Entity analysis device and method based on dual processing using AI, and smart farm monitoring system using same

Abstract

The invention discloses an entity analysis device and method based on dual processing using artificial intelligence, and a smart farm monitoring system using the same. An entity analysis method based on dual processing using artificial intelligence according to one embodiment of the present invention may comprise the steps of: collecting target data including image data and sensing data relating to an entity to be monitored by using a boundary device; executing, by the boundary device, an initial analysis of inputting the target data to a first model based on pre-learned artificial intelligence and outputting first analysis data relating to an action pattern of the entity; and when the first analysis data meets a preset detailed analysis condition, transmitting the first analysis data and the target data to a main server.

Description

Technical Field

[0001] This invention relates to an entity analysis apparatus and method based on dual processing utilizing artificial intelligence.

[0002] The invention relates to a smart farm monitoring system. For example, it relates to an AI-based dual-processing system for monitoring and analyzing livestock behavior in a smart farm environment. Background Technology

[0003] With the accelerated integration of agriculture and ICT technologies, traditional agricultural and pastoral practices are undergoing a comprehensive transformation towards smart farms. In particular, leading smart farm countries around the world are comprehensively exploring the application of ICT technologies such as the Internet of Things, nanotechnology, big data, cloud computing, robotics, and drones in agriculture and animal husbandry. Currently, some countries are implementing farm intelligence by using various systems to perform tasks such as harvest calculation, pest and disease diagnosis, soil moisture measurement, indicator status measurement, harvest time diagnosis, and crop status monitoring.

[0004] On the other hand, taking smart farms in the dairy farming sector as an example, due to the continuous development of breeding technology,

[0005] Ruminants are becoming increasingly larger, and their rearing environments have shifted from simple, stilted facilities to large, dry spaces with compost yards. Furthermore, due to industry consolidation, large-scale farming is now prevalent in livestock operations. Currently, the ultimate goal of large-scale ruminant farming systems is to improve the quality of beef cattle and the yield and quality of milk in dairy cows while ensuring the health and comfort of the livestock.

[0006] To achieve this goal, it is necessary not only to study the physiological behavior of ruminants, but also...

[0007] Environmental optimization and management are of paramount importance.

[0008] Over the past few decades, studying the behavior of ruminants and the impact of their environment has become an essential process in livestock and veterinary medicine. To achieve these goals, considering ruminant behavior and their rearing environment, based on existing physiological research, has become particularly important. Ruminant behavior includes responses to other ruminants and other objects, which are related to overall responses and adaptations to various internal and external conditions. Through observation of ruminant behavioral systems, it is known that ruminants primarily engage in feeding behavior, resting behavior, social behavior, group behavior, estrus behavior, reproductive behavior, and offspring-related behavior. Considering these behaviors is crucial for improving productivity, which is one of the aforementioned target factors. Furthermore, environmental indicators such as cleanliness, temperature and humidity, and stocking density are increasingly becoming important factors in considering livestock productivity. For example, the design of feeding and watering areas affects the health and physiological state of ruminants.

[0009] Furthermore, even though extensive research has been conducted on the regulation of the physiological state and reproductive environment of ruminants and the feeding environment for ruminants has been specified, as the scale of breeding expands, more manpower is required to observe all ruminants. Although sensor products play an auxiliary role to some extent, in addition to the cumbersome process of attaching multiple sensors to a single animal, they can not only increase the stress on ruminants, but also cause the economic burden on farms to increase with the expansion of the breeding scale.

[0010] The background technology of this invention is disclosed in Korean Patent Publication No. 10-1329022. Summary of the Invention

[0011] Technical issues

[0012] In order to solve the problems of the prior art, the present invention aims to provide an entity analysis device and method based on dual processing using artificial intelligence, which can effectively monitor and analyze entity actions in a smart farm environment, and a smart farm monitoring system using the same.

[0013] However, the technical objectives to be addressed by the embodiments of the present invention are not limited to the above-mentioned technical objectives, and other technical objectives may exist.

[0014] Technical solution

[0015] As a technical solution for achieving the above-mentioned technical objectives, an embodiment of the present invention provides an entity analysis method based on dual processing using artificial intelligence, which may include the following steps: collecting target data, including image data and sensing data related to the entity being monitored, using a boundary device; performing an initial analysis using the boundary device, inputting the target data into a first model based on pre-learned artificial intelligence to output first analysis data related to the action pattern of the entity; and transmitting the first analysis data and the target data to a main server when the first analysis data meets preset detailed analysis conditions.

[0016] Furthermore, the aforementioned first analysis data may include information on the presence or absence of the aforementioned entities reflected in the aforementioned image data, as well as information on their location.

