Intelligent rattan palm identification and classification method and system

Abstract

The application provides a kind of intelligent palm cane identification classification method and system, it is related to data processing technical field, determine palm cane species set, according to palm cane species set extraction corresponding palm cane image data, output is sample image data set, carry out sample image data set label annotation, establish base recognizer library, based on palm cane species set in base recognizer library Random extraction of base recognizer, obtain multiple recognizer clusters, respectively supervise training of multiple the recognizer cluster, and multiple recognizer clusters are fused by ensemble learning method, obtain identification classification model, identification classification of target scene palm cane is carried out through identification classification model. The technical problem that the palm cane identification and classification method in the prior art relies on artificial experience and professional knowledge, resulting in low efficiency and low accuracy of identification. The technical effect of intelligent identification and classification of palm cane is achieved, and the efficiency and accuracy of palm cane species identification are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to an intelligent rattan identification and classification method and system. BACKGROUND

[0002] Rattan occupies an important position in forestry, horticulture, medicinal materials and other fields due to its unique growth habit and wide application value. Rattan is not only an important economic crop, but also has important uses in home furnishing, handicrafts and medicinal materials. With the emphasis on rattan industry, how to effectively identify and classify different types of rattan has become an important research topic. However, rattan has many types and complex growth environment, often with great variability, which brings great challenges to identification and classification. Traditional rattan identification and classification methods mainly rely on manual experience and professional knowledge, which has the defects of low efficiency and difficulty in ensuring the accuracy and consistency of identification.

[0003] The rattan identification and classification method in the prior art relies on manual experience and professional knowledge, resulting in the technical problems of low efficiency and low accuracy of identification. SUMMARY

[0004] The present application provides an intelligent rattan identification and classification method and system to solve the technical problems of low efficiency and low accuracy of identification caused by the rattan identification and classification method in the prior art relying on manual experience and professional knowledge.

[0005] In view of the above problems, the present application provides an intelligent rattan identification and classification method and system.

[0006] In a first aspect of the present application, an intelligent rattan identification and classification method is provided, which comprises: determining a rattan species set by interacting with a rattan knowledge base, and extracting image data of corresponding rattan according to the rattan species set, and outputting as a sample image data set; performing label annotation of the sample image data set to obtain an annotated image data set; establishing a base recognizer library, and performing random extraction of base recognizers in the base recognizer library based on the rattan species set to obtain multiple recognizer clusters; taking the annotated image data set as a sample space, respectively performing supervised training of multiple recognizer clusters, and fusing multiple recognizer clusters through an ensemble learning method to obtain an identification and classification model; and performing identification and classification of rattan in a target scene through the identification and classification model.

[0007] In a second aspect of the present application, an intelligent rattan identification and classification system is provided, which comprises: an image data extraction module configured to determine a rattan species set by interacting with a rattan knowledge base, and extract image data of corresponding rattan according to the rattan species set, and output as a sample image data set; an image data labeling module configured to perform label labeling on the sample image data set, and obtain a labeled image data set; a base recognizer extraction module configured to establish a base recognizer library, and perform random extraction of base recognizers in the base recognizer library based on the rattan species set, and obtain a plurality of recognizer clusters; a classification model obtaining module configured to take the labeled image data set as a sample space, perform supervised training of a plurality of the recognizer clusters respectively, and obtain an identification and classification model by integrating a plurality of the recognizer clusters through an ensemble learning method; and a data classification module configured to perform identification and classification of rattan in a target scene through the identification and classification model.

[0008] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0009] The method provided by the embodiments of the present application determines a rattan species set by interacting with a rattan knowledge base, and extracts image data of corresponding rattan according to the rattan species set, and outputs as a sample image data set; performs label labeling on the sample image data set, and obtains a labeled image data set; establishes a base recognizer library, and performs random extraction of base recognizers in the base recognizer library based on the rattan species set, and obtains a plurality of recognizer clusters; takes the labeled image data set as a sample space, performs supervised training of a plurality of the recognizer clusters respectively, and obtains an identification and classification model by integrating a plurality of the recognizer clusters through an ensemble learning method; and performs identification and classification of rattan in a target scene through the identification and classification model. The technical effect of intelligent identification and classification of rattan is achieved, and the rattan species identification efficiency and accuracy are improved. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0011] Figure 1 A flowchart of an intelligent rattan identification and classification method provided by the present application is shown in the figure.

