasks are allocated according to the linkage mode;

[0052] S4, in the multi-user linkage real-time interpretation module and the intelligent interpretation auxiliary module, the users participating in the interpretation perform multi-user linkage real-time interpretation in the corresponding linkage mode interface.

[0053] The beneficial effects of the present invention are:

[0054] (1) Improve interpretation efficiency: The interpretation results can be synchronized to the interfaces of all participants in real time to ensure the consistency and timeliness of information, thereby speeding up the interpretation process.

[0055] (2) Personalization and flexibility: The system can customize the processing of target element types in remote sensing images according to the division of responsibilities of different users, ensuring that each user can focus on his or her area of ​​expertise.

[0056] (3) Intelligent task allocation: Combining the advantages of multiple linkage modes, flexibly switching according to user capabilities and task requirements to achieve optimal task allocation.

[0057] (4) Real-time synchronization and feedback mechanism: The interpretation results can be synchronized to the interface of all participants in real time to ensure the consistency and timeliness of information, thereby speeding up the interpretation process.

[0058] In summary, the present invention significantly improves the performance of large-size image target interpretation systems, enhances the accuracy and practicability of remote sensing image analysis, and promotes the application of remote sensing technology in more practical scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a composition diagram of the system of the present invention.

[0060] Figure 2 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present invention clearer and more understandable, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings.

[0062] The present invention proposes a large-size image target linkage strict judgment system based on identity and responsibility. The system relies on a high-performance computing platform, adopts WebGIS and HTTP technologies, and is based on a B / S (browser / server) architecture. It realizes the linkage and strict judgment capability of large-size image targets, ensures the stable operation and efficient processing capability of the system, optimizes the human-computer collaboration process, and improves the quality and work efficiency of target interpretation.

[0063] A large-size image target linkage strict judgment system based on identity and responsibility, such as Figure 1 As shown, it includes an identity responsibility matrix engine module, an intelligent linkage task scheduling center module, a multi-user linkage real-time interpretation module, and an intelligent interpretation auxiliary module.

[0064] Identity Responsibility Matrix Engine Module

[0065] The identity responsibility matrix engine module achieves efficient allocation of resources and tasks within the system through advanced user management, dynamic management of identity responsibilities, voting based on identity responsibilities, and feedback mechanisms. It further combines deep learning and data analysis to provide accurate user capability assessment and dynamic identity responsibility adjustment, significantly improving work efficiency and team collaboration.

[0066] The identity responsibility matrix engine module includes user management, identity responsibility dynamic management, identity responsibility-based voting mechanism, and identity responsibility-based feedback mechanism.

[0067] (1) User Management

[0068] User management enables adding new users, deleting users, modifying users, and querying users.

[0069] (2) Dynamic management of identity responsibilities

[0070] Dynamic management of identity responsibilities enables flexible allocation and real-time optimization of user identity responsibilities.

[0071] The identity responsibility is to be responsible for the interpretation of a certain target type in the system, which is divided into three levels: primary, intermediate and advanced. For example, primary aircraft interpretation, intermediate aircraft interpretation, advanced aircraft interpretation, primary ship interpretation, intermediate ship interpretation, advanced ship interpretation, etc.

[0072] The responsibilities of the primary reader are:

[0073] Daily interpretation: Responsible for daily interpretation tasks and interpretation of common targets.

[0074] Feedback issues: For interpretation targets that cannot be determined after voting, feedback will be provided to the intermediate interpretation through the feedback mechanism.

[0075] Accept the result: accept the feedback of the intermediate or advanced interpretation and complete the target interpretation.

[0076] Initiate voting: For uncertain interpretation target attributes, initiate voting among junior readers.

[0077] Participate in voting: Independently evaluate and vote on uncertain interpretation targets.

[0078] The responsibilities of the intermediate reader are:

[0079] Daily interpretation: Responsible for daily interpretation tasks and interpretation of common targets.

[0080] Feedback issues: For interpretation targets that cannot be determined after voting, feedback is provided to senior interpreters through the feedback mechanism.

[0081] Accept results: accept the feedback results of the advanced interpretation and complete the target interpretation.

[0082] Receive feedback: Accept feedback on primary interpretations and make interpretations of uncertain targets.