[0017] Furthermore, the aforementioned first analysis data may include first classification information related to the behavior type of the aforementioned entity.

[0018] Furthermore, the detailed analysis conditions mentioned above may include conditions corresponding to the first category information and preset abnormal or important behaviors.

[0019] Furthermore, the aforementioned second analysis data may include second classification information related to the behavioral type of the aforementioned entity.

[0020] Furthermore, the number of categories in the second category information can be greater than the number of categories in the first category information.

[0021] Furthermore, the first model described above can be a lightweight model constructed by taking into account the resource information of the aforementioned boundary devices.

[0022] Furthermore, the aforementioned second model may include multiple models pre-built based on multiple different network structures.

[0023] Furthermore, the second set of analytical data mentioned above can be determined by considering the consistency of the output data of each of the aforementioned multiple models.

[0024] On the other hand, an embodiment of the present invention of an entity analysis method based on dual processing using artificial intelligence may include the following steps: performing an initial analysis of first analysis data related to the action pattern of the entity by inputting target data, including image data and sensing data related to the entity being monitored, into a first model based on pre-learned artificial intelligence using a boundary device; receiving the first analysis data and the target data from the boundary device when the first analysis data meets preset detailed analysis conditions; and performing a detailed analysis of second analysis data related to the action pattern of the entity by inputting at least one of the first analysis data and the target data into a second model based on pre-learned artificial intelligence.

[0025] Furthermore, the steps of performing the detailed analysis described above can be used to determine the second analysis data by considering the consistency of the output data of the various models.

[0026] On the other hand, a smart farm monitoring system according to an embodiment of the present invention may include: a boundary device that collects target data including image data and sensing data related to an entity being monitored, performs an initial analysis by inputting the target data into a first model based on pre-learned artificial intelligence to output first analysis data related to the action pattern of the entity, and transmits the first analysis data and the target data to a main server when the first analysis data meets preset detailed analysis conditions; and a main server that performs a detailed analysis by inputting at least one of the first analysis data and the target data into a second model based on pre-learned artificial intelligence to derive second analysis data related to the action pattern of the entity.

[0027] The above technical solutions are merely examples and should not be construed as limiting the scope of the invention. In addition to the exemplary embodiments described above, further embodiments may be included in the accompanying drawings and detailed descriptions.

[0028] The effects of the invention

[0029] According to the technical solution described above, the present invention can provide an entity analysis device and method based on dual processing using artificial intelligence, which can effectively monitor and analyze entity actions in a smart farm environment, and a smart farm monitoring system using the same.

[0030] According to the technical solution described above, the present invention performs basic data analysis through a boundary device and filters and transmits important data to the server. Even in resource-constrained environments, the dual data processing channels that perform advanced analysis maximize the efficiency of data processing and minimize the server load and data transmission costs.

[0031] According to the technical solution described above, the present invention can construct a smart farm monitoring system to cope with various types of livestock and various breeding environments. It can continuously detect the health status, stress status, etc. of each entity as the monitoring object and quickly assist in realizing appropriate responses.

[0032] According to the technical solution described above, the present invention can simplify the integration of smart farm monitoring systems by implementing a modular design for farms of various sizes and dual types of smart farms. Managers can customize the elements required for each smart farm by setting the corresponding settings for each system.

[0033] However, the effects that can be obtained by the present invention are not limited to the above-described effects, and other effects may exist. Attached Figure Description

[0034] Figure 1 This is a simplified structural diagram of a smart farm monitoring system according to an embodiment of the present invention.

[0035] Figure 2 This is a conceptual diagram illustrating the workflow of a smart farm monitoring system according to an embodiment of the present invention.

[0036] Figure 3 This is a conceptual diagram illustrating how artificial intelligence-based entity recognition utilizes image data associated with entities being monitored.

[0037] Figure 4 This is a simplified structural diagram of the main server of a smart farm monitoring system according to an embodiment of the present invention.

[0038] Figure 5 This is a simplified structural diagram of the boundary device of a smart farm monitoring system according to an embodiment of the present invention.

[0039] Figure 6 A flowchart illustrating the workflow of an embodiment of the present invention, which utilizes a boundary device to perform entity analysis based on dual processing using artificial intelligence.

[0040] Figure 7 A flowchart illustrating the workflow of an entity analysis method based on dual processing utilizing artificial intelligence, according to an embodiment of the present invention, executed using a main server.

[0041] Figure 8 A detailed workflow diagram for the analysis process used to derive second analysis data related to the entity's behavior patterns.

