CN-122020095-A - Intelligent climate intelligent tea garden intelligent monitoring system

CN122020095ACN 122020095 ACN122020095 ACN 122020095ACN-122020095-A

Abstract

The invention relates to the technical field of ecological environment monitoring and information fusion quantization, and discloses a climate intelligent tea garden intelligent monitoring system which comprises a sensing module, an edge consistency module, an inference module and an active monitoring module, wherein the sensing module is used for collecting multi-source observation data and agronomic event data of tea garden plots, the edge consistency module is used for carrying out consistency detection on the multi-source observation data, the inference module is used for carrying out dynamic assimilation estimation on hidden states of the tea garden under the condition of introducing observation credibility, and the active monitoring module is used for calculating expected information gain of uncertainty reduction caused by different collection actions based on state uncertainty information output by the inference module, and accordingly generating a collection scheduling strategy at the next moment. According to the method, in the edge consistency module, time window statistical references are respectively constructed for the multi-source observed data marked at the same time, and the consistency violation degree is calculated based on the cross-sensor consistency constraint set, so that abnormal or low-consistency observed data is prevented from directly participating in state estimation.

Inventors

HU JUNMING
WU ZEHUA
ZHENG FUHAI
HU JINRUI
WU QIANTAO
LI YUXIANG
LI TINGTING
ZHANG JUNHUI
Liu Meikun
TAN JIANJIAN

Assignees

广西壮族自治区农业科学院

Dates

Publication Date: 20260512
Application Date: 20260319

Claims (10)

1. An intelligent climate monitoring system for a tea garden, comprising: The sensing module is used for collecting multi-source observation data and agronomic event data of tea garden plots, wherein the multi-source observation data comprises environment sensor data and tea garden image characteristic data; The edge consistency module is used for carrying out consistency detection on the multi-source observation data and generating corresponding observation credibility for the multi-source observation data at each moment; The inference module is used for carrying out dynamic assimilation estimation on the hidden state of the tea garden under the condition of introducing the observation credibility, and generating a tea garden risk assessment result and a corresponding causal evidence chain based on a pre-constructed causal map and a constraint sign regression model; And the active monitoring module is used for calculating the expected information gain of different acquisition actions on the uncertainty reduction based on the state uncertainty information output by the inference module, and generating an acquisition scheduling strategy at the next moment according to the expected information gain.
2. The intelligent climate tea garden monitoring system according to claim 1, wherein in the edge consistency module, the step of consistency detecting the multi-source observation data comprises: Acquiring multisource observation data of the same land block and under the same time mark; constructing a corresponding time window statistical benchmark for each type of observation data; Calculating the deviation amount between each observation data and the corresponding time window statistical standard; Based on a preset cross-sensor consistency constraint set, carrying out consistency constraint detection on multi-source observation data marked at the same time to obtain a consistency violation degree; calculating the observation credibility under the corresponding time mark according to the deviation amount and the consistency violation degree; By setting a threshold range of observation credibility, the numerical stability of the state updating process is ensured.
3. The intelligent climate tea garden monitoring system according to claim 2, wherein the step of performing consistency constraint detection on the multi-source observation data under the same time mark based on a preset set of cross-sensor consistency constraints comprises: respectively setting physical reasonable value range constraints for multiple environmental sensor data in the same land block; Setting monotone relation constraint for multiple environmental sensor data with physical or physiological association relation; Setting a combination consistency constraint aiming at the combination relation of at least two types of environmental sensor data; unifying the physical reasonable value range constraint, the monotonic relation constraint and the combined consistency constraint to form the cross-sensor consistency constraint set.
4. The intelligent climate tea garden monitoring system according to claim 1, wherein in the inference module, the step of constructing a causal map comprises: constructing a climate driving variable set, a soil process variable set, a tea tree phenotype variable set and a management intervention variable set; respectively taking variables in the climate driving variable set, the soil process variable set, the tea tree phenotype variable set and the management intervention variable set as nodes of a causal map; Establishing allowed directional connection relations between the nodes according to a preset causal direction rule; Recording the allowed directed connections as a causal adjacency matrix, forming the causal graph.
5. The intelligent climate monitoring system according to claim 1, wherein in the inference module, the step of dynamically assimilating and estimating the hidden state of the tea garden comprises: Establishing a state space model comprising state variables and observation variables; introducing the agronomic event data as external input items into the state space model; according to the physical period of the current tea garden land block, adjusting state transition parameters or process noise distribution parameters in the state space model; In a state prediction stage, introducing the external input item to perform state prediction on the state variable; in a state updating stage, introducing the observation credibility into a construction process of observation noise covariance, and carrying out weighted updating on observation data; and outputting a state estimation result and a state covariance matrix under the corresponding time mark.
6. The intelligent climate tea garden monitoring system according to claim 1, wherein in the inference module, the step of generating tea garden risk assessment results and corresponding causal evidence links based on pre-constructed causal graph and constrained symbolic regression models comprises: determining a causal map node corresponding to the target risk variable; extracting a candidate input variable set with causal parent-child relation with the target risk variable based on the causal adjacency matrix; Constructing a symbol expression search space based on the candidate input variable set; Generating a candidate expression within the symbol expression search space; Applying a structure constraint rule to the structure of the candidate expression when generating the candidate expression, wherein the structure constraint rule at least comprises defining the maximum expression tree depth of the candidate expression, defining the operator priority order and defining the maximum interaction order between candidate input variables; And screening the candidate expressions based on a preset constraint set to obtain a risk assessment expression.
7. The intelligent climate tea garden monitoring system according to claim 6, wherein the step of screening the candidate expressions based on a predetermined set of constraints to obtain risk assessment expressions comprises: Setting monotonicity constraint on at least one type of candidate input variable; setting a non-negative constraint on the target risk variable; setting an output range constraint on a target risk variable; setting a minimum effective lag time constraint on management intervention variables; And unifying the monotonicity constraint, the nonnegative constraint, the output range constraint and the minimum effective lag time constraint to form the preset constraint set.
8. The intelligent climate tea garden monitoring system according to claim 1, wherein in the inference module, the step of generating causal evidence links comprises: Determining a target node corresponding to a target risk variable in the causal map; searching the causal graph for a set of directed paths directed by candidate root nodes to the target node; Constructing a corresponding path node sequence for each directed path; calculating the path contribution degree of each directed path based on the local sensitivity between each path node; Performing intervention calculation on the variables corresponding to the candidate root cause nodes based on intervention analysis or inverse fact reasoning to obtain a target risk variable change result under the intervention condition, and performing consistency verification on the path contribution degree corresponding to the directed path based on the change result; And outputting the causal evidence chain according to the path contribution degree of each directional path.
9. The intelligent climate tea garden monitoring system according to claim 1, wherein in the active monitoring module, the step of calculating desired information gain for uncertainty reduction by different collection actions based on the state uncertainty information output by the inference module comprises: constructing an acquisition action set comprising a plurality of candidate acquisition actions; Based on the state covariance matrix, respectively estimating uncertainty change amounts corresponding to each candidate acquisition action; Calculating expected information gain of each candidate acquisition action based on the uncertainty change; And under the condition that the resource constraint condition is met, selecting the candidate acquisition action with the maximum expected information gain as an acquisition scheduling strategy at the next moment.
10. The intelligent climate tea garden monitoring system according to claim 9, wherein in the active monitoring module, the step of constructing the collection of collection actions comprising a plurality of candidate collection actions comprises: constructing an environment sensor sampling frequency adjustment action; constructing a tea garden image acquisition device to trigger actions; constructing a specified land block supplementary inspection task action; And uniformly forming the sampling frequency adjustment action of the environment sensor, the triggering action of the tea garden image acquisition equipment and the action of the specified land block supplementary inspection task into the acquisition action set.