[0012] Figure 2 A structure diagram of an intelligent rattan identification and classification system provided by the present application is shown in the figure.

[0013] Reference signs: image data extraction module 11, image data labeling module 12, base identifier extraction module 13, classification model obtaining module 14, data classification module 15. DETAILED DESCRIPTION

[0014] The application provides an intelligent palm cane identification and classification method and system, which is used to solve the technical problem that the palm cane identification and classification method in the prior art relies on artificial experience and professional knowledge, resulting in low efficiency and low accuracy of identification. The technical effect of intelligent identification and classification of palm cane is achieved, and the efficiency and accuracy of palm cane species identification are improved.

[0015] Hereinafter, the technical solutions in the application will be described clearly and completely with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the application, rather than all the embodiments of the application. It should be understood that the application is not limited by the example embodiments described herein. Based on the embodiments of the application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the application. In addition, it should be noted that, for the convenience of description, only the parts related to the application are shown in the drawings, not all.

[0016] In one embodiment, as shown in the accompanying drawings, the application provides an intelligent palm cane identification and classification method, which comprises: Figure 1

[0017] The interactive palm cane knowledge base determines a palm cane species set, and extracts image data of the corresponding palm cane according to the palm cane species set, and outputs a sample image data set.

[0018] ​Specifically, palm rattan is a kind of liana belonging to the palm family, with a wide variety of species. Due to its unique growth habits and extensive application value, it plays an important role in forestry, horticulture, medicinal materials and other fields. Through the interactive interface, the palm rattan knowledge base is queried to determine the set of palm rattan species with specific requirements or conditions. The palm rattan knowledge base is a dynamic database that contains rich palm rattan-related information such as species name, morphological characteristics, ecological habits, geographical distribution, etc. The palm rattan-related information is verified and updated by experts to ensure its accuracy and reliability. The palm rattan species set includes a collection of different palm rattan species, varieties or cultivation types, such as industrial palm rattan, ornamental palm rattan, etc. According to the determined palm rattan species set, image data covering multiple conditions is extracted from the palm rattan knowledge base, including different growth stages such as seedling stage and mature stage, different shooting angles such as top and side, different lighting conditions such as strong light and weak light, and different growth environments such as humid and dry. The extracted image data is integrated and output as a sample image data set, which contains high-quality and comprehensive palm rattan images, providing a solid foundation and guarantee for intelligent palm rattan recognition and classification.

[0019] Further, the interactive palm rattan knowledge base determines the palm rattan species set, and extracts the corresponding palm rattan image data according to the palm rattan species set, and outputs a sample image data set, including: according to the scene environment information of the target scene, interacting with the palm rattan knowledge base, obtaining the palm rattan species set; traversing the palm rattan species set to determine the stage characteristics of each species of palm rattan, and determining the corresponding stage sampling node set; taking the stage sampling node set and the palm rattan species as extraction constraints, extracting multiple palm rattan images based on big data, and storing them classifiedly, generating the sample image data set.

