Peach tree dry rot prediction method based on dynamic and static environment monitoring

Abstract

The invention discloses a peach tree dry rot prediction method based on dynamic and static environment monitoring, and the method comprises the steps: peach tree cultivation: the peach trees of a test group are cultivated based on a cross-ditch one-side cultivation mode; constructing a dynamic and static ring monitoring system; preprocessing the data; algorithm processing: processing the peach tree branch and leaf image data by using a branch and leaf distribution algorithm and obtaining a branch and leaf analysis result by the cloud server, processing the under-tree weed image data by using a weed distribution algorithm and obtaining a weed analysis result by the cloud server, and processing the soil moisture content data by using a cloud server; the cloud server carries out processing in a mode of comparing historical track data based on healthy soil of the peach tree, and obtains a soil moisture content analysis result; based on the branch and leaf analysis result, the weed analysis result and the soil moisture content analysis result, a dynamic and static prediction model about the peach tree dry rot is constructed in a weighted mode, and the prediction model is subjected to rot prediction.

Description

Technical Field

[0001] The invention relates to the field of peach tree trunk rot, and in particular to a peach tree trunk rot prediction method based on dynamic and static environment monitoring. Background Art

[0002] Peach rot is one of the main diseases of peach trees, also known as peach root rot. This disease can damage the root system of peach trees, which in turn affects the growth and development of the entire tree. Rot is usually caused by pathogenic microorganisms in the soil, and its main symptom is the appearance of brown rot spots on the roots.

[0003] The formation of peach rot lesions is related to environmental factors, pathogenic microorganisms and the disease resistance of peach trees themselves. First of all, environmental factors include excessive soil moisture, poor drainage, and the presence of pathogens in the soil. These factors will cause pathogenic microorganisms to reproduce rapidly, thereby causing diseases. Secondly, the most common pathogenic microorganisms are Pythium and Rhizoctonia solani. Once they invade the roots of peach trees, they will cause rot of the root tissue, causing great harm to the trees. In addition, the uneven distribution of branches and leaves cannot promote ventilation and light, further providing conditions for the growth of pathogenic microorganisms.

[0004] To deal with peach rot, a common cultivation measure is the one-sided cultivation mode. The benefits of the one-sided cultivation mode are: 1. It weakens the upward growth potential of the peach tree and turns the growth potential into fruiting potential. The tree grows by the ditch with half of its roots, which is equivalent to limiting a part of the root area and helps to ease the tree's vigor; 2. It allows every branch and leaf of the peach tree to receive light evenly; 3. Reasonable dense planting saves space.

[0005] Although the one-sided cultivation mode can better alleviate the rot caused by tree vigor, the dynamic environment (high density of branches and leaves, many weeds) and static environment (soil agitation) of peach tree growth are the main causes of peach trunk rot. Based on this, the present invention proposes a peach trunk rot prediction method based on dynamic and static environment monitoring. By applying this method combining dynamic and static environment monitoring to the peach test group under the one-sided cultivation mode, the dynamic and static environment of the peach trees under the one-sided cultivation mode can be predicted and analyzed by artificial intelligence, and an early warning can be provided in the early stage of the onset of peach trunk rot, further ensuring the scientific growth of peach trees. At the same time, it also provides a basis for field management of peach trees without relying on the experience of fruit farmers. Summary of the invention

[0006] The purpose of the present invention is to provide a peach tree trunk rot prediction method based on dynamic and static environment monitoring to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: a peach tree trunk rot prediction method based on dynamic and static environment monitoring, comprising the steps of:

[0008] S1: Peach tree cultivation, the peach trees in the experimental group were cultivated based on the cross-ditch one-sided cultivation mode;

[0009] S2: Construction of a dynamic and static environmental monitoring system. An edge computer for data preprocessing and forwarding is constructed along the open area of ​​the experimental group. A dynamic environmental monitoring module for collecting dynamic environmental data and a static environmental monitoring module for collecting static environmental data are arranged along both sides of the ditch. The edge computer can communicate with the cloud server. The dynamic environmental data includes image data of peach tree branches and leaves and image data of weeds under the tree. The static environmental data is soil moisture data. The dynamic environmental monitoring module and the static environmental monitoring module are connected to the edge computer through wired / wireless connection, and the dynamic and static environmental collection data are sent to the edge computer.

