Machine vision-based tender shoot identification method and system

Abstract

The invention discloses a tender shoot identification method and system based on machine vision, and relates to the technical field of image processing and identification, and the method comprises the steps: capturing the image data of to-be-identified tea leaves at different visual angles; performing instance segmentation on the image data by adopting an instance segmentation algorithm to respectively obtain a tea set and a stem set; constructing a graph structure, and obtaining a tree structure through pruning; performing feature calculation on the tea set, adding a dichotomy label, and constructing a multi-dimensional input vector; and inputting the multi-dimensional input vector into a tender shoot identification and classification model to obtain an identification result of the tea tender shoots. Based on machine vision and an advanced algorithm model, a large number of tea images can be rapidly processed and analyzed, automatic identification of the tea tender shoots is achieved, and compared with a traditional manual identification mode, the identification efficiency is greatly improved, the labor cost and the time cost are saved, and the identification efficiency is improved. Powerful support is provided for automatic operation of subsequent links such as tea leaf picking.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing and recognition, and more particularly to a method and system for identifying young sprouts based on machine vision. Background Art

[0002] As a tea variety that is deeply loved by consumers, white tea occupies an important position in the tea market. Its quality and taste depend to a large extent on the picked raw materials. The picking of white tea mainly focuses on the newly sprouted tender buds. These tender buds are rich in nutrients and unique flavor substances, and are the key factors in determining the quality of white tea.

[0003] For a long time, the picking of white tea has mainly relied on manual labor. In the actual picking process, tea farmers need to bend over in the tea garden for a long time, looking for and picking the buds that meet the standards from tree to tree and branch to branch. Currently, a large amount of manual labor is used in the picking process, which is costly, labor-intensive, and in a harsh environment. Automated and precise picking is a major development trend in the future. The premise of automated picking is that the equipment can accurately identify the buds.

[0004] Therefore, how to provide a machine vision-based tender shoot identification method and system to provide picking guidance for automated precision picking equipment is an urgent problem that technicians in this field need to solve. Summary of the invention

[0005] In view of this, the present invention provides a method and system for identifying young tea buds based on machine vision, which can quickly process and analyze a large number of tea images to achieve automatic identification of young tea buds.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a method for identifying young sprouts based on machine vision, comprising:

[0007] Capture image data of tea leaves to be identified at different viewing angles;

[0008] Using an instance segmentation algorithm to perform instance segmentation on the image data to obtain a tea set; at the same time, segmenting the stems of the tea leaves to extract the main trunk to obtain a stem set;

[0009] Constructing a graph structure based on the tea set and the stem set, and then obtaining a tree structure through pruning;

[0010] Based on the tree structure, feature calculation is performed on the tea set, and binary classification labels are added to construct a multi-dimensional input vector;

[0011] A bud identification and classification model is constructed, and the multi-dimensional input vector is input into the bud identification and classification model to obtain the identification result of tea buds.

[0012] Preferably, the binary classification labels include young buds and green leaves of tea leaves.

[0013] Preferably, constructing a graph structure based on the tea set and the stem set includes:

[0014] The centroid A of the tea leaves center is a leaf node, and the two endpoints of the stem set are path B end , and use the path length as the path weight β to match the centroid A center To Path B end The nearest end of , builds the graph structure.

[0015] Preferably, capturing image data of tea leaves to be identified at different viewing angles includes:

[0016] A plurality of rigidly connected industrial cameras are used to make their fields of view intersect, and the tea leaves are located within the intersecting fields of view.

[0017] Preferably, performing feature calculation on the tea set includes:

[0018] Calculate the growth position, size, aspect ratio and HSV channel histogram of the tea leaves.

[0019] Preferably, SVM is used for binary classification.

[0020] Preferably, a bud recognition system based on machine vision comprises:

[0021] A tea image capture module, used to capture image data of tea leaves to be identified at different viewing angles;

[0022] An instance segmentation module is used to perform instance segmentation on the image data using an instance segmentation algorithm to obtain a tea set; and to segment the stems of the tea leaves to extract the main trunk to obtain a stem set;

[0023] A tree structure acquisition module, used to construct a graph structure based on the tea set and the stem set, and then acquire a tree structure by pruning;

[0024] A feature calculation module, used to calculate the features of the tea set based on the tree structure, add binary classification labels, and construct a multi-dimensional input vector;

[0025] The tea bud recognition module is used to construct a bud recognition classification model, input the multi-dimensional input vector into the bud recognition classification model, and obtain the recognition result of the tea bud.