[0020] Specifically, scene environment information of a target scene is obtained, the target scene refers to a place or environment that actually needs to be identified, and the scene environment information refers to various environmental variables related to the target scene, including geographical location, climate conditions, and industrial demand, such as temperature, humidity, precipitation, and market demand for palm rattan and cultivation mode. Through interaction with the palm rattan knowledge base, combined with the scene environment information of the target scene, the palm rattan species set that meets the scene is filtered from the palm rattan knowledge base, ensuring that the palm rattan data set can fully represent the diversity of palm rattan. Each palm rattan in the palm rattan species set is analyzed in detail, and the growth stage characteristics of each palm rattan are determined by traversing the species set. For example, palm rattan may include different stages such as seedling stage, growth stage, and mature stage, each stage has different morphological characteristics and growth conditions. The stage characteristics refer to the characteristics exhibited by the palm rattan at each stage of its life cycle, such as leaf morphology, vine length, color change, etc. According to the stage characteristics of each palm rattan, the corresponding stage sampling node set is determined. The stage sampling node set refers to selecting appropriate time points for data sampling at different growth stages, and each node represents a specific sampling time point, ensuring that effective and reliable palm rattan image data can be obtained at different growth stages. After determining the palm rattan species set and the stage sampling node set, the stage sampling node set and the palm rattan species are used as constraint conditions, and based on big data technology, multiple palm rattan-related images are extracted from the palm rattan knowledge base. The extracted image data includes different growth stages, different angles, different light conditions, and other situations of palm rattan, ensuring the comprehensiveness of the image data. The multiple palm rattan images extracted are classified according to the species and growth stage of the palm rattan, and are effectively stored to generate a sample image data set, providing comprehensive and reliable data support for intelligent palm rattan identification and classification.

[0021] Label annotation of the sample image data set is performed to obtain a labeled image data set.

[0022] Further, label annotation of the sample image data set is performed to obtain a labeled image data set, wherein the labels of the labeled image data set at least include palm rattan species, growth period state, maturity, and disease condition.

[0023] Specifically, after obtaining the sample image dataset, according to image recognition technology, bioinformatics and other methods, specific labels are added to each sample image in the sample image dataset using image labeling tools such as LabelImg, CVAT, etc. The labeled labels at least include palm cane species, growth period state, maturity and disease condition. Palm cane species refers to the palm cane species to which the labeled image belongs. The growth period state labels the growth state of the palm cane at the image time, such as seedling stage, growth stage, mature stage, etc. Maturity refers to whether the palm cane has reached full maturity or is in a certain immature state, for example: green fruit stage or mature stage. Disease condition refers to the type and degree of disease if the palm cane in the image has disease, for example, yellow spot disease, leaf spot disease, etc. Optionally, the specific labeling process is as follows: through image recognition technology, combined with the appearance characteristics of palm cane, the species of palm cane is accurately judged, and palm cane species labeling is performed; the size, color and other characteristics of the leaves and vines of the palm cane in the image are analyzed to determine which growth stage it is in, and the growth period state labeling is performed; according to the size, color and thickness of the palm cane fruits in the image, the maturity is labeled; through image analysis, it is identified whether the palm cane leaves have disease spots or signs of pests and diseases, and qualitative and quantitative analysis of the disease is performed, and the specific disease type and impact degree are labeled. For example, label the leaf yellow spot disease, slight disease. Labeling the sample image dataset provides accurate and comprehensive data support for intelligent palm cane recognition, thereby improving the efficiency and accuracy of palm cane recognition.

[0024] A base recognizer library is established, and random extraction of base recognizers is performed based on the palm cane species set in the base recognizer library to obtain a plurality of recognizer clusters.

[0025] Further, a base recognizer library is established, and random extraction of base recognizers is performed based on the palm cane species set in the base recognizer library to obtain a plurality of recognizer clusters, including: defining a set of available model structures, and based on the set of available model structures, establishing a plurality of base recognizers to obtain the base recognizer library; performing severity evaluation on the multi-dimensional feature information corresponding to the palm cane species set, and defining an adjustment coefficient set based on the severity evaluation result; based on the preset cluster size and the adjustment coefficient set, repeatedly randomly extracting from the base recognizer library, and outputting the extracted content as a plurality of recognizer clusters.