[0042] Explanation of reference numerals in the attached figures

[0043] 10: Smart Farm Monitoring System;

[0044] 100: Main server;

[0045] 110: Data Receiving Unit;

[0046] 120: Detailed Analysis Department;

[0047] 130: Analytical Information Provision Department;

[0048] 200: Boundary device;

[0049] 210: Collection Department;

[0050] 220: Initial Analysis Section;

[0051] 230: Ministry of Communications;

[0052] 21: Measurement module;

[0053] 22: Camera module;

[0054] 300: User terminal;

[0055] 20: Internet. Detailed Implementation

[0056] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, so that those skilled in the art can easily implement the present invention. However, the present invention can be implemented through various different embodiments and is not limited to the embodiments described herein. Moreover, for the purpose of clearly illustrating the present invention, parts unrelated to the description have been omitted from the drawings, and similar reference numerals have been given to similar parts throughout this specification.

[0057] Throughout this specification, when it is said that one part is "connected" to another part, it not only means "direct connection" but may also include "electrical connection" or "indirect connection" with other devices in between.

[0058] Throughout this specification, when referring to a part as being "above," "upper part," "upper end," "lower," "lower part," or "lower end" of another component, it may include situations where one component is in contact with another component, or situations where there are other components between the two components.

[0059] Throughout this specification, when a part is said to "include" another structural element, it means, unless specifically stated otherwise, that other structural elements are also included, and not excluded.

[0060] This invention relates to an entity analysis apparatus and method based on dual processing using artificial intelligence, and a smart farm monitoring system utilizing the same.

[0061] Figure 1 This is a simplified structural diagram of a smart farm monitoring system according to an embodiment of the present invention.

[0062] Reference Figure 1 An embodiment of the smart farm monitoring system 10 of the present invention may include a main server 100, a border device 200, and a user terminal 300. Furthermore, referring to... Figure 1 The smart farm monitoring system 10 may include: a measurement module 21, which measures and transmits various sensing data of the target space (e.g., farm, livestock shed, smart farm, etc.) as the monitoring object along the side of the boundary device 200; and a camera module 22, which captures and transmits image data of the target space.

[0063] The measurement module 21, camera module 22, main server 100, border device 200, and user terminal 300 can communicate with each other via network 20. Network 20 refers to the connection structure that enables the sending and receiving of information between various nodes such as terminals and servers. As an example, network 20 may include 3GPP (3rd Generation Partnership Project) networks, Long Term Evolution (LTE) networks, 5G networks, WIMAX (World Interoperability for Microwave Access) networks, the Internet, local area networks (LANs), wireless local area networks (WLANs), wide area networks (WANs), personal area networks (PANs), Wi-Fi networks, Bluetooth networks, satellite broadcasting networks, analog broadcasting networks, and digital multimedia broadcasting (DMB) networks, but is not limited to these.

[0064] For example, the user terminal 300 can be any type of wireless communication device, such as a smartphone, smartpad, tablet computer, personal communication system (PCS) terminal, global system for mobile communication (GSM) terminal, personal digital cellular (PDC) terminal, personal handyphone system (PHS) terminal, personal digital assistant (PDA) terminal, International Mobile Telecommunication-2000 (IMT-2000) terminal, W-CDMA (W-Code Division Multiple Access) terminal, or Wibro (Wireless Broadband Internet) terminal.

[0065] For reference, in describing the embodiments of the present invention, the user terminal 300 may be a terminal held by the manager of the target space (e.g., farm, livestock shed, smart farm, etc.) managed by the smart farm monitoring system 10 disclosed in the present invention.

[0066] The following is for reference Figure 2 This invention describes the data processing channel and system structure of the smart farm monitoring system 10 disclosed in this invention.

[0067] Figure 2 This is a conceptual diagram illustrating the workflow of a smart farm monitoring system according to an embodiment of the present invention.

[0068] Reference Figure 2 A smart farm monitoring system 10 according to an embodiment of the present invention may include: a boundary device 200, which is respectively configured in the target space that is the monitoring object; and a main server 100, which obtains data collected in each target space from the boundary device 200, analyzes the obtained data and provides analysis results.

[0069] Specifically, the edge device 200 can collect target data and perform initial analysis. This target data includes image data and sensor data related to the entity being monitored. The initial analysis outputs first analysis data related to the entity's behavior patterns by inputting the collected target data into a first model based on pre-learned artificial intelligence. Furthermore, when the first analysis data meets preset detailed analysis conditions, the edge device 200 can transmit the first analysis data and the target data to the main server 100.