[0083] Initiate voting: For uncertain interpretation target attributes, initiate voting among intermediate interpreters.

[0084] Participate in voting: Independently evaluate and vote on uncertain interpretation targets.

[0085] The responsibilities of a senior reader are:

[0086] Daily interpretation: Responsible for daily interpretation tasks and interpretation of common targets.

[0087] Receive feedback: Receive feedback on intermediate interpretations and make interpretations on uncertain targets.

[0088] Initiate voting: For uncertain interpretation target attributes, initiate voting among senior interpreters.

[0089] Participate in voting: Independently evaluate and vote on uncertain interpretation targets.

[0090] Final decision: For targets that cannot be determined after voting, the final target attributes are determined.

[0091] Dynamic management of identity responsibilities includes static management of identity responsibilities and dynamic adjustment.

[0092] Static management of identity responsibilities manages identity responsibilities through adding, deleting, modifying, and querying, and supports many-to-many association of identity responsibilities with users.

[0093] For example, primary aircraft interpretation is bound to multiple users (user A-primary aircraft interpretation, user B-primary aircraft interpretation, user C-primary aircraft interpretation), and a user can also have multiple identity responsibilities (user A-primary aircraft interpretation, user A-intermediate ship interpretation).

[0094] Dynamic adjustment of identity responsibilities tracks the target interpretation completion status (average interpretation time, average interpretation accuracy, number of completed interpretation targets, number of independently completed interpretation targets) of users under different identity responsibilities, evaluates the target interpretation ability of users under different identity responsibilities through hierarchical analysis method, generates detailed performance reports, and dynamically recommends adjustments to identity responsibilities based on analysis results and task requirements to improve work efficiency.

[0095] For example, if user A's identity responsibility is primary aircraft interpretation, when the average aircraft interpretation time index is A, the average interpretation accuracy index is A, the number of aircraft interpretations completed is A, and the number of aircraft interpretations completed independently is A, it will be dynamically adjusted to intermediate aircraft interpretation.

[0096] (3) Voting mechanism based on identity responsibilities

[0097] When a reading user encounters an uncertain reading target, he or she can initiate a vote for the identity responsibility at the same level. If the ratio of the votes for a certain attribute value of the target to the number of users participating in the vote exceeds 1 / 2, the target attribute value is accepted as the reading value of the target attribute; if the ratio of the votes for a certain attribute value of the target to the number of users participating in the vote does not exceed 1 / 2, feedback is given to the identity responsibility at a higher level.

[0098] For example, when a user whose identity responsibility is primary aircraft interpretation interprets the type of an aircraft, if the interpreter cannot determine whether it is a carrier-based aircraft / fighter, a vote can be initiated for the type attribute value of the aircraft target, which is carrier-based aircraft / fighter. The user whose identity responsibility is primary aircraft interpretation will receive a voting notification. If the number of votes that agree that the aircraft type is a fighter exceeds the ratio of the number of users who participated in the vote, the aircraft target type is a fighter. If the number of votes that agree that the aircraft type is uncertain exceeds the ratio of the number of users who participated in the vote, feedback will be given to the user whose identity responsibility is intermediate aircraft interpretation.

[0099] (4) Feedback mechanism based on identity responsibilities

[0100] When the interpretation user still cannot determine the attributes of the interpretation target after voting based on the voting mechanism of identity responsibility, feedback is given to the identity responsibility of a higher level, and a interpretation user with identity responsibility of a higher level is randomly obtained to interpret the target; when the interpretation user with identity responsibility of a higher level cannot determine the attributes of the target, a vote is initiated with identity responsibility of the same level; when the voting with identity responsibility of the same level still cannot determine the attributes of the target, feedback is continued to be given to the interpretation user with identity responsibility of a higher level.

[0101] Intelligent linkage task scheduling central module

[0102] The intelligent linkage task scheduling central module realizes intelligent matching based on identity responsibilities and multi-mode intelligent linkage task allocation.

[0103] The intelligent linkage task scheduling central module includes identity-responsibility-based intelligent matching and multi-mode intelligent linkage task allocation.

