CN-122024076-A - Greenhouse cash crop growth monitoring method and system based on edge cloud cooperation

CN122024076ACN 122024076 ACN122024076 ACN 122024076ACN-122024076-A

Abstract

The application discloses a greenhouse cash crop growth monitoring method and system based on edge cloud cooperation, and relates to the technical field of intelligent agriculture, wherein the method comprises the steps of deploying a distributed multi-source fixed probe and establishing a single-greenhouse time sequence monitoring data set; the method comprises the steps of inputting a lightweight wavelet segmentation network to segment a crop area to obtain a single-shed crop weak change enhancement feature set, uploading the single-shed crop weak change enhancement feature set to a cloud, constructing a commonality feature library and a personalized feature library to obtain an area-level multi-shed fusion feature database, building an edge cloud collaborative global growth monitoring model, training to obtain a global general growth monitoring model, sending the global general growth monitoring model to an edge end, finely adjusting the global general growth monitoring model to obtain a single-shed personalized adaptation monitoring model, generating a single-shed growth monitoring report based on a single-shed time sequence monitoring data set reasoning, and generating a multi-shed growth state thermodynamic diagram and weak change incentive association diagram after integration to obtain a multi-shed collaborative monitoring and farming decision scheme. The application solves the problem of adaptability of greenhouse scenes and improves the recognition capability of weak growth changes.

Inventors

FENG LEILEI
CHEN YANG
YANG ZHENHUA
ZHANG WEN
XUE MINGKE

Assignees

陕西农林职业技术大学

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (9)