[0026] Specifically, first, a set of available model structures is determined, which are verified model architectures containing various machine learning models suitable for rattan identification, such as convolutional neural networks, deep neural networks, and transfer learning models, representing different network structures and parameter configurations, capable of capturing different features in rattan images. According to the set of available model structures, multiple base identifiers are established, each of which is initialized with a model structure in the set and trained through a labeled image dataset for performing a specific rattan species or feature classification task. The multiple base identifiers are integrated to form a library containing multiple base identifiers. Multi-dimensional feature information corresponding to the set of rattan species is extracted from rattan images through a feature extraction algorithm, including intrinsic features, value features, and typical planting proportions, etc. Intrinsic features such as leaf morphology and vine length, value features such as economic value and market demand, and typical planting proportions such as the distribution of different rattan species in the target scene. The extracted multi-dimensional feature information is cleaned, normalized or standardized to ensure data consistency, and then the correlation coefficient or mutual information between each feature and rattan species is calculated for multi-dimensional feature information. The severity evaluation refers to a weighted calculation based on multi-dimensional features to determine the importance of different rattan species. According to the analysis results, the severity evaluation results are obtained. Further, based on the severity evaluation results, an adjustment coefficient set is defined, which is a parameter determined according to the evaluation results for adjusting the randomness and weight of model selection, representing the range of up and down fluctuations. A preset cluster size is determined, which refers to the number of base identifiers included in each identifier cluster. An identifier cluster is a collection of multiple base identifiers, each cluster has unique classification ability and combination characteristics. For example, an identifier cluster contains M base identifiers, M is a positive integer. According to the cluster size and adjustment coefficient set, random extraction is performed in the base identifier library, which adopts a replaceable strategy, i.e. after extraction, the base identifier can still participate in the next extraction to ensure diversity. Through multiple extractions, multiple identifier clusters are formed, each of which is used for subsequent classification model training and ensemble learning. By constructing the base identifier library, rich model diversity is provided, which helps to improve the performance of rattan intelligent identification. Through severity evaluation and adjustment coefficient set, the importance of features is considered in the random extraction process, making the extracted identifier clusters more consistent with actual needs, further improving the accuracy and stability of rattan identification.

[0027] Further, the multi-dimensional feature information corresponding to the rattan species set is subjected to severity evaluation, and an adjustment coefficient set is defined based on the severity evaluation result, including: interacting with the rattan knowledge base to match the intrinsic feature information corresponding to the rattan species set; interacting with the rattan transaction platform to obtain the value feature information corresponding to the rattan species set; interacting with the target scene to obtain the typical planting proportion information corresponding to the rattan species set; performing weighted evaluation based on the intrinsic feature information, the value feature information and the typical planting proportion information to obtain the severity evaluation result; calculating the average value of the severity evaluation result as a severity baseline, and calculating the ratio of the severity evaluation result to the severity baseline, and outputting as the adjustment coefficient set.

[0028] Specifically, the intrinsic feature information of the rattan species set is obtained by interacting with the rattan knowledge base. The intrinsic feature information refers to the natural attributes of rattan, which is used to reflect the identification difficulty, such as cane length, leaf shape and size, form complexity, texture significance, etc. These features directly reflect the biological characteristics of rattan. The value feature information corresponding to the rattan species set is obtained by interacting with the rattan transaction platform, i.e. by interacting with the online or offline rattan transaction market. The value feature refers to information reflecting market demand and economic benefits, for example, a certain rattan has high price due to its strong durability. The typical planting proportion information of rattan species in the target scene is extracted by analyzing the actual target area or environment. The typical planting proportion information refers to the planting area ratio or quantity distribution of each type of rattan in a specific area, reflecting the actual demand of the target scene. According to the task requirements and actual importance, each feature is weighted, and a weighted formula is used to comprehensively evaluate each rattan species. Weighted evaluation refers to assigning weights according to the importance of features, calculating the comprehensive score, and obtaining the severity evaluation result. Severity evaluation result = (intrinsic feature value x w1) + (value feature value x w2) + (planting proportion feature value x w3), wherein w1, w2 and w3 are weights. The average value of the severity evaluation results of all species is calculated, and the average value calculation result is taken as the severity baseline. The severity baseline represents the average level of importance of all species and is a reference value for adjustment coefficient calculation. The ratio of the severity evaluation score of each rattan species to the severity baseline is calculated to obtain the adjustment coefficient set, wherein the adjustment coefficient = severity evaluation score / severity baseline. The adjustment coefficient is a floating ratio value according to the importance of the species relative to the average level, which is used to dynamically adjust the recognizer selection probability or quantity. By integrating intrinsic features, value features and planting proportion features, a comprehensive severity evaluation is formed, which can dynamically adapt the recognizer configuration of different species according to actual requirements. Through the dynamic adjustment mechanism of the adjustment coefficient, the rattan recognition flexibility and adaptability are enhanced, thereby improving the rattan classification efficiency and accuracy.