[0070] In other words, the boundary device 200 disclosed in this invention exists in the target space. By setting up a measurement module 21 and a camera module 22 in the target space to monitor the actions, locations, and health status of entities that are monitored (e.g., dairy cows, calving cows, growing cows, dairy cows, cattle, calves, pigs, horses, chickens, sheep, etc., and crops raised in the target space), target data is collected and initial data processing is applied. A basic analysis (initial analysis) considering the resource information (e.g., computing power, etc.) of the boundary device 200 can be performed using a pre-learned first model (lightweight model).

[0071] For reference, the initial data processing of the target data may include extensive preprocessing work, such as noise removal, data normalization, and initial data classification.

[0072] Furthermore, the boundary device 200 distinguishes between normal and abnormal conditions occurring in the target space based on the analysis results of the first model (lightweight model). Alternatively, it can filter and transmit target data collected based on the important condition identification results and / or first analysis data initially analyzed by the first model to the main server 100. In order to filter the data transmitted to the main server 100, the selected data can be optimized and compressed using a specified important data selection criterion.

[0073] Furthermore, the main server 100 can perform detailed analysis, which derives second analysis data related to the entity's action pattern by inputting at least one of the first analysis data and target data received from the boundary device 200 into a second model based on pre-learned artificial intelligence.

[0074] In other words, the main server 100 disclosed in this invention can perform the following functions: receive data from the border device 200; when considering the resources of the border device 200, use a pre-learned second model to perform advanced artificial intelligence-based analysis (detailed analysis) on the received data that is difficult for the border device 200 to perform; and provide the results of the advanced analysis (detailed analysis) to the user terminal 300.

[0075] On the other hand, exemplified, such as Figure 2 As shown, when the smart farm monitoring system 10 analyzes and manages data configured in multiple target spaces (e.g., ...), Figure 2 When the boundary device 200 of a smart farm (such as "Farm A", "Farm B" etc.) collects data, the managers corresponding to each target space may hold user terminals 300 respectively, but are not limited to this.

[0076] The functions and operations of the main server 100 and the boundary device 200 of the smart farm monitoring system 10 disclosed in this invention are described in detail below. First, the data transmission process of data collection, initial analysis (basic analysis), and selection and optimization filtering of important data using the boundary device 200 is described. Then, the process of the main server 100 performing detailed analysis (advanced analysis) using the data received from the boundary device 200 and providing the analysis results to the user terminal 300 is described.

[0077] First, the border device 200 can be used to collect target data, which includes image data and sensor data related to the entity being monitored.

[0078] Furthermore, the boundary device 200 can be used to perform initial analysis, which outputs first analysis data related to the action patterns of entities by inputting target data into a first model based on pre-learned artificial intelligence.

[0079] On the other hand, the boundary device 200 disclosed in this invention can perform initial analysis using a first model, which is a lightweight model constructed taking into account the resource information of the boundary device 200. Relatedly, the first model can be an edge target model with lightweight design considering the processing performance, storage capacity, and processing speed of the boundary device 200, and can be effectively executed even by a boundary device 200 with limited computing power.

[0080] For example, compared to the second model used by the main server 100 to perform detailed analysis (advanced analysis), the first model may be designed to handle input data of a relatively small size or include multiple layers, with fewer layers than the second model, and may not include functions of relatively high complexity, and may have a relatively small number of specified computation iterations, but is not limited to these.

[0081] Furthermore, according to one embodiment of the present invention, the first model can be a model whose confidence value for the predicted category is set relatively lower than that of the second model. Relatedly, a model with a higher confidence value can have a higher degree of accuracy in predicting the corresponding category. However, in the case of the boundary device 200, when considering resources, it is more important to prevent the omission of specific important situations (e.g., specific types of actions of entities) than to provide highly accurate analysis results. Therefore, even if the possibility of misdiagnosis (incorrect category classification) is relatively high, the confidence value can be applied to the first model to rigorously prevent undiagnosed situations / events (omission of important situations / events).

[0082] In other words, according to one embodiment of the present invention, the confidence value (e.g., the first threshold) applied to the first model (lightweight model) for initial analysis can be relatively small compared to the confidence value (e.g., the second threshold) applied to the second model (server-side model) used for detailed analysis.

[0083] Specifically, the boundary device 200 can export the presence and location information of entities reflected in the image data as the first analysis data.

[0084] In this regard, Figure 3 This is a conceptual diagram illustrating how artificial intelligence-based entity recognition utilizes image data associated with entities being monitored.

[0085] Reference Figure 3 The boundary device 200 utilizes a first model of the image analysis model type to represent entities in the object image 1 of the specified frame included in the image data, and can specify a first boundary region B-1 including the identified entities to derive the presence or absence information of the entities and the location information of the entities.