1. The greenhouse cash crop growth monitoring method based on edge cloud cooperation is characterized by comprising the following steps of: Collecting multispectral images through a fixed probe, and carrying out standardized pretreatment on the collected data to obtain a single-shed standardized time sequence monitoring data set; Inputting the single-shed standardized time sequence monitoring data set into a trained lightweight wavelet segmentation network for crop region segmentation, fusing data by a difference characteristic compensation module, and carrying out characteristic strengthening treatment by combining a channel attention mechanism to obtain a single-shed crop weak change strengthening characteristic set; Uploading the single-shed crop weak change enhancement feature set to a cloud end through a wireless network, constructing a common feature library and a personalized feature library, and performing normalization and feature fusion processing on data in the common feature library to obtain a regional multi-shed fusion feature database; establishing an edge cloud collaborative global growth monitoring model, and obtaining a regional global general growth monitoring model after training by utilizing the regional multi-shed fusion characteristic database; issuing the regional global general growth monitoring model to the edge end of a greenhouse, and fine-tuning the regional global general growth monitoring model by utilizing the single-greenhouse crop weak change enhancement feature set to obtain a single-greenhouse personalized adaptation monitoring model; inputting the single-shed standardized time sequence monitoring data set into the single-shed personalized adaptive monitoring model for real-time reasoning, and finishing crop growth state classification and weak change area space positioning to generate a single-shed refined growth monitoring report; Integrating the single-shed refined growth monitoring report of each shed at the cloud end, generating a regional multi-shed growth state thermodynamic diagram and a weak change incentive association diagram, and further generating an agronomic operation suggestion to obtain a regional monitoring result.
2. The greenhouse cash crop growth monitoring method based on edge cloud cooperation as claimed in claim 1, wherein the fixed probe further collects data of air temperature and humidity, soil moisture content and chlorophyll concentration, and the standardized preprocessing of the collected data comprises probe baseline calibration, time stamp synchronization, radiation consistency correction and noise removal.
3. The greenhouse cash crop growth monitoring method based on edge cloud cooperation as claimed in claim 1, wherein the lightweight wavelet division network comprises an encoder and a decoder, the encoder adopts a 2-level wavelet decomposition structure, feature dimension reduction and extraction are completed through depth separable convolution, and the decoder adopts an up-sampling and feature fusion structure to fuse low-level detail features extracted by the encoder with high-level semantic features to obtain a binary mask of a crop area.
4. The greenhouse economic crop growth monitoring method based on edge cloud cooperation as claimed in claim 1, wherein the common feature library stores common data of multi-greenhouse crop growth, the individual feature library independently stores exclusive data of each greenhouse, and in the feature fusion process, the cloud performs normalization processing on the data in the common feature library, fuses common features of regional-level crop growth, and obtains the regional-level multi-greenhouse fusion feature database, wherein the data in the individual feature library keeps an original state.
5. The greenhouse economic crop growth monitoring method based on edge cloud cooperation as claimed in claim 1, wherein the backbone of the Bian Yun cooperation global growth monitoring model is a lightweight convolutional neural network, the lightweight convolutional neural network comprises an input layer, a common feature extraction layer and a growth state classification layer, the input layer receives multi-shed fusion feature data, the common feature extraction layer extracts general features of regional crop growth through 4 layers of lightweight convolutional units, and the growth state classification layer classifies crop growth states into four stages, namely health, slight stress, moderate stress and severe stress, and outputs crop growth state probability and weak change positioning coordinates.
6. The greenhouse cash crop growth monitoring method based on edge cloud coordination of claim 5, wherein when the regional global general growth monitoring model is fine-tuned, all parameters of the common feature extraction layer are frozen, and the growth state classification layer of the regional global general growth monitoring model is fine-tuned only by using the single greenhouse crop weak change enhancement feature set in the individual feature library.
7. The greenhouse economic crop growth monitoring method based on edge cloud cooperation according to claim 1, wherein in the process of generating the single greenhouse refined growth monitoring report, the single greenhouse personalized adaptation monitoring model generates original reasoning data based on the single greenhouse standardized time sequence monitoring data set, the original reasoning data comprises results of real-time growth state grading, weak change area positioning and stress type judgment, based on a visualization module built-in at the edge end, a greenhouse crop growth state distribution diagram and a crop time sequence growth trend curve are generated according to the original reasoning data, and the original reasoning data, the greenhouse crop growth state distribution diagram and the crop time sequence growth trend curve form the single greenhouse refined growth monitoring report.
8. The greenhouse cash crop growth monitoring method based on edge cloud cooperation as claimed in claim 1, wherein the cloud performs comparison analysis on the single greenhouse refined growth monitoring report from three dimensions of growth state comparison, weak change rule comparison and environment association comparison, and generates the regional multi-greenhouse growth state thermodynamic diagram and the weak change cause association diagram.
9. A system for applying the greenhouse cash crop growth monitoring method based on edge-cloud coordination as claimed in any one of claims 1 to 8, characterized in that the system comprises: the data acquisition module is a fixed probe and is used for acquiring multispectral images, and carrying out standardized pretreatment on the acquired data to obtain a single-shed standardized time sequence monitoring data set; The data enhancement module is used for inputting the single-shed standardized time sequence monitoring data set into a trained lightweight wavelet segmentation network to segment crop areas, fusing data through the difference characteristic compensation module, and carrying out characteristic enhancement processing by combining a channel attention mechanism to obtain a single-shed crop weak change enhancement characteristic set; the database establishing module is used for uploading the single-shed crop weak change enhancement feature set to the cloud end through a wireless network, constructing a common feature library and a personalized feature library, and performing normalization and feature fusion processing on data in the common feature library to obtain a regional multi-shed fusion feature database; The monitoring model construction module is used for establishing an edge cloud collaborative global growth monitoring model, and obtaining a regional global general growth monitoring model after training by utilizing the regional multi-shed fusion characteristic database; The model fine adjustment module is used for issuing the regional global general growth monitoring model to the edge end of the greenhouse, and fine adjustment is carried out on the regional global general growth monitoring model by utilizing the single-greenhouse crop weak change enhancement characteristic set to obtain a single-greenhouse personalized adaptation monitoring model; The report generation module is used for inputting the single-shed standardized time sequence monitoring data set into the single-shed personalized adaptive monitoring model to perform real-time reasoning, so as to complete crop growth state classification and weak change area space positioning and generate a single-shed refined growth monitoring report; And the report integration module is used for integrating the single-shed refined growth monitoring report of each shed at the cloud end, generating a regional multi-shed growth state thermodynamic diagram and a weak change incentive association diagram, and further generating an agricultural operation suggestion to obtain a regional monitoring result.