[0029] The labeled image data set is used as a sample space, and supervised training of multiple identifier clusters is performed separately. The multiple identifier clusters are fused through an ensemble learning method to obtain a recognition classification model.

[0030] The palm vines in the target scene are identified and classified using the identification and classification model.

[0031] Specifically, the labeled image dataset is used as the sample space. The sample space refers to the training data used to train the model. The labeled image dataset has been accurately labeled and contains label information such as the type, growth period, maturity, and disease condition of the palm vine. Using the labeled image dataset, supervised training is performed for each recognizer cluster. Supervised training refers to training the model on the labeled sample data, optimizing the model parameters by inputting image data and corresponding labels, so that the model can recognize the relationship between the image and the label. Each recognizer cluster contains multiple base recognizers, which are responsible for classifying different palm vine species, stages or disease conditions. The training process is to use the sample image dataset as input for each recognizer cluster. The sample image dataset contains input images, such as images of palm vines and corresponding labels. The training goal is to learn the relationship between the input image and the label so that the base recognizer can accurately predict the corresponding label on the new input image. Using traditional supervised learning methods for training, each base recognizer extracts features from the sample space. The cross-entropy loss function is used to measure the difference between the model prediction and the true label. The base recognizer calculates the prediction through forward propagation. Based on the loss function results, the gradient is calculated through backpropagation. The model weights are adjusted using optimization algorithms such as SGD and Adam. When the model accuracy reaches convergence, iterations are terminated, resulting in a trained recognizer cluster. After training, each recognizer cluster learns the optimal weight parameters to accurately classify the input image. Using ensemble learning methods such as voting and weighted averaging, the trained multiple recognizer clusters are fused to generate a final overall recognition and classification model. By combining multiple models, this recognition and classification model reduces the bias and variance of individual models and improves the performance of the overall model. During the fusion process, the weights are dynamically adjusted based on the accuracy of each recognizer cluster, so that the overall model optimizes the recognition performance for each palm rattan species, stage, and disease condition. By integrating multiple recognizer clusters, the bias and overfitting problems that may exist in a single model are reduced, and multiple tasks can be processed simultaneously, thereby improving the overall processing capability and task adaptability, and providing accurate, efficient and stable palm vine classification and recognition functions.

[0032] Further, image data in the target scene is collected through shooting, sensor data or existing databases, and palm vine images in the target scene are taken as input data to the trained recognition classification model. The recognition classification model includes multiple base identifier clusters, and the outputs of multiple base identifiers are fused through ensemble learning to obtain the final classification result. Multiple base identifier clusters in the recognition classification model respectively identify and classify the input image and fuse, and output specific category information of the palm vine in the target scene, including palm vine species, growth stage, maturity and presence of disease. Through ensemble learning, multiple base identifier clusters can complement each other to improve the classification accuracy of palm vine species and growth stage. Through automatic image recognition and classification, the recognition classification model can efficiently and accurately classify and identify palm vines in the target scene, greatly reducing the cost and time of manual inspection.