[0086] For example, the boundary device 200 may perform analysis using a first model of the image analysis model type, such as identifying bounding boxes corresponding to entities appearing in the object image 1, identifying key points representing specific body parts of entities, and segmentation to classify the object types of the various parts constituting the object image 1, but is not limited to these.

[0087] Furthermore, referring to Figure 3 The boundary device 200 identifies a preset identification object (e.g., an ear tag, sensor, label, identification tag, etc. attached to the ear of an entity), and can specify a second boundary region B-2 that includes the identified identification object within the first boundary region B-1.

[0088] On the other hand, according to an embodiment of the present invention, the boundary device 200 may use a first model, which is an artificial intelligence-based image analysis model, to export the shape information of objects within the identified object image 1 as first analysis data. Relatedly, the object shape information may specifically include at least one of the object's color information and segmentation model information. The main server 100, which acquires this object shape information as the first analysis data, may use a database (not shown) containing entity identification information corresponding to multiple entities active in the target space, and label identification information corresponding to the type of shape information for multiple objects generated to distinguish shape information, to specify (identify) entities appearing in the first analysis data and image data related to the first analysis data.

[0089] As another example, the boundary device 200 can export first classification information related to the behavior type of an entity as first analysis data. Furthermore, if the first analysis data meets preset detailed analysis conditions, the boundary device 200 can transmit the first analysis data and target data to the main server 100.

[0090] In this regard, according to an embodiment of the present invention, if the first classification information corresponds to a preset abnormal or important behavior, the boundary device 200 determines that the detailed analysis conditions are met and can transmit the first analysis data and target data to the main server 100.

[0091] Furthermore, for example, when the entity is a dairy cow among livestock, the aforementioned abnormal or important behaviors may include climbing on the back (reproductive behavior), intake (ingesting food and water), abnormal behaviors caused by disease and stress, etc., but are not limited to these, and may be set differently depending on the type of entity located in the target space.

[0092] Next, we will explain the main server 100.

[0093] Based on the results of the initial analysis performed by the boundary device 200, if the first analysis data meets the preset detailed analysis conditions, the main server 100 can receive the first analysis data and target data from the boundary device 200. The target data includes image data and sensing data related to the entity being monitored. The detailed analysis outputs the first analysis data related to the entity's action pattern by inputting the target data into a first model based on pre-learned artificial intelligence.

[0094] Furthermore, the main server 100 can perform detailed analysis of the second analysis data related to the entity's action patterns by inputting at least one of the first analysis data and the target data into the second model based on pre-learned artificial intelligence.

[0095] Specifically, the main server 100 can utilize a second model to determine the second analysis data by considering the consistency of the output data of multiple models. The second model includes multiple models pre-built based on multiple different network structures. On the other hand, among the multiple models with different network structures, for example, one model is designed based on a convolutional neural network (CNN), and another model can be designed based on a transform structure, but it is not limited to this. The multiple models that the second model disclosed in this invention may include can apply various types of artificial intelligence algorithms and structures such as currently known or later developed deep learning networks, machine learning algorithms, supervised / unsupervised learning algorithms, etc.

[0096] In this regard, according to an embodiment of the present invention, the main server 100 may generate the analysis object data of the second model from the dataset received from the border device 200, which includes at least a portion of the target data and the first analysis data.

[0097] Furthermore, the main server 100 can individually input the same generated dataset into the multiple models included in the second model to obtain the prediction results of each of the multiple models.

[0098] Furthermore, the main server 100 can verify whether the prediction results derived from multiple models are the same. Specifically, if the prediction results derived from multiple models are all the same, the main server 100 can determine the same prediction results derived from multiple models as the second analysis data of the second model. In contrast, if at least some of the prediction results derived from multiple models are different, the main server 100 removes the existing prediction results derived from multiple models and can perform detailed analysis again using the data re-received from the boundary device 200 (in other words, the target data and the first analysis data).

[0099] On the other hand, in relation to this multi-model access method, compared to the second model which includes multiple models, the smart farm monitoring system 10 disclosed in this invention allows multiple models to have the same dataset and perform analysis independently. The analysis result is considered reliable only if multiple models predict the same category. In contrast, if only one model predicts a specific category, its result is considered an incorrect answer. Considering that minimizing erroneous results and ultimately providing highly reliable and accurate results is particularly important in commercial environments such as smart farms, unlike the first model, it is understood that the second model adopts an access method that improves the reliability of results by using multiple models, rather than relying on a single model.

[0100] Furthermore, the main server 100 can transmit the second analysis data derived from the detailed analysis by the main server 100 to the user terminal 300.

