Search

CN-122020253-A - Farm plant universe intelligent monitoring system fused with Internet of things

CN122020253ACN 122020253 ACN122020253 ACN 122020253ACN-122020253-A

Abstract

The invention discloses a global intelligent monitoring system of a farm, which relates to the technical field of intelligent agriculture, and comprises a quantitative evaluation module, a comprehensive attribution reporting module, a piglet aggregation behavior prediction module, a piglet aggregation area spatial concentration prediction module and a control strategy module, wherein the quantitative evaluation module is used for acquiring historical motion track data of piglet individuals of the farm, extracting multidimensional space-time characteristics of piglet individual motion, establishing a piglet individual full-time period dynamic association network, quantitatively evaluating piglet area spatial concentration in a time period, the comprehensive attribution reporting module is used for establishing a piglet aggregation behavior prediction model, predicting piglet future period group spatial concentration, calculating an aggregation area risk index, combining the dynamic association network to identify key piglet individuals, introducing compensation factors for carrying out association analysis, generating a piglet individual behavior-environment condition-resource distribution comprehensive attribution report, and the intervention strategy module is used for integrating the piglet future period group spatial concentration and the comprehensive attribution report, judging future aggregation risk level and generating a farm intervention strategy. The method and the device realize accurate prediction of the future aggregation risk.

Inventors

  • SHAN PEI
  • CHEN DAOLI

Assignees

  • 深圳市朗锐恒科技开发有限公司

Dates

Publication Date
20260512
Application Date
20260202

Claims (10)