[0010] S3: Data preprocessing: The edge computer obtains dynamic and static environmental data and soil moisture data and preprocesses the above data respectively. The preprocessing includes data cleaning, denoising and formatting to delete irrelevant data or duplicate data in the dynamic and static environmental data and soil moisture data, and format them uniformly and upload them to the cloud server.

[0011] S4: Algorithm processing: for the image data of peach tree branches and leaves, the cloud server uses the branch and leaf distribution algorithm to process and obtain the branch and leaf analysis results; for the image data of weeds under the tree, the cloud server uses the weed distribution algorithm to process and obtain the weed analysis results; for the soil moisture data, the cloud server uses the historical trajectory data comparison method based on the healthy soil of the peach tree to process and obtain the soil moisture analysis results;

[0012] S5: Prediction model construction, based on the weighted analysis results of branches and leaves, weeds and soil moisture analysis, a dynamic and static prediction model for peach trunk rot is constructed, and the prediction model is used to predict the rot.

[0013] Preferably, the one-sided cultivation mode across the ditch in S1 refers to a cultivation mode based on tree shape, which requires a single main branch, no main trunk, no side branches, and direct fruiting on the main branch. All main branches in the entire test group tend to one side, the main branches grow straight and obliquely, the tree body is fan-shaped like a fishbone, and the peach tree grows on the edge of the ditch with half of the roots on one side.

[0014] Preferably, the dynamic environment monitoring module includes a multi-axis servo and an image acquisition device mounted on the multi-axis servo, and the static environment monitoring module is an integrated soil moisture module, and it includes at least a soil temperature and humidity sensor, a soil pH sensor, a soil conductivity sensor, and a soil nutrient sensor.

[0015] Preferably, the specific process of the branch-leaf distribution algorithm in S4 includes:

[0016] S401a: Normalize the image size to obtain image intensity information;

[0017] S402a: Extract basic visual features of the image using a feature extraction base network based on VGGNet;

[0018] S403a: performing segmentation of branches, leaves and trunk of the peach tree based on MaskRCNN to obtain contour information of branches, leaves and trunk of the peach tree;

[0019] S404a: constructing a preliminary segmentation network for semantically segmenting different regions in the image, and obtaining a semantic segmentation result;

[0020] S405a: constructing a RNN network, and then taking the peach tree branch and trunk segmentation results, semantic segmentation results and image intensity information as RNN inputs, and establishing spatial associations on the image in a composition manner;

[0021] S406a: Connect the initial segmentation network and the RNN network to build an end-to-end segmentation network;

[0022] S407a: After the segmentation network is constructed, a unified training is performed to obtain a trained segmentation network;

[0023] S408a: Segment the branches, leaves and trunk of the peach tree using the segmentation network.

[0024] S409a: Based on the segmented branches and trunk regions of the peach tree, a density algorithm is used to determine the density of the branches and leaves of the peach tree.

[0025] Preferably, in the MaskRCNN of S403a, by introducing the Inception module into ResNet50, it is added after the ReLU layer of the three parts of res2_, res3_, and res4_, respectively, and three branches are formed by stacking convolutions of 3×3 and 1×1 sizes. The first branch is composed of a 1×1 convolution layer, the second branch is composed of a 1×1 convolution layer followed by a 3×3 convolution layer, and the third branch is composed of a 1×1 convolution layer followed by two 3×3 convolution layers, and finally the outputs of the three branches are fused.

[0026] Preferably, the density algorithm in S409a refers to determining the density by calculating the area of ​​the branches and leaves of the peach tree.

[0027] Preferably, the specific process of the weed distribution algorithm in S4 includes:

[0028] S401b: align, repair, encode and enhance the image of weeds under trees;

[0029] S403b: Construct a parallel convolutional multi-scale feature processing layer;

[0030] S403b: construct a multimodal feature fusion layer, specifically, concatenate the multi-scale color texture features and the multi-scale spatial geometry features through the multimodal feature fusion layer to obtain a multimodal feature fusion result, and the calculation formula is:

[0031] Fea fusion =concat(Fea1,Fea2)

[0032] In the formula, Fea fusion It is the result of multi-modal feature fusion, Fea1 is the multi-scale color texture feature, and Fea2 is the multi-scale spatial geometry feature;

[0033] S404b: construct a parallel convolutional feature extractor, perform parallel convolutional feature extractor construction, and obtain a parallel convolutional feature extractor;

[0034] S405b: extracting features from the farmland RGB image and the three-channel depth image through a parallel convolution feature extractor to obtain multimodal features of farmland weeds;