[0104] (1) Intelligent matching based on identity and responsibilities

[0105] Select the target to be read, and sort the users under a target that meets the requirements in ascending order according to the user analysis results based on identity responsibilities and the urgency of the task, where:

[0106] The urgency of tasks is divided into general, urgent, and urgent;

[0107] Define an array M to represent the average reading time of each user (the time it takes to read a target): M = [M 1 ,M 2 ,…,M i ];

[0108] Define an array N to represent the average reading accuracy of each user (the accuracy of reading a target): N = [N 1 ,N 2 ,…,N i ];

[0109] Define an array Q to represent the unfinished task amount of each user (participating in the interpretation of a certain target in a large-size image, but not yet completed): Q = [Q 1 ,Q 2 ,…,Q i ];

[0110] Define an array P to represent the number of completed interpretation targets for each user: P = [P 1 ,P 2 ,…,P i ];

[0111] Define an array R to represent the number of independent completions of each user's interpretation goals (this goal has not been voted on or feedback mechanism): R = [R 1 ,R 2 ,…,R i ];

[0112] Where i is the number of users;

[0113] make represents the index value of the jth element in array X after X is sorted in ascending order (index starts from 1), let represents the index value of the jth element in array X after X is sorted in descending order (index starts from 1), then the comprehensive score of each user can be calculated by the following formula:

[0114]

[0115] Among them, ω M ,ω N ,ω Q ,ω P ,ω R They are the weights of the average interpretation time, average interpretation accuracy, the amount of tasks yet to be completed, the number of completed interpretation targets, and the number of independently completed interpretation targets (the targets have not been voted on or feedback mechanisms have not been implemented);

[0116] Finally, according to S j Sort all users in ascending order based on the values ​​of to obtain the final sorted array S:

[0117] S=sort(S 1 ,S 2 ,...,S i );

[0118] Among them, when the urgency of the task is normal:

[0119] ω M =100,ω N =1000,ω Q =10000,ω P =10,ωR =1;

[0120] Among them, when the urgency of the task is urgent:

[0121] ω M =1000,ω N =10000,ω Q =1,ω P =100,ω R =10;

[0122] Among them, when the urgency of the task is extremely urgent:

[0123] ω M =10000,ω N =1000,ω Q =1,ω P =10,ω R =100.

[0124] (2) Multi-mode intelligent linkage task allocation

[0125] Realize task allocation in three linkage modes: target linkage, regional linkage and intelligent linkage, among which:

[0126] Task allocation in target linkage mode: According to the list of interpretation users output by intelligent matching based on identity responsibilities, the best interpretation user for each interpretation target (i.e. the user ranked highest) is selected, and the corresponding target interpretation task is assigned to the interpretation user.

[0127] Task allocation in regional linkage mode: According to the list of interpretation users output by intelligent matching based on identity responsibilities, select the number of users n who participate in target interpretation, divide the entire large-size image into task areas corresponding to the number n evenly, and assign the task areas to the users who participate in target interpretation.

[0128] Task allocation under intelligent linkage mode: According to the list of interpretation users output by intelligent matching based on identity responsibilities, the list of interpretation targets that can be used in the target linkage mode is screened out through the pre-set user analysis based on identity responsibilities, and the corresponding optimal target interpretation personnel are calculated. For the remaining interpretation targets and personnel that cannot be allocated through the target linkage mode, task allocation is carried out according to the regional linkage mode, that is, the large-size image is evenly divided into multiple task areas, and the task areas are evenly allocated to the corresponding interpretation users according to the number of remaining users who can participate in the interpretation.

[0129] Multi-user linkage real-time interpretation module

[0130] The multi-user linkage real-time interpretation module realizes large-size image display, multi-user linkage target interpretation, and real-time synchronization of target interpretation results, including:

[0131] Large-size image display: By converting large-size images into Cloud optimized GeoTIFF files (a GeoTIFF image format optimized for cloud storage and network transmission) and storing them in Minio object storage (a high-performance object storage system), the standard S3 protocol (Amazon S3 API, a REST FUL Web service interface for interacting with object storage) is provided for loading and displaying on the browser, supporting free zooming, panning, rolling shutters, gradients, and simultaneous display of multiple image windows; global display / task area display switching can also be performed according to different interpretation requirements.