  1. 1. Farm plant's universe intelligent monitoring system that thing networking fused, its characterized in that includes: The system comprises a quantitative evaluation module, a comprehensive attribution report module and an intervention strategy module; The comprehensive attribution reporting module is electrically connected with the quantitative evaluation module, and the intervention strategy module is electrically connected with the comprehensive attribution reporting module; the quantitative evaluation module is used for acquiring individual historical motion trail data of piglets in a farm, extracting individual motion multidimensional space-time characteristics of the piglets, establishing an individual full-time dynamic association network of the piglets, and quantitatively evaluating the regional space concentration degree of the piglets in a time-division manner; The comprehensive attribution reporting module is used for establishing a piglet aggregation behavior prediction model, predicting the piglet future period group space aggregation degree, calculating an aggregation region risk index, combining with a piglet individual full-period dynamic association network to identify key piglet individuals causing abnormal aggregation behavior, introducing an environmental state and a resource allocation variable as compensation factors to perform association analysis, and generating a piglet individual behavior-environmental condition-resource distribution comprehensive attribution report; The intervention strategy module integrates the regional space aggregation degree of the group in the future period of the piglet and the comprehensive attribution report of individual behavior of the piglet, environmental conditions and resource distribution, judges the future aggregation risk level of the farm, and automatically generates the self-adaptive hierarchical zoning intervention strategy of the farm.
  2. 2. The intelligent monitoring system for the universe of the farm integrated by the Internet of things according to claim 1, wherein the quantitative evaluation module comprises: the individual historical motion trail data unit is used for acquiring individual IDs and real-time positions of piglets based on low-power-consumption blue tooth neck marks worn by the piglets, establishing three-dimensional space coordinates of the piglets by taking a farm angle as an origin, registering the three-dimensional space coordinates with the coordinates by unifying time stamps, removing noise and abnormal values, and acquiring historical motion trail data of the piglets; The space motion feature extraction unit is used for setting a fixed sliding time window based on piglet individual historical motion track data, calculating piglet individual moving speed, acceleration and angular velocity of each sampling point in the time window, counting statistics of mean and variance of the piglet individual moving speed, acceleration and angular velocity as piglet individual motion features, calculating piglet individual moving area in the time window according to piglet individual historical motion track data by a convex hull method, counting area access frequency and average residence time in the time window, evaluating piglet gathering space preference in each area, and extracting piglet individual space motion features; And the multidimensional space-time characteristic unit is used for counting the times and duration of aggregation of piglet individuals and other piglet individuals in the same area in a time window based on the historical movement track data of the piglet individuals, extracting the social characteristics of the piglet individuals, and integrating the kinematic characteristics of the piglet individuals and the spatial movement characteristics of the piglet individuals to obtain the multidimensional space-time characteristics of the piglet individual movement.
  3. 3. The intelligent monitoring system for farm farms integrated by the internet of things according to claim 2, wherein the quantitative evaluation module further comprises: a space proximity strength unit, setting a distance threshold value, calculating a space distance matrix of all piglets in a time window, and counting the accumulated time length occupation ratio of each pair of piglet space distances smaller than the threshold value as the inter-individual space proximity strength of the piglets; The comprehensive motion similarity unit is used for calculating the pearson correlation coefficient of the motion direction angle sequence among piglet individuals in the time window based on the historical motion track data of the piglet individuals, taking the pearson correlation coefficient as the direction similarity among piglet individuals, and calculating the comprehensive motion similarity by combining the speed vector similarity among piglet individuals; calculating the piglet cosine-to-cosine similarity of the piglet relative to the speed vector at any moment, and taking an average value at all moments in a time window to obtain the piglet inter-individual speed vector similarity; The dynamic association network building unit takes each piglet as a node, takes individual spatial motion characteristics of the piglets as node attributes, calculates the inter-piglet association strength weight by weighting and fusing the inter-piglet spatial proximity strength and the comprehensive motion similarity, takes the inter-piglet association strength weight as an edge weight, builds an individual full-period dynamic association network of the piglets, learns the time evolution mode of the nodes and the edges, and identifies the inter-piglet social role change.
  4. 4. The intelligent monitoring system for farm farms integrated with the internet of things according to claim 3, wherein the quantitative evaluation module further comprises: The computing unit of the Morgan index, divide the farm into n m size regular grids, according to the individual full-time dynamic association network of piglets, count the number of piglets in each grid in the time window, obtain the preliminary space density of piglets of each grid, calculate the global space density of piglets of the farm, utilize Morgan index to calculate the formula, evaluate the overall gathering trend of piglet space distribution of the farm quantitatively, calculate the local Morgan index and discern the local hot spot area, judge the grid with the local Morgan index greater than 0 as the high-high gathering hot spot area, judge the grid with the local Morgan index less than 0 as the low-low gathering cold spot area; If the Morgan index of a certain region is larger than 0, the region is judged to be piglet aggregation distribution, if the Morgan index of the certain region is smaller than 0, the piglet dispersion distribution of the region is judged, and if the Morgan index of the certain region is equal to 0, the piglet random distribution of the region is judged.
  5. 5. The intelligent monitoring system for farm farms integrated with the internet of things according to claim 4, wherein the quantitative evaluation module further comprises: And the clustering unit is used for carrying out clustering analysis on the screened areas by taking the grid central coordinates of the local hot spot areas as input and utilizing a DBSCAN density clustering algorithm, identifying piglet clusters which are in spatial communication, determining the central position of each cluster, calculating the ratio of the average piglet density of each cluster to the total area covered by the clusters to the total area of the farm to obtain piglet aggregation indexes, and generating a piglet aggregation degree thermodynamic diagram.