[0033] Further, the label annotation of the sample image dataset is performed to obtain an annotated image dataset, and then the method further comprises: performing image enhancement on the annotated image dataset to generate a strengthened image dataset; extracting palm vine edge information through edge detection; performing image segmentation based on the edge information and identifying and determining a palm vine pixel region; and performing light processing on the annotated image dataset based on the palm vine pixel region.

[0034] Specifically, the annotated image dataset is subjected to image enhancement through various data enhancement methods to generate a strengthened image dataset. The data enhancement methods include geometric transformation and color transformation. The geometric transformation includes rotation, flipping, cropping and scaling to simulate palm vine images under different shooting angles and distances. The color transformation includes changing brightness, contrast and saturation to simulate different lighting conditions. The image enhancement refers to transforming and processing the image to expand the size and diversity of the training dataset. The strengthened image dataset refers to a new image set generated through enhancement, which is more diverse and representative. The Canny edge detection algorithm, Sobel operator or Laplace operator is used to identify the regions with significant pixel value changes in the image for palm vine image edge detection to extract palm vine edge information. Then, according to the edge detection result, the image is divided into a target region and a background region. The target region is the palm vine pixel region. The image segmentation method can be selected. The region growing method is used to gradually expand the region from the initial seed point according to the similarity of adjacent pixels, or a semantic segmentation model is used to accurately segment the palm vine region. After the palm vine pixel region is determined, the light processing of the annotated image dataset is performed to simplify the image content and only retain the pixel region related to the palm vine, thereby reducing noise interference and improving the model training efficiency. The generated strengthened image dataset and the light-processed annotated image dataset significantly improve the quality of the training data and improve the model recognition accuracy and reliability.

[0035] Further, based on the palm fiber pixel region, the light processing of the labeled image dataset is performed, including: extracting the pixel coordinate features of the palm fiber pixel region; mapping the pixel coordinate features to the labeled image dataset, and removing the pixels not belonging to the mapping result to realize the light processing.

[0036] Specifically, after image segmentation, the segmented palm fiber pixel region is obtained, and then the segmented image is traversed to extract the pixel point coordinates of the palm fiber pixel region, and the pixel coordinate features of the palm fiber pixel region are obtained. The palm fiber pixel region refers to a group of continuous pixels in the palm fiber image, which represents the location of the target object. The pixel coordinate feature refers to the position coordinates of each target pixel in the image, which is represented by (x, y). The extracted palm fiber pixel region coordinate features are mapped to the labeled image dataset, each image in the labeled dataset is checked one by one to ensure that the target region is completely aligned with the pixel coordinates, the target region is reconstructed through the coordinate features, and only the pixels matching the pixel coordinates are retained in the labeled image dataset. After the mapping is completed, the pixels in the image that do not belong to the target region, i.e. the background region or other interference parts, are removed, and the non-target region pixels are set to a specific value, such as zero or completely removed, thereby simplifying the image structure and realizing the light processing of the labeled image dataset. Through the light processing, only the palm fiber related pixel region is retained in each image, so that the model focuses more on the target features. By accurately extracting the target region pixels and removing the background, the interference of irrelevant information on the model training is significantly reduced, so that the model can focus on learning the key features of the target region, and the accuracy of recognition and classification is improved.

[0037] In summary, the intelligent palm fiber recognition and classification method provided by the embodiments has the following technical effects:

[0038] 1. The intelligent palm fiber recognition and classification method can automatically and quickly recognize palm fibers, significantly improving the recognition efficiency and accuracy, and reducing the cost of manual recognition.

[0039] 2. Accurate recognition and classification can help promote the development of palm fiber breeding, cultivation, processing and other industries, and improve the economic value and ecological benefits of palm fiber resources.