[0132] Multi-user linkage target interpretation: supports real-time collaboration among multiple users, and can annotate and fill in attributes of the same image at the same time; supports users to add marks and annotations to the image, draw the bounding box position of the target, and manually fill in the relevant attribute values ​​of the target. The rule library verifies whether it complies with the relevant rules. If it does, the target is saved. If it does not, it prompts that the relevant attribute value does not comply with a certain rule. Other users participating in the interpretation task can see these operations in real time.

[0133] The rule base is user-defined validation rules.

[0134] For example, date format, numerical range, latitude and longitude accuracy, and conventional cultivated land and lakes will not appear at sea level, glaciers will not appear in tropical areas, and ships will not appear in city centers.

[0135] Real-time synchronization of target interpretation results: By synchronizing the target interpretation results to the system in real time, all participating users and management users can instantly view the latest interpretation results. At the same time, the historical records of each target interpretation are saved for easy backtracking and review. Version control is supported, making it easy for users to view and restore previous versions of target interpretation results.

[0136] Intelligent interpretation auxiliary module

[0137] The intelligent interpretation auxiliary module uses the target history library, target knowledge library, target situation library, image search engine and similar target extraction to assist in target interpretation, including:

[0138] Target history library: Provides auxiliary interpretation function by querying historically saved target information.

[0139] Target knowledge base: Provides information about related targets by retrieving knowledge graphs and performing text analysis.

[0140] Target situation database: provides auxiliary interpretation function by retrieving the target's historical trajectory and historical activity information.

[0141] Image search engine: By cropping the slice image of the target to be interpreted, searching for similar targets in the history library, knowledge base and situation library, and displaying them on the interpretation interface, it provides auxiliary interpretation function.

[0142] Similar target extraction: After the user draws a target on a large-size image, the SAM2 model (an advanced image and video segmentation model, an upgraded version of the SAM model) is used to extract the target contours similar to the drawn target on the large-size image, making it easier for users to identify similar targets.

[0143] A large-size image target linkage strict judgment method based on identity responsibility is implemented by the large-size image target linkage strict judgment system based on identity responsibility, such as Figure 2 As shown, the following steps are included:

[0144] S1, create users, create identity responsibilities and bind users in the identity responsibility matrix engine module.

[0145] Create users A1, A2, A3, A4, B1, B2, B3, C1, C2, and C3, and create identity responsibilities of primary aircraft interpretation, intermediate aircraft interpretation, and advanced aircraft interpretation. Bind the identity responsibilities of users A1, A2, A3, and A4 to primary aircraft interpretation, bind the identity responsibilities of users B1, B2, and B3 to intermediate aircraft interpretation, and bind the identity responsibilities of users C1, C2, and C3 to advanced aircraft interpretation.

[0146] S2, input large size image.

[0147] S3: Perform intelligent matching based on identity responsibilities in the intelligent linkage task scheduling central module, then select the linkage mode and assign tasks according to the linkage mode. The specific steps include:

[0148] S31, select the aircraft as the target of interpretation and the general mission urgency, and according to the analysis results of users based on identity and responsibility, arrange the users whose identity and responsibility are primary aircraft interpretation in ascending order: A1-A2-A3-A4-B1-B2-B3-C1-C2-C3, where the ranking calculation formula is as follows:

[0149]

[0150] Among them, ω M ,ω N ,ω Q ,ω P ,ω R are the weights of the average interpretation time, average interpretation accuracy, unfinished tasks, number of completed interpretation targets, and number of independently completed interpretation targets (the target has not been voted and feedback mechanism), ω M =100,ω N=1000,ω Q =10000,ω P =10,ω R =1.

[0151] S32, selecting a linkage mode, and performing task allocation according to the linkage mode. The specific steps include:

[0152] S321, if the target linkage mode is selected, the best interpretation user for each target is selected (i.e., the user ranked first) according to the interpretation user list output by the identity-responsibility-based intelligent matching, and the corresponding target interpretation task is assigned to the interpretation user, that is, the aircraft target interpretation task is assigned to the interpretation user A1;

[0153] S322, if the regional linkage mode is selected, according to the interpretation user list output by the intelligent matching based on identity responsibilities, select the number of users participating in the target interpretation as 4, divide the entire large-size image into 4 task areas evenly, and assign the task areas to the users A1-A2-A3-A4 participating in the target interpretation;

[0154] S323, if the intelligent linkage mode is selected, according to the interpretation user list output by the intelligent matching based on identity responsibilities, according to the pre-set user analysis threshold based on identity responsibilities, the interpretation target list that can be used for the target linkage mode - aircraft is screened out, and the corresponding optimal target interpretation personnel - A1 is calculated, and the remaining interpretation target list and the remaining interpretation personnel are assigned tasks according to the regional linkage mode (if there are no interpretation targets left, no task assignment is performed).