  6. 6. The intelligent monitoring system for the universe of the farm integrated by the Internet of things according to claim 1, wherein the comprehensive attribution reporting module comprises the following components: The prediction unit is used for acquiring piglet aggregation indexes of past h time windows based on a farm database to establish a sequence, analyzing the aggregation conditions of piglets in different time periods, combining social features, environmental features and resource features as input, establishing a piglet aggregation behavior prediction model, learning a piglet aggregation index change rule along with time and an influence mode of each feature factor on piglet aggregation behavior, and predicting the piglet future period group space aggregation degree; the social characteristics are that according to a piglet individual full-time dynamic association network, learning nodes and edges evolve modes along with time, and identifying the social role change among piglet individuals; the environmental characteristics are time sequence data of temperature and humidity environmental states of the past h time windows; The resource characteristics are specifically time sequence data of recent usage rates of a trough and a water trough in past h time windows.
  7. 7. The intelligent monitoring system for farm farms integrated with the internet of things according to claim 6, wherein the comprehensive attribution reporting module further comprises: The risk level unit is used for calculating the deviation between the current temperature and humidity of the region and the optimal region of the piglet based on the regional space aggregation degree of the group of the piglet in the future period of prediction, calculating the residual resource quantity and queuing condition of the trough and the trough near the aggregation point as the sufficient resource coefficient, fusing the sufficient resource coefficient and the sufficient resource coefficient as compensation factors, calculating the regional risk index, setting the low, medium and high three-level risk threshold values and generating the aggregation cluster risk level; The screening unit is used for acquiring full-time dynamic association networks of piglet individuals in the first J time windows aiming at the clustering cluster marked as high risk level, calculating the number of other piglet individuals directly connected in the network by each piglet individual in the time window, and identifying piglet individuals playing a role in the network in the time window to obtain A candidate key piglet individuals causing abnormal clustering behaviors; The individual identification unit sets candidate individuals which actually enter the area earliest as an active state based on the initial moments of the first J time windows, sets other key piglet individuals as an inactive state, iterates along time steps, aims at the active node at the last moment of each time step to activate the inactive neighbor node to the active state, establishes a piglet influence propagation model, obtains piglet individuals with the highest activation success probability, and identifies the key piglet individuals which cause abnormal aggregation behavior; the active state is defined as having entered a high risk prediction aggregation area; The judgment standard of whether the activation is successful or not is that the activated node actually enters the high risk aggregation area in the subsequent time step, and the identification basis of the key individual is that the matching degree of the successfully activated node set and the real aggregation cluster members is highest when the key individual is used for simulating a piglet individual.
  8. 8. The intelligent monitoring system for farm farms integrated with the internet of things according to claim 7, wherein the comprehensive attribution reporting module further comprises: The report generation unit is used for calculating SHAP values of social features, environmental features and resource features based on the clustered clusters marked as high risk levels, performing time synchronization and alignment, calculating contribution degree Shapley values of each factor to the clustered risk, quantifying the contribution degree of each feature to the clustered cluster prediction of the high risk levels, and performing integrated analysis on the contribution degree of each feature to the clustered clusters of the high risk levels, the environmental deviation coefficient and the resource sufficiency coefficient evaluation result of the clustered clusters of the clustered risk levels and key piglet individuals causing abnormal clustered behaviors to generate piglet individual behavior-environmental condition-resource distribution comprehensive attribution reports.
  9. 9. The intelligent monitoring system for the universe of the farm integrated by the Internet of things according to claim 8, wherein the intervention strategy module specifically comprises: The generation unit takes regional space aggregation degree and regional risk index of the piglet future period group as input, combines piglet individual behavior-environmental condition-resource distribution comprehensive attribution report, judges future aggregation risk level of the farm according to set low, medium and high three-level risk threshold values, and automatically generates self-adaptive hierarchical zoning intervention strategy of the farm.
  10. 10. The internet of things fused farm universe intelligent monitoring system of claim 9, wherein the intervention strategy module further comprises: The low risk unit judges that the region is in a low risk level if the regional space aggregation degree of the group of piglets in the future period is low and the regional risk index is less than or equal to a low threshold value, marks the region as blue, and records real-time data; The medium risk unit judges that the region is in a medium risk level if the regional space aggregation degree of the group of piglets in the future period is medium and the low threshold value is less than or equal to the regional risk index and less than or equal to the medium threshold value, marks the region as yellow, automatically calculates and executes compensatory environmental parameters to adjust the temperature rise and fall and adjust ventilation according to the direction and the size of the environmental deviation coefficient, generates backlogs in a management background, and reminds management personnel to check the heat preservation facilities; The high risk unit judges that the region is in a high risk level if the regional space aggregation degree of the group of piglets in the future period is high and the regional risk index is more than or equal to a high threshold value, marks the region as orange, automatically highlights the individual IDs and positions of key piglets triggering the risk, and recommends individual inspection or guidance; And the emergency risk unit judges that the region is in an emergency risk level if the regional space aggregation degree of the group of piglets in the future period is extremely high and the regional risk index is extremely high, marks the region as red, automatically instructs to start a standby drinking water point or trough, informs a feeder to perform manual intervention to the aggregation region, and ensures sufficient resource supply.