[0040] Embodiment two, based on the same inventive concept as the intelligent palm fiber recognition and classification method in the foregoing embodiments, as shown in Figure 2 The present application provides an intelligent palm fiber recognition and classification system, wherein the system comprises:

[0041] The image data extraction module 11 is used to determine a palm vine species set by interacting with the palm vine knowledge base, and extract image data of corresponding palm vines according to the palm vine species set, and output as a sample image data set; the image data labeling module 12 is used to perform label labeling on the sample image data set, and obtain a labeled image data set; the base recognizer extraction module 13 is used to establish a base recognizer library, and perform random extraction of base recognizers in the base recognizer library based on the palm vine species set, and obtain a plurality of recognizer clusters; the classification model obtaining module 14 is used to take the labeled image data set as a sample space, respectively perform supervised training on a plurality of the recognizer clusters, and fuse a plurality of the recognizer clusters through an ensemble learning method, and obtain a recognition classification model; and the data classification module 15 is used to perform recognition and classification of target scene palm vines through the recognition classification model.

[0042] Further, the image data extraction module 11 is further used to perform the following steps: according to scene environment information of a target scene, interacting with the palm vine knowledge base, obtaining the palm vine species set; traversing the palm vine species set to determine the stage characteristics of each species of palm vine, and determining a corresponding stage sampling node set; taking the stage sampling node set and the palm vine species as extraction constraints, extracting a plurality of palm vine images based on big data, and classifying and storing to generate the sample image data set.

[0043] Further, the base recognizer extraction module 13 is further used to perform the following steps: defining an available model structure set, and establishing a plurality of base recognizers based on the available model structure set, and obtaining the base recognizer library; performing severity evaluation on multi-dimensional feature information corresponding to the palm vine species set, and defining an adjustment coefficient set based on the severity evaluation result; based on a preset cluster size and the adjustment coefficient set, performing repeatable random extraction in the base recognizer library, and outputting the extraction content as a plurality of the recognizer clusters.

[0044] Further, the base recognizer extraction module 13 is further used to perform the following steps: interacting with the palm vine knowledge base, matching intrinsic feature information corresponding to the palm vine species set; interacting with a palm vine transaction platform, obtaining value feature information corresponding to the palm vine species set; interacting with a target scene, obtaining typical planting proportion information corresponding to the palm vine species set; performing weighted evaluation based on the intrinsic feature information, the value feature information and the typical planting proportion information, and obtaining the severity evaluation result; calculating an average value of the severity evaluation result as a severity baseline, and calculating a ratio of the severity evaluation result to the severity baseline, and outputting as the adjustment coefficient set.

[0045] Further, the image data labeling module 12 is further used to perform the following steps: the labels of the labeled image data set at least include palm vine species, growth period state, maturity and disease condition.

[0046] Further, the image data labeling module 12 is further configured to perform the following steps: performing image enhancement on the labeled image dataset to generate an enhanced image dataset; extracting palm vine edge information by edge detection; performing image segmentation based on the edge information to identify and determine palm vine pixel regions; and performing light processing on the labeled image dataset based on the palm vine pixel regions.

[0047] Further, the image data labeling module 12 is further configured to perform the following steps: extracting pixel coordinate features of the palm vine pixel regions; mapping the pixel coordinate features to the labeled image dataset; and removing pixels not belonging to the mapping results to achieve light processing.

[0048] The above merely provides the preferred embodiments of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0049] The specification and drawings merely illustrate the exemplary embodiments of the present application, and should be considered as covering any and all modifications, variations, combinations or equivalents that fall within the scope of the present application. Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technologies, the present application is intended to include these modifications and variations.

Claims

1. An intelligent palm vine identification and classification method, characterized in that: The method comprises: The interactive palm rattan knowledge base determines a palm rattan species set, and extracts corresponding palm rattan image data according to the palm rattan species set, and outputs the data as a sample image dataset; Perform label annotation on the sample image dataset to obtain an annotated image dataset; Establishing a base identifier library, and randomly extracting base identifiers based on the palm rattan species set in the base identifier library to obtain a plurality of identifier clusters; Using the labeled image dataset as a sample space, supervised training is performed on a plurality of the identifier clusters, and the plurality of identifier clusters are fused through an ensemble learning method to obtain a recognition classification model; Identify and classify the target scene palm vines using the identification and classification model; A base identifier library is established, and base identifiers are randomly extracted from the base identifier library based on the palm rattan species set to obtain multiple identifier clusters, including: Defining an available model structure set, and establishing a plurality of base identifiers based on the available model structure set, and obtaining the base identifier library; Obtaining multidimensional feature information corresponding to the set of palm rattan species for heavy evaluation, and defining an adjustment coefficient set based on the heavy evaluation result; Based on a preset cluster size and the adjustment coefficient set, repeatable random extraction is performed on the base identifier library, and the extracted content is output as a plurality of the identifier clusters.