[0155] S4, in the multi-user linkage real-time interpretation module and the intelligent interpretation auxiliary module, the users participating in the interpretation perform multi-user linkage real-time interpretation in the corresponding linkage mode interface, and the specific steps include:

[0156] S41, the browser loads and displays the large-size image, and performs real-time enhanced display according to user operations;

[0157] S42, if the user is interpreting in the target linkage mode, only the assigned target type can be interpreted and saved;

[0158] S43, if the user interprets in the area linkage mode, only the assigned task area can be interpreted and saved;

[0159] S44, by drawing the target outline, calling the image search engine to query similar targets from the history library, the situation library and the knowledge library, and displaying them on the interpretation interface to achieve auxiliary interpretation;

[0160] S45, through keyword retrieval, query the corresponding information from the history library, situation library and knowledge library, and display it on the interpretation interface to achieve auxiliary interpretation;

[0161] S46, by inputting the attribute value, retrieving information similar to the attribute value from the history library, situation library and knowledge library, and displaying it on the interpretation interface to achieve auxiliary interpretation;

[0162] S47, draw the target outline, input the corresponding attribute value, and complete the target interpretation;

[0163] S48, after the target is interpreted, the relevant rule base is called for verification. If the verification passes, the target is saved. If the verification fails, it is prompted that a certain attribute value has not passed the corresponding rule verification;

[0164] S49, after the target is stored in the database, similar target interpretation is performed;

[0165] S410, according to the outline drawn by the target, cutting the target slice, calling the SAM2 model to extract similar target outline information;

[0166] S411, when the user encounters an undetermined target, the user determines the target's attributes through the identity responsibility-based voting mechanism and the identity responsibility-based feedback mechanism in the identity responsibility matrix engine module. The specific steps include:

[0167] S411.1, when the interpretation user A1 encounters an interpretation target that cannot be determined, he can initiate a vote among the same level of identity responsibilities, that is, initiate a vote among users A1-A2-A3-A4;

[0168] S411.2, if the ratio of voting to voting users for a certain attribute value of the target exceeds 1 / 2, the target attribute value is accepted as the judgment value of the target attribute;

[0169] S411.3, if the ratio of voting to voting users for a certain attribute value of the target does not exceed 1 / 2, feedback is given to the identity responsibility at a higher level, that is, A1 gives feedback to B1 or B2 or B3;

[0170] S411.4, if the user with identity responsibilities at a higher level cannot determine the target attribute, a vote can be initiated for the identity responsibilities at the same level, that is, a vote can be initiated among users B1-B2-B3;

[0171] S412, all operations in the target interpretation process are synchronously sent and displayed on the browser interface of other users who participate in the interpretation task.

[0172] The contents not described in detail in the specification of the present invention belong to the currently known technologies in the art.

Claims

1. A large-size image target linkage strict judgment system based on identity and responsibility, characterized in that: It includes identity responsibility matrix engine module, intelligent linkage task scheduling center module, multi-user linkage real-time interpretation module, and intelligent interpretation auxiliary module, among which: The identity responsibility matrix engine module realizes the efficient allocation of resources and tasks within the system, and further combines data analysis to provide accurate user capability assessment and dynamic identity responsibility adjustment; The intelligent linkage task scheduling hub module realizes intelligent matching based on identity responsibilities and multi-mode intelligent linkage task allocation; The multi-user linkage real-time interpretation module realizes large-size image display, multi-user linkage target interpretation, and real-time synchronization of target interpretation results; The intelligent interpretation assistance module uses the target history library, target knowledge base, target situation library, image search engine and similar target extraction to assist in target interpretation.