[0101] Figure 4 This is a simplified structural diagram of the main server of a smart farm monitoring system according to an embodiment of the present invention.

[0102] Reference Figure 4 The main server 100 may include a data receiving unit 110, a detailed analysis unit 120, and an analysis information providing unit 130.

[0103] Based on the results of the initial analysis performed by the boundary device 200, if the first analysis data meets the preset detailed analysis conditions, the data receiving unit 110 can receive the first analysis data and target data from the boundary device 200. The target data includes image data and sensing data related to the entity being monitored. The detailed analysis outputs the first analysis data related to the entity's action pattern by inputting the target data into a first model based on pre-learned artificial intelligence.

[0104] The detailed analysis unit 120 can perform detailed analysis by inputting at least one of the first analysis data and the target data into a second model based on pre-learned artificial intelligence to derive detailed analysis data related to the action patterns of the entity.

[0105] Specifically, the detailed analysis unit 120 can use a second model to consider the consistency of the output data of multiple models to determine the second analysis data. The second model includes multiple models pre-built based on multiple different network structures.

[0106] In this regard, according to an embodiment of the present invention, the detailed analysis unit 120 may generate a dataset received from the boundary device 200, including at least a portion of the target data and the first analysis data, as the analysis object data of the second model.

[0107] Furthermore, the detailed analysis unit 120 can individually input the same generated dataset into the multiple models included in the second model to obtain the prediction results of each of the multiple models.

[0108] Furthermore, the detailed analysis unit 120 can verify whether the prediction results derived from multiple models are the same. Specifically, if the prediction results derived from multiple models are all the same, the detailed analysis unit 120 can determine the same prediction results derived from multiple models as the second analysis data of the second model. In contrast, if at least some of the prediction results derived from multiple models are different, the detailed analysis unit 120 removes the existing prediction results derived from multiple models and can perform detailed analysis again using the data re-received from the boundary device 200 (in other words, the target data and the first analysis data).

[0109] The analysis information providing unit 130 can transmit second analysis data derived from detailed analysis by the main server 100 to the user terminal 300.

[0110] Figure 5 This is a simplified structural diagram of the boundary device of a smart farm monitoring system according to an embodiment of the present invention.

[0111] Reference Figure 5 The boundary device 200 may include a collection unit 210 and an initial analysis unit 220.

[0112] The boundary device 200 can collect target data through the collection unit 210, which includes image data and sensing data related to the entity being monitored.

[0113] The boundary device 200 can perform initial analysis through the initial analysis unit 220, inputting target data into a first model based on pre-learned artificial intelligence to output first analysis data related to the action pattern of the entity.

[0114] Examplely, the initial analysis unit 220 may perform an initial analysis using a first model, which is a lightweight model constructed taking into account the resource information of the boundary device 200.

[0115] Specifically, the initial analysis unit 220 can export information on the presence or absence of entities and their location in the image data as first analysis data. As another example, the initial analysis unit 220 can export first classification information related to the behavior type of the entity as first analysis data.

[0116] If the first analysis data meets the preset detailed analysis conditions, the communication unit 230 can transmit the first analysis data and target data to the main server 100.

[0117] Specifically, according to an embodiment of the present invention, if the first classification information corresponds to a preset abnormal behavior or important behavior, it is determined that the detailed analysis conditions are met, and the communication unit 230 can transmit the first analysis data and target data to the main server 100.

[0118] The workflow of the present invention will now be briefly described based on the above detailed description.

[0119] Figure 6 A flowchart illustrating the workflow of an embodiment of the present invention, which utilizes a boundary device to perform entity analysis based on dual processing using artificial intelligence.

[0120] Figure 6 The entity analysis method based on dual processing utilizing artificial intelligence, as shown, can be executed by the boundary device 200 described above. Therefore, even if the following description is omitted, the description related to the boundary device 200 can also be applied to the entity analysis method based on dual processing utilizing artificial intelligence.

[0121] Reference Figure 6 In step S11, the boundary device 200 can collect target data through the collection unit 210. The target data includes image data and sensing data related to the entity being monitored.

[0122] Next, in step S12, the boundary device 200 can perform initial analysis through the initial analysis unit 220, inputting target data into the first model based on pre-learned artificial intelligence to output first analysis data related to the action pattern of the entity.

[0123] Examplely, in step S12, the initial analysis unit 220 may perform an initial analysis using a first model, which is a lightweight model constructed taking into account the resource information of the boundary device 200.

[0124] Specifically, in step S12, the initial analysis unit 220 can export information on the presence or absence of entities reflected in the image data and the location information of the entities as first analysis data. As another example, in step S12, the initial analysis unit 220 can export first classification information related to the behavior type of the entities as first analysis data.