Description

Greenhouse cash crop growth monitoring method and system based on edge cloud cooperation Technical Field The application relates to the technical field of intelligent agriculture, in particular to a method for monitoring the growth state of economic crops in a greenhouse by adopting a side cloud cooperation technology. Background Facility agriculture is a core component of modern agriculture in China, and greenhouse economic crops become key carriers for increasing income of farmers and improving quality of agriculture by virtue of high-yield and out-of-season planting. The closed and controllable growth environment in the greenhouse ensures that the growth state of crops is highly sensitive to the change of microenvironments such as temperature and humidity, water and fertilizer, illumination and the like, and accurate, real-time and refined growth monitoring is a core premise for realizing intelligent planting in the greenhouse and improving the quality and yield of crops. The traditional greenhouse monitoring mainly comprises manual inspection, has the defects of low efficiency, strong subjectivity, incapability of timely finding early growth abnormality and the like, and is difficult to meet the requirement of large-scale facility agriculture, and along with the development of intelligent agricultural technology, intelligent technologies such as remote sensing monitoring, unmanned aerial vehicle inspection, sensor monitoring and the like are gradually applied to the field of crop growth monitoring, so that the intelligent greenhouse monitoring becomes the main stream direction of industrial research and application. Patent CN120375218a is a representative technology of current field crop growth monitoring, the technology uses an unmanned plane as a core data acquisition carrier, carries a multispectral sensor to obtain a time sequence remote sensing image of a large-area farmland, completes crop area segmentation and growth difference analysis through a deep learning network, and realizes automatic monitoring of states such as open-air farmland crop growth vigor, plant diseases and insect pests. The patent technology is designed aiming at the open scene of a large-area open farmland, relies on the large-scale cruising ability of an unmanned aerial vehicle, solves the problems of large field manual monitoring difficulty and wide range of pain points, and has the following two defects when being applied to monitoring of greenhouse cash crops: First, the weak growth change monitoring of crops is insensitive. The algorithm model of the patent is designed aiming at large-area and remarkable growth changes (such as whole drought and large-scale diseases and insect pests) of open farmland crops, the crops in a greenhouse are in a stable environment of manual regulation and control, the growth changes are gentle, the weak growth changes such as early fertilizer deficiency, slight leaf yellowing and early diseases and insect pests are monitoring cores, but the patent algorithm cannot capture the fine features, so that early growth abnormality detection is missed, and the optimal control and regulation time is missed. Second, the application scenario is severely mismatched. Unmanned aerial vehicle can't take off and land safely in the big-arch shelter, stably cruises, can't realize full canopy cover collection, just does not possess the feasibility of big-arch shelter deployment from the hardware level. Disclosure of Invention The embodiment of the application provides a greenhouse cash crop growth monitoring method and system based on edge cloud cooperation, which are used for solving the two problems in the prior art. In one aspect, the embodiment of the application provides a greenhouse cash crop growth monitoring method based on edge cloud cooperation, which comprises the following steps: deploying a distributed multi-source fixed probe in a greenhouse, collecting multispectral image, air temperature and humidity, soil moisture content and chlorophyll concentration data, and carrying out standardized pretreatment on the collected data to obtain a single-greenhouse standardized time sequence monitoring data set; inputting the single-shed standardized time sequence monitoring data set into a trained lightweight wavelet segmentation network for crop region segmentation, fusing multispectral images, chlorophyll concentration data and environmental data through a difference characteristic compensation module, and carrying out characteristic enhancement processing by combining a channel attention mechanism to obtain a single-shed crop weak change enhancement characteristic set; Uploading the single-shed crop weak change enhancement feature set to a cloud end through a wireless network, constructing a commonality feature library and a personalized feature library, and performing normalization and feature fusion processing on data in the commonality feature library to obtain a regional multi-shed fusion feature datab