2. The large-size image target linkage strict judgment system based on identity and responsibility according to claim 1 is characterized in that: The identity responsibility matrix engine module includes user management, dynamic management of identity responsibilities, voting mechanism based on identity responsibilities, and feedback mechanism based on identity responsibilities; user management implements the functions of adding users, deleting users, modifying users, and querying users; Dynamic management of identity responsibilities realizes flexible allocation and real-time optimization of user identity responsibilities; the voting mechanism based on identity responsibilities means that when a judgment user encounters an uncertain judgment target, he / she initiates a vote with identity responsibilities of the same level. If the ratio of the vote for a certain attribute value of the target to the number of users participating in the vote exceeds 1 / 2, the target attribute value is accepted as the judgment value of the target attribute; if the ratio of the vote for a certain attribute value of the target to the number of users participating in the vote does not exceed 1 / 2, feedback is given to the identity responsibilities of a higher level; the feedback mechanism based on identity responsibilities means that when the judgment user still cannot determine the attribute of the judgment target after voting based on the voting mechanism of identity responsibilities, he / she gives feedback to the identity responsibilities of a higher level, and randomly obtains a judgment user with a higher-level identity responsibility to judge the target; when the judgment user with the identity responsibility of a higher level cannot determine the target attribute, a vote with the identity responsibility of the same level is initiated; when the voting with the identity responsibility of the same level still cannot determine the target attribute, feedback is continued to be given to the judgment user with the identity responsibility of a higher level.

3. The large-size image target linkage strict judgment system based on identity and responsibility according to claim 2 is characterized in that: Dynamic management of identity responsibilities includes static management and dynamic adjustment of identity responsibilities; static management of identity responsibilities manages identity responsibilities through adding, deleting, modifying, and querying, and supports many-to-many association of identity responsibilities with users; dynamic adjustment of identity responsibilities refers to tracking the completion of target interpretation by users under different identity responsibilities, evaluating the target interpretation ability of users under different identity responsibilities through hierarchical analysis method, generating detailed performance reports, and dynamically proposing adjustments to identity responsibilities based on analysis results and task requirements.

4. The large-size image target linkage strict judgment system based on identity and responsibility according to claim 1 is characterized in that: The intelligent linkage task scheduling central module includes intelligent matching based on identity responsibilities and multi-mode intelligent linkage task allocation; intelligent matching based on identity responsibilities refers to selecting the interpretation target, and arranging the users under a certain interpretation target that meet the requirements in ascending order according to the user analysis results based on identity responsibilities and the urgency of the task; Multi-mode intelligent linkage task allocation realizes task allocation under three linkage modes: target linkage, regional linkage and intelligent linkage.

5. The large-size image target linkage strict judgment system based on identity and responsibility according to claim 4 is characterized in that: The intelligent matching based on identity and responsibilities selects the interpretation target. According to the user analysis results based on identity and responsibilities and the urgency of the task, the users under a certain interpretation target that meet the requirements are arranged in ascending order, where: The urgency of tasks is divided into general, urgent, and urgent; Define an array M to represent the average reading time of each user: M = [M1, M2, ..., M i ]; Define an array N to represent the average accuracy of each user: N = [N1, N2, ..., N i ]; Define an array Q to represent the amount of unfinished tasks for each user: Q = [Q1, Q2, ..., Q i ]; Define an array P to represent the number of completed interpretation targets for each user: P = [P1, O2, ..., P i ]; Define an array R to represent the number of independent interpretation targets completed by each user: R = [R1, R2, ..., R i ]; Where i is the number of users; make represents the index value of the jth element in array X after X is sorted in ascending order. Let represents the index value of the jth element in array X after X is sorted in descending order. The comprehensive score of each user can be calculated by the following formula: Among them, ω M ,ω N ,ω Q ,ω P ,ω R They are the weights of average interpretation time, average interpretation accuracy, unfinished tasks, number of completed interpretation targets, and number of independently completed interpretation targets; According to S j Sort all users in ascending order based on the values ​​of to obtain the final sorted array S: S=sort(S1,S2,...,S i ); Among them, when the urgency of the task is normal: oh M =100,oh N =1000,oh Q =10000,oh P =10, ω R =1; Among them, when the urgency of the task is urgent: oh M =1000,oh N =10000,oh Q =1,ω P =100,oh R =10; Among them, when the urgency of the task is extremely urgent: oh M =10000,oh N =1000,oh Q =1,ω P =10, ω R = 100.