[0125] Subsequently, in step S13, if the first analysis data meets the preset detailed analysis conditions, the communication unit 230 can transmit the first analysis data and target data to the main server 100.

[0126] Specifically, according to an embodiment of the present invention, in step S13, if the first classification information corresponds to a preset abnormal behavior or important behavior, it is determined that the detailed analysis conditions are met, and the communication unit 230 can transmit the first analysis data and target data to the main server 100.

[0127] In the above description, steps S11 to S13 can be further divided into additional steps or combined into fewer steps according to the embodiments of the present invention. Furthermore, some steps can be omitted or the order of the steps can be changed as needed.

[0128] Figure 7 A flowchart illustrating the workflow of an entity analysis method based on dual processing utilizing artificial intelligence, according to an embodiment of the present invention, executed using a main server.

[0129] Figure 7 The entity analysis method based on dual processing utilizing artificial intelligence, as shown above, can be executed by the main server 100 described above. Therefore, even if the following description is omitted, the description related to the main server 100 can also be applied to the entity analysis method based on dual processing utilizing artificial intelligence.

[0130] Reference Figure 7 In step S21, based on the results of the initial analysis performed using the boundary device 200, if the first analysis data meets the preset detailed analysis conditions, the data receiving unit 110 can receive the first analysis data and target data from the boundary device 200. The target data includes image data and sensing data related to the entity being monitored. The detailed analysis outputs the first analysis data related to the entity's action pattern by inputting the target data into a first model based on pre-learning artificial intelligence.

[0131] Next, in step S22, the detailed analysis unit 120 may perform detailed analysis by inputting at least one of the first analysis data and the target data into the second model based on pre-learned artificial intelligence to derive the second analysis data related to the entity's action pattern.

[0132] Specifically, in step S22, the detailed analysis unit 120 can use the second model to consider the consistency of the output data of multiple models to determine the second analysis data. The second model includes multiple models pre-built based on multiple different network structures.

[0133] Furthermore, according to an embodiment of the present invention, in step S22, the detailed analysis unit 120 exports second analysis data containing second classification information related to the behavior type of the entity, and the number of categories of the exported second classification information may be greater than the number of categories of the first classification information.

[0134] In the above description, steps S21 to S22 can be further divided into additional steps or combined into fewer steps according to the embodiments of the present invention. Furthermore, some steps can be omitted or the order of the steps can be changed as needed.

[0135] Figure 8A detailed workflow diagram for the analysis process used to derive second analysis data related to the entity's behavior patterns.

[0136] Figure 8 The detailed analysis process for deriving second analysis data related to the entity's behavior patterns, as shown above, can be executed by the main server 100 described above. Therefore, even if the following description is omitted, the description related to the main server 100 can also be applied to entity analysis methods based on dual processing utilizing artificial intelligence.

[0137] Reference Figure 8 In step S221, the detailed analysis unit 120 can generate the dataset received from the boundary device 200, which includes at least a portion of the target data and the first analysis data, as the analysis object data of the second model.

[0138] Next, in step S222, the detailed analysis unit 120 may input the same dataset generated in step S221 into each of the multiple models included in the second model.

[0139] Subsequently, in step S223, the detailed analysis unit 120 can obtain the prediction results of multiple models.

[0140] Then, in step S224, the detailed analysis unit 120 can verify whether the prediction results derived from multiple models are the same.

[0141] Based on the judgment result of step S244, if the prediction results derived from multiple models are all the same, in step S225, the detailed analysis unit 120 can determine the same prediction results derived from multiple models as the second analysis data of the second model.

[0142] In contrast, based on the judgment result of step S224, if at least some of the prediction results derived from multiple models are different, then in step S226, the detailed analysis unit 120 removes the existing prediction results derived from multiple models respectively, and can perform detailed analysis again using the data re-received from the boundary device 200 (in other words, the target data and the first analysis data).

[0143] In the above description, steps S221 and S226 can be further divided into additional steps or combined into fewer steps according to the embodiments of the present invention. Furthermore, some steps can be omitted or the order of the steps can be changed as needed.