[0035] S406b: Outputting weed identification analysis results according to the multimodal features.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] The present invention applies the method of combining dynamic and static environment monitoring to a peach tree test group under a one-sided cultivation mode, can predict and analyze the dynamic and static environment of the peach trees under the one-sided cultivation mode by means of artificial intelligence, can provide an early warning in the early stage of peach rot, further ensures the scientific growth of peach trees, and also provides a basis for field management of peach trees without relying on the experience of fruit farmers. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a cultivation mode of the present invention and a prediction flow chart based on dynamic and static environment monitoring. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] See also Figure 1 The present invention provides a technical solution: a peach tree trunk rot prediction method based on dynamic and static environment monitoring, comprising the steps of:

[0041] S1: Peach tree cultivation, the peach trees in the experimental group were cultivated based on the cross-ditch one-sided cultivation mode;

[0042] S2: Construction of a dynamic and static environmental monitoring system. An edge computer for data preprocessing and forwarding is constructed along the open area of ​​the experimental group. A dynamic environmental monitoring module for collecting dynamic environmental data and a static environmental monitoring module for collecting static environmental data are arranged along both sides of the ditch. The edge computer can communicate with the cloud server. The dynamic environmental data includes image data of peach tree branches and leaves and image data of weeds under the tree. The static environmental data is soil moisture data. The dynamic environmental monitoring module and the static environmental monitoring module are connected to the edge computer through wired / wireless connection, and the dynamic and static environmental collection data are sent to the edge computer.

[0043] S3: Data preprocessing: The edge computer obtains dynamic and static environmental data and soil moisture data and preprocesses the above data respectively. The preprocessing includes data cleaning, denoising and formatting to delete irrelevant data or duplicate data in the dynamic and static environmental data and soil moisture data, and format them uniformly and upload them to the cloud server.

[0044] S4: Algorithm processing: for the image data of peach tree branches and leaves, the cloud server uses the branch and leaf distribution algorithm to process and obtain the branch and leaf analysis results; for the image data of weeds under the tree, the cloud server uses the weed distribution algorithm to process and obtain the weed analysis results; for the soil moisture data, the cloud server uses the historical trajectory data comparison method based on the healthy soil of the peach tree to process and obtain the soil moisture analysis results;

[0045] S5: Prediction model construction, based on the weighted analysis results of branches and leaves, weeds and soil moisture analysis, a dynamic and static prediction model for peach trunk rot is constructed, and the prediction model is used to predict the rot.

[0046] In this embodiment, the one-sided cultivation mode across the ditch in S1 refers to a cultivation mode based on tree shape, which requires a single main branch, no main trunk, no side branches, and direct fruiting on the main branch. All main branches in the entire experimental group tend to one side, the main branches grow straight and obliquely, and the tree body is fan-shaped like a fishbone. The peach tree grows on the edge of the ditch with half of the roots on one side.

[0047] In this embodiment, the dynamic environment monitoring module includes a multi-axis servo and an image acquisition device mounted on the multi-axis servo, and the static environment monitoring module is an integrated soil moisture module, and it includes at least a soil temperature and humidity sensor, a soil pH sensor, a soil conductivity sensor, and a soil nutrient sensor.

[0048] In this embodiment, since the disease resistance of the peach tree itself will also affect the occurrence and development of rot, peach trees with weaker disease resistance are more easily attacked by pathogenic microorganisms. Therefore, the static environmental monitoring module mainly uses the above-mentioned soil sensors to perform soil detection, including temperature and humidity, pH, conductivity, nutrients, etc., and can provide a basis for peach trunk rot based on static environmental factors.

[0049] In this embodiment, the main purpose of the branch and leaf distribution algorithm is to identify the luxuriance (density) of the branches and leaves of peach trees under the cross-ditch one-sided cultivation mode. When the branch and leaf luxuriance (density) is detected to be high and it is predicted that it may affect the occurrence of rot, fruit farmers can adjust the distribution of branches and leaves to promote ventilation and light, thereby reducing the breeding conditions of pathogenic microorganisms.