2. The intelligent palm vine identification and classification method according to claim 1, characterized in that: The interactive rattan knowledge base determines a set of rattan species, and extracts corresponding rattan image data according to the set of rattan species, and outputs the data as a sample image dataset, including: According to the scene environment information of the target scene, interacting with the palm rattan knowledge base to obtain the palm rattan species set; Traversing the palm rattan species set to determine the stage characteristics of each type of palm rattan, and determining a corresponding stage sampling node set; Taking the stage sampling node set and the palm vine species as extraction constraints, a plurality of palm vine images based on big data are extracted and classified and stored to generate the sample image data set.

3. The intelligent palm vine identification and classification method according to claim 2, characterized in that: Obtain multidimensional feature information corresponding to the set of palm rattan species for heavy evaluation, and define an adjustment coefficient set based on the heavy evaluation result, including: Interacting with the palm rattan knowledge base to match the intrinsic feature information corresponding to the palm rattan species set; Interacting with the palm rattan trading platform to obtain value characteristic information corresponding to the palm rattan species set; Interacting with the target scene, obtaining typical planting ratio information corresponding to the set of palm and rattan species; Performing a weighted evaluation based on the intrinsic feature information, the value feature information, and the typical planting ratio information to obtain the heavy assessment result; An average value of the severe assessment results is calculated as a severe baseline, and a ratio of the severe assessment results to the severe baseline is calculated and output as the adjustment coefficient set.

4. The intelligent palm vine identification and classification method according to claim 1, characterized in that: The sample image dataset is labeled to obtain a labeled image dataset, wherein the labels of the labeled image dataset at least include palm rattan species, growth period, maturity, and disease status.

5. The intelligent palm vine identification and classification method according to claim 1, characterized in that: Perform labeling of the sample image dataset to obtain the labeled image dataset, and then further include: Perform image enhancement on the labeled image dataset to generate an enhanced image dataset; Extract palm vine edge information through edge detection; Perform image segmentation based on edge information and identify and determine the palm vine pixel area; Lightweight processing of the labeled image dataset is performed based on the palm vine pixel area.

6. The intelligent palm vine identification and classification method according to claim 5, characterized in that: Performing lightweight processing on the labeled image dataset based on the palm vine pixel area includes: Extracting pixel coordinate features of the palm vine pixel area; The pixel coordinate features are mapped to the annotated image dataset, and pixels that do not belong to the mapping results are eliminated to achieve lightweight processing.

7. An intelligent palm rattan identification and classification system, characterized in that: The intelligent palm rattan identification and classification system is used to implement the steps of the method according to any one of claims 1 to 6, comprising: An image data extraction module is used to interact with a palm rattan knowledge base to determine a palm rattan species set, and extract corresponding palm rattan image data according to the palm rattan species set, and output the data as a sample image dataset; An image data annotation module, configured to annotate the sample image dataset with labels and obtain an annotated image dataset; a base identifier extraction module, configured to establish a base identifier library and randomly extract base identifiers based on the palm rattan species set in the base identifier library to obtain a plurality of identifier clusters; a classification model acquisition module, configured to perform supervised training of the plurality of the identifier clusters using the labeled image dataset as a sample space, and fuse the plurality of the identifier clusters through an ensemble learning method to obtain a recognition classification model; The data classification module is used to identify and classify palm vines in the target scene through the identification and classification model.

Images