6. The large-size image target linkage strict judgment system based on identity and responsibility according to claim 4 is characterized in that: Multi-mode intelligent linkage task allocation realizes task allocation in three linkage modes: target linkage, regional linkage and intelligent linkage, among which: Task allocation in target linkage mode: According to the list of interpretation users output by intelligent matching based on identity responsibilities, the best interpretation user for each interpretation target is selected, and the corresponding target interpretation task is assigned to the interpretation user; Task allocation in regional linkage mode: According to the list of users output by intelligent matching based on identity responsibilities, select the number of users n who participate in target interpretation, divide the entire large-size image into task areas corresponding to the number n, and allocate the task areas to the users who participate in target interpretation; Task allocation under intelligent linkage mode: According to the list of interpretation users output by intelligent matching based on identity responsibilities, the list of interpretation targets that can be used in the target linkage mode is screened out through the pre-set user analysis threshold based on identity responsibilities, and the corresponding optimal target interpretation personnel are calculated. For the remaining interpretation targets and personnel that cannot be allocated through the target linkage mode, task allocation is carried out according to the regional linkage mode, and the task area is evenly allocated to the corresponding interpretation users according to the number of remaining users who can participate in the interpretation.

7. The large-size image target linkage strict judgment system based on identity and responsibility according to claim 1 is characterized in that: The multi-user linkage real-time interpretation module realizes large-size image display, multi-user linkage target interpretation, and real-time synchronization of target interpretation results, including: Large-size image display: By converting large-size images into Cloud optimized GeoTIFF files and storing them in Minio object storage, the standard S3 protocol is provided for loading and displaying them on the browser. It supports free zooming, panning, rolling shutters, gradients, and simultaneous display of multiple image windows. It also switches between global display and task area display according to different interpretation requirements. Multi-user linkage target interpretation: supports real-time collaboration among multiple users, and can annotate and fill in attributes of the same image at the same time; supports users to add marks and annotations to the image, draw the bounding box position of the target, and manually fill in the relevant attribute values ​​of the target. The rule library verifies whether it complies with the relevant rules. If it complies, the target is saved. If it does not comply, it prompts that the relevant attribute value does not comply with a certain rule. Other users participating in the interpretation task can see these operations in real time; Real-time synchronization of target interpretation results: By synchronizing the target interpretation results to the system in real time, all participating users and management users can instantly view the latest interpretation results. At the same time, the historical records of each target interpretation are saved for easy backtracking and review. Version control is supported, allowing users to view and restore previous versions of target interpretation results.

8. The large-size image target linkage strict judgment system based on identity and responsibility according to claim 7 is characterized in that: The rule base is user-defined validation rules.

9. The large-size image target linkage strict judgment system based on identity and responsibility according to claim 1 is characterized in that: The intelligent interpretation auxiliary module uses the target history library, target knowledge library, target situation library, image search engine and similar target extraction to assist in target interpretation, including: Target history library: Provides auxiliary interpretation function by querying historically saved target information; Target knowledge base: provides information about relevant targets by retrieving knowledge graphs and analyzing text; Target situation database: provides auxiliary interpretation function by retrieving the target's historical trajectory and historical activity information; Image search engine: By cutting and interpreting the sliced ​​images of the target, it searches for similar targets in the history library, knowledge base, and situation library, and displays them on the interpretation interface to provide auxiliary interpretation functions; Similar target extraction: After the user draws a target on a large-size image, the SAM2 model is used to extract the target contours similar to the drawn target on the large-size image, making it easier for users to identify similar targets.

10. A method for linking and strictly judging large-size image targets based on identity responsibilities, implemented according to the system for linking and strictly judging large-size image targets based on identity responsibilities according to any one of claims 1 to 9, characterized in that: The following steps are involved: S1, create users, create identity responsibilities and bind users in the identity responsibility matrix engine module; S2, input large size image; S3, in the intelligent linkage task scheduling central module, intelligent matching based on identity responsibilities is performed, and then the linkage mode is selected and tasks are allocated according to the linkage mode; S4, in the multi-user linkage real-time interpretation module and the intelligent interpretation auxiliary module, the users participating in the interpretation perform multi-user linkage real-time interpretation in the corresponding linkage mode interface.

Images