[0026] Through the above technical solutions, it can be known that compared with the prior art, the present invention discloses a method and system for identifying young buds based on machine vision, including: capturing image data of tea leaves to be identified under different viewing angles; using an instance segmentation algorithm to perform instance segmentation on the image data to obtain a tea set; segmenting the stems of the tea leaves at the same time, extracting the trunk, and obtaining a stem set; constructing a graph structure based on the tea set and the stem set, and then obtaining a tree structure through pruning; based on the tree structure, performing feature calculation on the tea set, and adding binary classification labels to construct a multi-dimensional input vector; constructing a young bud identification classification model, and inputting the multi-dimensional input vector into the young bud identification classification model to obtain the identification result of tea young buds. The present invention is based on machine vision and advanced algorithm models, and can quickly process and analyze a large number of tea images to realize the automatic identification of tea young buds. Compared with the traditional manual identification method, it greatly improves the identification efficiency, saves labor costs and time costs, and provides strong support for the automation of subsequent links such as tea picking. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0028] Figure 1 A schematic flow chart of a method for identifying young sprouts based on machine vision provided by an embodiment of the present invention;

[0029] Figure 2 A schematic diagram of capturing tea leaves to be identified at different viewing angles provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] 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.

[0031] The embodiment of the present invention discloses a method and system for identifying young buds based on machine vision, which captures image data of tea leaves to be identified from different viewing angles; uses an instance segmentation algorithm to perform instance segmentation on the image data, and obtains a tea set and a stem set respectively; constructs a graph structure, and then obtains a tree structure through pruning; performs feature calculation on the tea set, and adds binary classification labels to construct a multi-dimensional input vector; inputs the multi-dimensional input vector into a young bud identification classification model to obtain the identification result of young tea leaves. Based on machine vision and advanced algorithm models, the present invention can quickly process and analyze a large number of tea images, and realize automatic identification of young tea leaves. Compared with traditional manual identification methods, it greatly improves the identification efficiency, saves labor costs and time costs, and provides strong support for the automation of subsequent links such as tea picking.

[0032] In a specific embodiment of the present invention, a method for identifying young sprouts based on machine vision is provided. Figure 1 As shown, including:

[0033] Capture image data of tea leaves to be identified at different viewing angles;

[0034] Using an instance segmentation algorithm to perform instance segmentation on the image data to obtain a tea set; at the same time, segmenting the stems of the tea leaves to extract the main trunk to obtain a stem set;

[0035] Constructing a graph structure based on the tea set and the stem set, and then obtaining a tree structure through pruning;

[0036] Based on the tree structure, feature calculation is performed on the tea set, and binary classification labels are added to construct a multi-dimensional input vector;

[0037] A bud identification and classification model is constructed, and the multi-dimensional input vector is input into the bud identification and classification model to obtain the identification result of tea buds.

[0038] Specifically, the instance segmentation algorithm includes but is not limited to the yolov8 algorithm.

[0039] Specifically, the image data of tea leaves to be identified under different viewing angles are captured, including:

[0040] Use multiple rigidly connected industrial cameras to make their fields of view intersect, and the tea leaves are located in the intersecting fields of view. Figure 2 As shown, four industrial cameras are used to capture the tea leaves to be identified from different perspectives to avoid incomplete perspectives such as occlusion.

[0041] Specifically, for the image data of tea leaves obtained by cameras 1 to 4, an instance segmentation algorithm is used to implement instance segmentation of tea leaves. Here, whether or not the leaves are tender buds is not distinguished, and only tea leaves set A are segmented. At the same time, the stems of the tea leaves are segmented, the main trunk is extracted, and the stem set B is segmented.

[0042] Specifically, 1. Collect 10,000 pictures of young tea buds and construct a standard data set DATASET;

[0043] 2. Manual labeling using LabelMe;

[0044] 3. Use polygons to classify tea leaves and stems in LabelMe;

[0045] 4. Divide DATASET into training set, test set and validation set according to 8:1:1 for model training and validation;

[0046] 5. Select YOLOv8x-seg as the segmentation model for training;

[0047] 6. Save the best weight file during training as the result.

[0048] Specifically, constructing a graph structure based on the tea set and the stem set includes:

[0049] Extract the tea set A after instance segmentation, and take the centroid A of the tea set A as center is a leaf node; centroid A center It is the geometric center of the pixels of each leaf image after instance segmentation; use the centroid A center It can better show the spatial relationship between tea leaves and stems;

[0050] The two endpoints of stem set B are path B end , and the path length is used as the path weight β, which is used to calculate the weighted distance from the leaf node to the root node in the future; calculating the weighted distance can better represent the spatial location attribute of the leaf node;

[0051] Matching centroid A center To Path B end The nearest end of the graph structure Graph(A center , B end ), and then obtain the tree structure through pruning. Among them, path B end It contains two endpoints, and the centroid of the tea leaf is the leaf node of this path; the tea leaf is the leaf node, the stem is the path, and the nodes and paths are associated to form a graph structure. By pruning and merging the nodes where the stems intersect, it is ensured that each node has only one parent node, thus forming a tree structure.