[0050] In this embodiment, the specific process of the branch-leaf distribution algorithm in S4 includes:

[0051] S401a: Normalize the image size to obtain image intensity information;

[0052] S402a: Extract basic visual features of the image using a feature extraction base network based on VGGNet;

[0053] S403a: performing segmentation of branches, leaves and trunk of the peach tree based on MaskRCNN to obtain contour information of branches, leaves and trunk of the peach tree;

[0054] S404a: constructing a preliminary segmentation network for semantically segmenting different regions in the image, and obtaining a result of semantic segmentation;

[0055] S405a: constructing a RNN network, and then taking the peach tree branch and trunk segmentation results, semantic segmentation results and image intensity information as RNN inputs, and establishing spatial associations on the image in a composition manner;

[0056] S406a: Connect the initial segmentation network and the RNN network to build an end-to-end segmentation network;

[0057] S407a: After the segmentation network is constructed, a unified training is performed to obtain a trained segmentation network;

[0058] S408a: Segment the branches, leaves and trunk of the peach tree using the segmentation network.

[0059] S409a: Based on the segmented branches and trunk regions of the peach tree, a density algorithm is used to determine the density of the branches and leaves of the peach tree.

[0060] In this embodiment, in the MaskRCNN of S403a, the Inception module is introduced into ResNet50 and added after the ReLU layers of res2_, res3_, and res4_ respectively. Three branches are formed by stacking convolutions of sizes 3×3 and 1×1. The first branch is composed of a 1×1 convolution layer, the second branch is composed of a 1×1 convolution layer followed by a 3×3 convolution layer, and the third branch is composed of a 1×1 convolution layer followed by two 3×3 convolution layers. Finally, the outputs of the three branches are fused.

[0061] In this embodiment, the density algorithm in S409a refers to determining the density by calculating the area of ​​the branches and leaves of the peach tree.

[0062] In this embodiment, after soil treatment, some pathogens will overwinter on the weeds under the trees. Therefore, the weed distribution algorithm provided in this embodiment is mainly used to identify the regional weeds under the peach trees by image recognition. Since the orchard area is large, some weeds are hidden in the corners. Therefore, the specific distribution position of the weeds is detected by this weed distribution algorithm, and the weighted scoring model is used to determine whether it will affect the treatment of peach rot.

[0063] In this embodiment, the specific process of the weed distribution algorithm in S4 includes:

[0064] S401b: align, repair, encode and enhance the image of weeds under trees;

[0065] S403b: Construct a parallel convolutional multi-scale feature processing layer;

[0066] S403b: construct a multimodal feature fusion layer, specifically, concatenate the multi-scale color texture features and the multi-scale spatial geometry features through the multimodal feature fusion layer to obtain a multimodal feature fusion result, and the calculation formula is:

[0067] Fea fusion =concat(Fea1,Fea2)

[0068] In the formula, Fea fusion It is the result of multi-modal feature fusion, Fea1 is the multi-scale color texture feature, and Fea2 is the multi-scale spatial geometry feature;

[0069] S404b: construct a parallel convolutional feature extractor, perform parallel convolutional feature extractor construction, and obtain a parallel convolutional feature extractor;

[0070] S405b: extracting features from the farmland RGB image and the three-channel depth image through a parallel convolution feature extractor to obtain multimodal features of farmland weeds;

[0071] S406b: Outputting weed identification analysis results according to the multimodal features.

[0072] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0073] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0074] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0075] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0076] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0077] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0078] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks or optical disks.

[0079] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0080] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0081] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A peach tree trunk rot prediction method based on dynamic and static environmental monitoring, characterized in that: Includes steps: S1: Peach tree cultivation, the peach trees in the experimental group were cultivated based on the cross-ditch one-sided cultivation mode; S2: Construction of a dynamic and static environmental monitoring system. An edge computer for data preprocessing and forwarding is constructed along the open area of ​​the experimental group. A dynamic environmental monitoring module for collecting dynamic environmental data and a static environmental monitoring module for collecting static environmental data are arranged along both sides of the ditch. The edge computer can communicate with the cloud server. The dynamic environmental data includes image data of peach tree branches and leaves and image data of weeds under the tree. The static environmental data is soil moisture data. The dynamic environmental monitoring module and the static environmental monitoring module are connected to the edge computer through wired / wireless connection, and the dynamic and static environmental collection data are sent to the edge computer. S3: Data preprocessing: The edge computer obtains dynamic and static environmental data and soil moisture data and preprocesses the above data respectively. The preprocessing includes data cleaning, denoising and formatting to delete irrelevant data or duplicate data in the dynamic and static environmental data and soil moisture data, and format them uniformly and upload them to the cloud server. S4: Algorithm processing: for the image data of peach tree branches and leaves, the cloud server uses the branch and leaf distribution algorithm to process and obtain the branch and leaf analysis results; for the image data of weeds under the tree, the cloud server uses the weed distribution algorithm to process and obtain the weed analysis results; for the soil moisture data, the cloud server uses the historical trajectory data comparison method based on the healthy soil of the peach tree to process and obtain the soil moisture analysis results; S5: Prediction model construction, based on the weighted analysis results of branches and leaves, weeds and soil moisture analysis, a dynamic and static prediction model for peach trunk rot is constructed, and the prediction model is used to predict the rot.