Description

Intelligent climate intelligent tea garden intelligent monitoring system Technical Field The invention relates to the technical field of ecological environment monitoring and information fusion quantification, in particular to an intelligent climate monitoring system for a tea garden. Background With the development of climate change, internet of things technology, environment sensing technology and agricultural informatization technology, an intelligent monitoring system for modern tea garden production management is gradually applied to scenes such as tea garden environment monitoring, crop growth state monitoring and production risk assessment. The existing intelligent monitoring system for the tea garden generally adopts a multi-source sensing data fusion mode, uniformly processes observation data from different types of sensors and image processing modules, and estimates the internal state of the tea garden by constructing a state estimation model or a prediction model. In the prior art, after various sensor observation data and image characteristic data are aligned in the time dimension, the sensor observation data and the image characteristic data are directly used as observation input to be introduced into a state estimation model, and the fixed observation noise parameters are utilized to update the multi-source observation data in a weighting manner, so that an estimation result of the environmental state of the tea garden or the crop growth state is obtained. Therefore, the inventor finds that the prior art at least has the following technical problems in the process of realizing the information fusion quantification related technical scheme that under the condition that the multisource observation data have cross-sensor inconsistency and short-time abnormal fluctuation, a quantification mechanism for the consistency degree of the multisource observation data under the same time mark is lacked, so that the stability and the credibility of the tea garden hidden state estimation result are affected. Disclosure of Invention In order to make up for the defects, the invention provides an intelligent monitoring system for a climate intelligent tea garden, which aims to solve the problem that the prior art lacks a quantification mechanism for marking the consistency degree of multi-source observation data at the same time. The invention provides a technical scheme that an intelligent monitoring system for a climate intelligent tea garden comprises: The sensing module is used for collecting multi-source observation data and agronomic event data of tea garden plots, wherein the multi-source observation data comprises environment sensor data and tea garden image characteristic data; The edge consistency module is used for carrying out consistency detection on the multi-source observation data and generating corresponding observation credibility for the multi-source observation data at each moment; The inference module is used for carrying out dynamic assimilation estimation on the hidden state of the tea garden under the condition of introducing the observation credibility, and generating a tea garden risk assessment result and a corresponding causal evidence chain based on a pre-constructed causal map and a constraint sign regression model; And the active monitoring module is used for calculating the expected information gain of different acquisition actions on the uncertainty reduction based on the state uncertainty information output by the inference module, and generating an acquisition scheduling strategy at the next moment according to the expected information gain. Preferably, in the edge consistency module, the step of performing consistency detection on the multi-source observed data includes: Acquiring multisource observation data of the same land block and under the same time mark; constructing a corresponding time window statistical benchmark for each type of observation data; Calculating the deviation amount between each observation data and the corresponding time window statistical standard; Based on a preset cross-sensor consistency constraint set, carrying out consistency constraint detection on multi-source observation data marked at the same time to obtain a consistency violation degree; calculating the observation credibility under the corresponding time mark according to the deviation amount and the consistency violation degree; By setting a threshold range of observation credibility, the numerical stability of the state updating process is ensured. Preferably, the step of performing consistency constraint detection on the multi-source observation data under the same time mark based on a preset cross-sensor consistency constraint set includes: respectively setting physical reasonable value range constraints for multiple environmental sensor data in the same land block; Setting monotone relation constraint for multiple environmental sensor data with physical or physiological association