[0052] Specifically, feature calculation is performed on the tea set, including:

[0053] Calculate the growth position, size, aspect ratio and HSV channel histogram of the tea leaves.

[0054] Specifically, the binary classification labels include young buds and green leaves of tea.

[0055] Specifically, the growth position Pos, size (pixel size) Size, aspect ratio AspectRatio, HSV channel histogram HSV and other features are calculated for the tea set A, and the two-category labels of tender buds and green leaves are added to construct a multi-dimensional input vector FeatureArray; wherein,

[0056] Pos(a)=∑distance(node,a center ), where node is a center All parent nodes, a∈A, a center ∈A center ;

[0057] Size = Count(pixel);

[0058] AspectRatio = Height / Width.

[0059] The image captured by the camera is an RGB image, and the process of converting it to HSV is as follows:

[0060] For 8 bits, R, G, and B are converted to floating point format and scaled to fit in the range 0 to 1:

[0061] V←max(R, G, B)

[0062]

[0063] If H<0, then H←H+360, output 0≤V≤1, 0≤S≤1, 0≤H≤360;

[0064] The values ​​are then converted to the target data type:

[0065] V←255V, S←255S, H←H / 2 (fits 0 to 255).

[0066] Specifically, SVM (Support Vector Machine) is used for binary classification, where the input is FeatureArray. Through training and verification, a classification model that supports sprout recognition is finally trained.

[0067] Specifically, the embodiment of the present invention is aimed at identifying young buds of white tea.

[0068] In a specific embodiment of the present invention, a machine vision-based sprout recognition system includes:

[0069] A tea image capture module, used to capture image data of tea leaves to be identified at different viewing angles;

[0070] An instance segmentation module is used to perform instance segmentation on the image data using an instance segmentation algorithm to obtain a tea set; and to segment the stems of the tea leaves to extract the main trunk to obtain a stem set;

[0071] A tree structure acquisition module, used to construct a graph structure based on the tea set and the stem set, and then acquire a tree structure by pruning;

[0072] A feature calculation module, used to calculate the features of the tea set based on the tree structure, add binary classification labels, and construct a multi-dimensional input vector;

[0073] The tea bud recognition module is used to construct a bud recognition classification model, input the multi-dimensional input vector into the bud recognition classification model, and obtain the recognition result of the tea bud.

[0074] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0075] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying young shoots based on machine vision, characterized in that: include: Capture image data of tea leaves to be identified at different viewing angles; Performing instance segmentation on the image data using an instance segmentation algorithm to obtain a tea set; At the same time, the tea stems are divided, the main trunk is extracted, and a stem set is obtained; Constructing a graph structure based on the tea set and the stem set, and then obtaining a tree structure through pruning; Based on the tree structure, feature calculation is performed on the tea set, and binary classification labels are added to construct a multi-dimensional input vector; A tender bud recognition and classification model is constructed, and the multi-dimensional input vector is input into the tender bud recognition and classification model to obtain a recognition result of tea tender buds.

2. The method for identifying young sprouts based on machine vision according to claim 1, characterized in that: The binary classification labels include young buds and green leaves of tea leaves.

3. The method for identifying young sprouts based on machine vision according to claim 1, characterized in that: Constructing a graph structure based on the tea set and the stem set includes: The centroid A of the tea leaves center is a leaf node, and the two endpoints of the stem set are path B end , and use the path length as the path weight β to match the centroid A center To Path B end The nearest end of , builds the graph structure.

4. The method for identifying tender shoots based on machine vision according to claim 1, characterized in that: Capture image data of tea leaves to be identified from different viewing angles, including: A plurality of rigidly connected industrial cameras are used to make their fields of view intersect, and the tea leaves are located within the intersecting fields of view.

5. The method for identifying tender shoots based on machine vision according to claim 1, characterized in that: Perform feature calculations on the tea collection, including: Calculate the growth position, size, aspect ratio and HSV channel histogram of the tea leaves.

6. The method for identifying tender shoots based on machine vision according to claim 1, characterized in that: SVM is used for binary classification.

7. A machine vision-based sprout recognition system, using the machine vision-based sprout recognition method according to any one of claims 1 to 6, characterized in that: include: A tea image capture module, used to capture image data of tea leaves to be identified at different viewing angles; An instance segmentation module is used to perform instance segmentation on the image data using an instance segmentation algorithm to obtain a tea set; and to segment the stems of the tea leaves to extract the main trunk to obtain a stem set; A tree structure acquisition module, used to construct a graph structure based on the tea set and the stem set, and then acquire a tree structure by pruning; A feature calculation module, used to calculate the features of the tea set based on the tree structure, add binary classification labels, and construct a multi-dimensional input vector; The tea bud recognition module is used to construct a bud recognition classification model, input the multi-dimensional input vector into the bud recognition classification model, and obtain the recognition result of the tea bud.

Images