2. A peach tree trunk rot prediction method based on dynamic and static environmental monitoring according to claim 1, characterized in that: The one-sided cultivation mode across the ditch in S1 refers to a cultivation mode based on tree shape, which requires that the tree shape has a single main branch, no main trunk, no side branches, and fruits are directly produced on the main branch. All main branches in the entire test group tend to one side, the main branches grow straight and obliquely, and the tree body is fan-shaped like a fishbone. The peach tree grows on the edge of the ditch with half of the roots on one side.

3. A peach tree trunk rot prediction method based on dynamic and static environmental monitoring according to claim 1, characterized in that: The dynamic environment monitoring module includes a multi-axis servo and an image collector mounted on the multi-axis servo. The static environment monitoring module is an integrated soil moisture module, and it includes at least a soil temperature and humidity sensor, a soil pH sensor, a soil conductivity sensor, and a soil nutrient sensor.

4. A peach tree trunk rot prediction method based on dynamic and static environmental monitoring according to claim 1, characterized in that: The specific process of the branch-leaf distribution algorithm in S4 includes: S401a: Normalize the image size to obtain image intensity information; S402a: Extract basic visual features of the image using a feature extraction base network based on VGGNet; S403a: performing segmentation of branches, leaves and trunk of the peach tree based on MaskRCNN to obtain contour information of branches, leaves and trunk of the peach tree; S404a: constructing a preliminary segmentation network for semantically segmenting different regions in the image, and obtaining a result of semantic segmentation; S405a: constructing a RNN network, and then taking the peach tree branch and trunk segmentation results, semantic segmentation results and image intensity information as RNN inputs, and establishing spatial associations on the image in a composition manner; S406a: Connect the initial segmentation network and the RNN network to build an end-to-end segmentation network; S407a: After the segmentation network is constructed, a unified training is performed to obtain a trained segmentation network; S408a: Segment the branches, leaves and trunk of the peach tree using the segmentation network. S409a: Based on the segmented branches and trunk regions of the peach tree, a density algorithm is used to determine the density of the branches and leaves of the peach tree.

5. A peach tree trunk rot prediction method based on dynamic and static environment monitoring according to claim 4, characterized in that: In the MaskRCNN of S403a, the Inception module is introduced into ResNet50 and added after the ReLU layers of res2_, res3_, and res4_ respectively. Three branches are formed by stacking convolutions of 3×3 and 1×1 sizes. The first branch is composed of a 1×1 convolution layer, the second branch is composed of a 1×1 convolution layer followed by a 3×3 convolution layer, and the third branch is composed of a 1×1 convolution layer followed by two 3×3 convolution layers. Finally, the outputs of the three branches are fused.

6. A peach tree trunk rot prediction method based on dynamic and static environment monitoring according to claim 4, characterized in that: The density algorithm in S409a refers to determining the density by calculating the area of ​​the branches and leaves of the peach tree.

7. The peach trunk rot prediction method based on dynamic and static environmental monitoring according to claim 1, characterized in that: The specific process of the weed distribution algorithm in S4 includes: S401b: align, repair, encode and enhance the image of weeds under the tree; S403b: Construct a parallel convolutional multi-scale feature processing layer; S403b: construct a multimodal feature fusion layer, specifically, concatenate the multi-scale color texture features and the multi-scale spatial geometry features through the multimodal feature fusion layer to obtain a multimodal feature fusion result, and the calculation formula is: Fea fusion =comcat(Fea1,Fea2) In the formula, Fea fusion It is the result of multi-modal feature fusion, Fea1 is the multi-scale color texture feature, and Fea2 is the multi-scale spatial geometry feature; S404b: construct a parallel convolutional feature extractor, perform parallel convolutional feature extractor construction, and obtain a parallel convolutional feature extractor; S405b: extracting features from the farmland RGB image and the three-channel depth image through a parallel convolution feature extractor to obtain multimodal features of farmland weeds; S406b: Outputting weed identification analysis results according to the multimodal features.

Images