[0144] An embodiment of the entity analysis method based on dual processing utilizing artificial intelligence according to an embodiment of the present invention can be implemented in the form of program instructions executable by a computer device and recorded on a computer-readable medium. The aforementioned computer-readable medium may individually or in combination include program instructions, data files, data structures, etc. The program instructions recorded on the aforementioned medium are specifically designed for the present invention, or may be known instructions that can be used by a person skilled in the art of computer software. For example, the computer-readable recording medium may include magnetic media such as hard disks, floppy disks, and magnetic disks; optical media such as CD-ROMs and DVDs; magneto-optical media such as floppy disks; and hardware devices specifically designed for storing and executing program instructions, such as read-only memory (ROM), random access memory (RAM), and flash memory. For example, in addition to machine language code generated by a compiler, the program instructions may include high-level language code that a computer can execute through an interpreter, etc. To perform the work of the present invention, the aforementioned hardware device may be used as one or more software modules, or vice versa.

[0145] Furthermore, the aforementioned entity analysis method based on dual processing utilizing artificial intelligence can also be implemented in the form of a computer-executable program or application stored on a recording medium.

[0146] The above description of the present invention is merely illustrative. Those skilled in the art should understand that modifications can be easily made through other embodiments without altering the technical concept or essential features of the invention. Therefore, the embodiments described above are merely illustrative at all levels and should not be construed as limiting. For example, the structural elements described as a single type can be implemented separately, and similarly, the dispersed structural elements can be implemented in combination.

[0147] Compared with the detailed description above, the scope of this invention should be based on the scope of the claims. All modifications or variations of the embodiments derived from the meaning, scope and equivalent concepts of the scope of the claims are within the scope of this invention.

Claims

1. An analysis method, which is an entity analysis method based on dual processing utilizing artificial intelligence, characterized in that, The analytical method includes the following steps: Target data is collected using edge devices, including image and sensor data related to the entities being monitored. The boundary device is used to perform an initial analysis, inputting the target data into a first model based on pre-learned artificial intelligence to output first analytical data related to the entity's behavior pattern; and When the first analysis data meets the preset detailed analysis conditions, the first analysis data and the target data are transmitted to the main server. The main server performs a detailed analysis of the second analysis data related to the entity's behavior patterns by inputting at least one of the first analysis data and the target data into a second model based on pre-learned artificial intelligence.

2. The analytical method according to claim 1, characterized in that, The first analysis data includes information on the presence or absence of the entity and its location reflected in the image data.

3. The analytical method according to claim 1, characterized in that, The first analysis data includes first classification information related to the behavior type of the entity. The detailed analysis conditions include conditions corresponding to the first classification information and preset abnormal or important behaviors.

4. The analytical method according to claim 3, characterized in that, The second analysis data includes second classification information related to the behavior type of the entity, and the number of categories in the second classification information is greater than the number of categories in the first classification information.

5. The analytical method according to claim 1, characterized in that, The first model is a lightweight model constructed taking into account the resource information of the boundary device.

6. The analytical method according to claim 1, characterized in that, The second model includes multiple models pre-built based on multiple different network structures. The second analytical data is determined by considering the consistency of the output data of the various models.

7. An analysis method, which is an entity analysis method based on dual processing utilizing artificial intelligence, characterized in that, The analytical method includes the following steps: The initial analysis results are obtained by using a boundary device to input target data, including image data and sensor data related to the entity being monitored, into a first model based on pre-learned artificial intelligence, and outputting first analysis data related to the action patterns of the entity. When the first analysis data meets preset detailed analysis conditions, the first analysis data and the target data are received from the boundary device. Perform a detailed analysis of the second analytical data related to the entity's behavior patterns by inputting at least one of the first analytical data and the target data into a second model based on pre-learned artificial intelligence.

8. The analytical method according to claim 7, characterized in that, The second model includes multiple models pre-built based on multiple different network structures. The detailed analysis step determines the second analysis data by considering the consistency of the output data of the various models.

9. The analytical method according to claim 7, characterized in that, The first analysis data includes information on the presence or absence of the entity and its location reflected in the image data.

10. The analytical method according to claim 7, characterized in that, The first analysis data includes first classification information related to the behavior type of the entity. The detailed analysis conditions include conditions corresponding to the first classification information and preset abnormal or important behaviors.

11. The analytical method according to claim 10, characterized in that, The second analysis data includes second classification information related to the behavior type of the entity, and the number of categories in the second classification information is greater than the number of categories in the first classification information.

12. A smart farm monitoring system, characterized in that, The smart farm monitoring system includes: A boundary device collects target data, including image data and sensor data related to an entity being monitored; performs initial analysis by inputting the target data into a first model of pre-learned artificial intelligence to output first analysis data related to the entity's behavior patterns; and when the first analysis data meets preset detailed analysis conditions, transmits the first analysis data and the target data to a main server. The main server performs a detailed analysis of the second analysis data related to the entity's behavior patterns by inputting at least one of the first analysis data and the target data into a second model based on pre-learned artificial intelligence.

Images