Description

Farm plant universe intelligent monitoring system fused with Internet of things Technical Field The invention relates to the technical field of intelligent agriculture, in particular to an intelligent monitoring system for the universe of a farm integrated by the Internet of things. Background The traditional farm breeding monitoring system is independent of an isolated environment sensor or manual inspection, lacks of real-time and continuous quantitative analysis of animal group dynamic behaviors, is difficult to capture space-time correlation characteristics of piglet individuals and groups, is generally based on a fixed threshold value of a single index, causes lag in risk judgment and is attributed to one side, and the prior art cannot effectively integrate multi-source heterogeneous data such as individual behavior tracks, environment states and resource distribution to comprehensively attribute, so that intervention measures are always passive and general, does not have differentiation and self-adaptive execution capacity based on risk grades, and severely restricts the accuracy and timeliness of breeding risk management. Disclosure of Invention In order to solve the technical problems, the intelligent monitoring system for the whole domain of the farm integrated by the Internet of things is provided, and the technical scheme solves the problems that in the traditional cultivation monitoring, risk early warning is lagged, due to one-sided and intervention measures are single, and self-adaptive accurate response is not available. In order to achieve the above purpose, the invention adopts the following technical scheme: an intelligent farm universe monitoring system integrated by the internet of things, comprising: The system comprises a quantitative evaluation module, a comprehensive attribution report module and an intervention strategy module; The comprehensive attribution reporting module is electrically connected with the quantitative evaluation module, and the intervention strategy module is electrically connected with the comprehensive attribution reporting module; the quantitative evaluation module is used for acquiring individual historical motion trail data of piglets in a farm, extracting individual motion multidimensional space-time characteristics of the piglets, establishing an individual full-time dynamic association network of the piglets, and quantitatively evaluating the regional space concentration degree of the piglets in a time-division manner; The comprehensive attribution reporting module is used for establishing a piglet aggregation behavior prediction model, predicting the piglet future period group space aggregation degree, calculating an aggregation region risk index, combining with a piglet individual full-period dynamic association network to identify key piglet individuals causing abnormal aggregation behavior, introducing an environmental state and a resource allocation variable as compensation factors to perform association analysis, and generating a piglet individual behavior-environmental condition-resource distribution comprehensive attribution report; The intervention strategy module integrates the regional space aggregation degree of the group in the future period of the piglet and the comprehensive attribution report of individual behavior of the piglet, environmental conditions and resource distribution, judges the future aggregation risk level of the farm, and automatically generates the self-adaptive hierarchical zoning intervention strategy of the farm. Preferably, the quantitative evaluation module specifically includes: the individual historical motion trail data unit is used for acquiring individual IDs and real-time positions of piglets based on low-power-consumption blue tooth neck marks worn by the piglets, establishing three-dimensional space coordinates of the piglets by taking a farm angle as an origin, registering the three-dimensional space coordinates with the coordinates by unifying time stamps, removing noise and abnormal values, and acquiring historical motion trail data of the piglets; The space motion feature extraction unit is used for setting a fixed sliding time window based on piglet individual historical motion track data, calculating piglet individual moving speed, acceleration and angular velocity of each sampling point in the time window, counting statistics of mean and variance of the piglet individual moving speed, acceleration and angular velocity as piglet individual motion features, calculating piglet individual moving area in the time window according to piglet individual historical motion track data by a convex hull method, counting area access frequency and average residence time in the time window, evaluating piglet gathering space preference in each area, and extracting piglet individual space motion features; And the multidimensional space-time characteristic unit is used for counting the times and duration of aggregation of