An irrigation decision system based on edge multi-modal perception fusion

Abstract

The application relates to the technical field of edge-end data processing, in particular to an irrigation decision system based on edge-end multi-modal perception fusion. The system comprises a data acquisition module, a window processing module, a recovery judgment module, an irrigation mode determination module, a residual pressure processing module and an instruction generation module arranged at the edge end; the data acquisition module acquires soil moisture state, crown layer state, valve state and pipeline pressure state data of a target irrigation partition; the window processing module intercepts data in an afternoon window, a sunset window and a morning window and generates an intraday backwater sequence; the recovery judgment module judges whether the target irrigation partition completes recovery in a morning low transpiration state; the irrigation mode determination module determines an irrigation mode; the residual pressure processing module corrects the irrigation mode or generates a pipeline abnormality prompt according to pipeline residual pressure decay data after the valve is closed; and the instruction generation module generates an irrigation decision instruction; the application can be used for edge-end partition irrigation decision.

Claims

1. An edge-based multi-modal perception fusion based irrigation decision system, characterized in that, This includes the following modules located at the edge: The data acquisition module is used to acquire multimodal sensing data of the target irrigation zone. The multimodal sensing data includes soil moisture status data, canopy status data, valve status data, and pipeline pressure status data. The multimodal sensing data also includes air temperature and humidity data, soil temperature data, and pipeline flow data. The soil moisture status data includes shallow soil moisture data and root zone soil moisture data. The canopy status data includes canopy temperature data and leaf posture data. The window processing module is used to extract window data of the target irrigation zone from the multimodal sensing data within the afternoon window, sunset window, and early morning window, respectively, and generate an intraday water return sequence containing water change items, canopy change items, and pipeline pressure change items according to the time sequence of the afternoon window, the sunset window, and the early morning window; the intraday water return sequence includes soil shallow water change items, soil root water change items, canopy state change items, and pipeline pressure change items corresponding to the afternoon window, the sunset window, and the early morning window, respectively; the window processing module generates the soil shallow water change items, the soil root water change items, the canopy state change items, and the pipeline pressure change items based on the difference in representative values ​​of adjacent windows; the canopy state change items include canopy... The temperature decrease from the afternoon window to the early morning window, and the leaf posture recovery from the sunset window to the early morning window; wherein, the decrease in canopy temperature is determined based on the difference between the representative value of canopy temperature in the afternoon window and the representative value of canopy temperature in the early morning window. When the difference is not less than the preset canopy temperature decrease threshold corresponding to the target irrigation zone, the canopy temperature is determined to be a decrease item; the recovery of leaf posture is determined based on the deviation of the representative values ​​of leaf posture in the sunset window and the representative values ​​of leaf posture in the early morning window from the preset normal posture range. When the decrease in the deviation is not less than the preset posture recovery threshold corresponding to the target irrigation zone, or when the representative value of leaf posture in the early morning window enters the preset normal posture range, the leaf posture is determined to be a recovery item; The recovery judgment module is used to determine whether the target irrigation zone has completed recovery under the low evaporation state in the early morning based on the intraday water return sequence. An irrigation method determination module is used to determine the irrigation method of the target irrigation zone based on the judgment result of the recovery judgment module and the intraday return water sequence; The residual pressure processing module is used to correct the irrigation method or generate pipeline abnormality prompts based on the pipeline residual pressure decay data after the valve is closed. The instruction generation module is used to generate irrigation decision instructions based on the processing results of the residual pressure processing module.

2. The irrigation decision-making system based on edge-end multimodal perception fusion according to claim 1, characterized in that, The window processing module determines the afternoon window, the sunset window, and the early morning window based on light intensity, air temperature, and air humidity. The afternoon window corresponds to a time period in which both light intensity and air temperature are higher than the adjacent time periods of the day. The sunset window corresponds to a time period in which light intensity continuously decreases and air temperature begins to decrease. The early morning window corresponds to a time period in which light intensity is lower than a preset light threshold and air temperature is in the low range of the day.

3. The irrigation decision-making system based on edge-end multimodal perception fusion according to claim 1, characterized in that, The recovery judgment module determines that the target irrigation zone has completed recovery when the canopy temperature is determined to be decreasing, the leaf posture is determined to be recovering, and the soil root moisture change is determined to be maintaining. The recovery judgment module determines that the target irrigation zone has not completed recovery when the canopy temperature is not determined to be a decreasing item, the leaf posture is not determined to be a recovering item, or the soil root moisture change item is determined to be a decreasing item.

4. The irrigation decision-making system based on edge-end multimodal perception fusion according to claim 3, characterized in that, The irrigation method determination module determines the morning irrigation method when the soil root moisture change item is decreasing and the canopy state change item does not indicate a decrease in canopy temperature and recovery of leaf posture; it determines the low-flow, long-duration irrigation method when the soil shallow moisture change item is not decreasing and the soil root moisture change item is decreasing; it determines the delayed irrigation method when the target irrigation zone has recovered and there is an abnormal canopy state within the afternoon window; and it determines the paused irrigation method when the pipeline pressure change item is abnormal.

5. The irrigation decision-making system based on edge-end multimodal perception fusion according to claim 1, characterized in that, The pipeline residual pressure attenuation data includes the initial pressure value after the valve is closed, the pressure drop magnitude, and the pressure drop duration. When the pressure drop magnitude is greater than the preset attenuation range, the residual pressure processing module limits the increase in irrigation duration and generates a leakage warning. When the pressure drop magnitude is less than the preset attenuation range and the soil root layer moisture does not change accordingly, the residual pressure processing module pauses the output of irrigation decision commands and generates a pipeline blockage warning.

6. The irrigation decision-making system based on edge-end multimodal perception fusion according to claim 1, characterized in that, The irrigation decision instruction includes the target irrigation zone number, irrigation start time, irrigation duration, irrigation segment interval, irrigation flow rate level, and abnormal prompt information; wherein, the irrigation start time is determined by the judgment result of the recovery judgment module, the irrigation duration is determined by the soil root layer moisture change item, and the irrigation segment interval and the irrigation flow rate level are determined by the difference between the soil shallow layer moisture change item and the soil root layer moisture change item.

Images