A tea garden tea picking planning system and method based on visual recognition

Abstract

The invention relates to the technical field of agricultural picking, and discloses a tea garden tea picking planning system and method based on visual recognition, a data acquisition module, which is used for collecting illumination images and day-night temperature difference data of M mu of tea fields in the tea garden every day; a graph structure module, which is used for constructing a topological map, each node of the topological map is mapped to one mu of tea fields, a graph shielding module, which is used for shielding some nodes of the topological map based on illumination images and picking time points of the tea fields to obtain an updated topological map, and a path planning module, which is used for determining the picking path of tea leaves in the tea garden based on the updated topological map; the invention collects and analyzes illumination images and day-night temperature difference data of the tea garden, and combines the characteristics of the tea growing environment to judge tea leaves that have just matured and have a high tea polyphenol content, so as to select tea leaves that have just matured and have a high tea polyphenol content for picking in a targeted manner.

Description

Technical Field

[0001] The present invention relates to the field of agricultural picking, and more specifically, to a tea garden tea picking planning system and method based on visual recognition. Background Art

[0002] In the traditional tea garden management and tea picking process, the experience and intuition of the pickers are usually the main basis for deciding the time to pick tea. However, this method is often difficult to achieve the desired effect when facing the complex natural environment and the growth conditions of tea trees, especially in ensuring the excellent quality of tea.

[0003] First of all, tea polyphenols are an important component of tea, and their content gradually increases as the tea leaves grow. However, the best time to pick tea is not when the tea polyphenol content is the highest, but when the tea leaves are just mature and have a high tea polyphenol content while still remaining fresh and tender. The traditional picking method cannot accurately select the picking time, which may lead to picking too early and insufficient tea polyphenol content; or picking too late and aging of the tea leaves; secondly, the water content of tea leaves also has an important impact on their subsequent processing and storage. When tea leaves are picked when they are too fresh and tender, although the tea polyphenol content is appropriate, the water content is high, and the tea leaves are easily damaged due to mechanical damage during transportation, affecting their appearance and final quality. Traditional picking methods usually make it difficult to effectively control the water content of tea leaves during picking, resulting in damage to the quality of tea leaves during picking and transportation. Summary of the invention

[0004] The present invention provides a tea garden tea picking planning system and method based on visual recognition, which solve the technical problems raised in the background technology.

[0005] In the first aspect, a tea garden tea picking planning system based on visual recognition comprises:

[0006] Data acquisition module: used to collect the illumination images and day-night temperature difference data of M mu of tea fields in the tea garden every day, and pre-process the illumination images and day-night temperature difference data to obtain the pre-processed illumination images and day-night temperature difference data; according to the pre-processed illumination images, identify the direct illumination time of each mu of tea fields during the day, and judge whether the tea fields are in the illumination area or the shadow area at the preset time;

[0007] Graph structure module: used to construct a topological graph, where each node of the topological graph is mapped to an acre of tea field. During the construction process, the direct illumination time and the temperature difference between day and night of each acre of tea field are used as the initial features of the corresponding node; wherein the edges between nodes are constructed based on the feature similarity between nodes, specifically including: setting a feature similarity threshold, when the feature similarity between any two nodes is greater than the preset feature similarity threshold, establishing an edge between the two nodes; determining the weight of the edge between the two nodes according to the minimum path length between the two acres of tea fields mapped by the two nodes;

[0008] Graph shielding module: used to shield some nodes of the topology graph based on the illumination image and the picking time of the tea field to obtain an updated topology graph;

[0009] Path planning module: used to determine the tea picking path in the tea garden based on the updated topological map.

[0010] Furthermore, based on the judgment result that each acre of tea field is in the illumination area or the shadow area at the preset moment, the continuous illumination time and the continuous shading time of each acre of tea field at the preset moment are calculated.

[0011] Furthermore, some nodes of the topology map are shielded based on the illumination image and the picking time of the tea field to obtain an updated topology map, including:

[0012] Based on the picking time of each acre of tea field, combined with the day and night temperature difference of each acre of tea field and the daily direct sunlight time of each acre of tea field, the tea growth factor of each acre of tea field is obtained, and the first shielding time of the node mapped by each acre of tea field is determined according to the tea growth factor;

[0013] Based on the continuous illumination time and continuous shielding time of each acre of tea field at a preset time, the moisture factor of tea leaves in each acre of tea field is calculated, and the second shielding time of the node mapped to each acre of tea field is determined according to the moisture factor;

[0014] Part of the nodes in the topology graph are shielded based on the first shielding time and the second shielding time to obtain an updated topology graph.

[0015] Furthermore, based on the illumination image and the picking time point of the tea field, some nodes of the topological map are shielded to obtain an updated topological map, which also includes:

[0016] According to the direct sunlight time and the temperature difference between day and night of each acre of tea field on that day, the current features of the nodes mapped to each acre of tea field are updated to obtain an updated topological map; wherein the current features are obtained based on the initial feature update in this way. Further, the picking path of tea leaves in the tea garden is determined based on the updated topological map, including:

[0017] Step S201: Calculate the tea growth score of each node according to the current characteristics of each node in the topological map;

[0018] Step S202: masking some nodes in the graph network according to the first masking time and the second masking time to determine the nodes to be selected;

[0019] Step S203: determining the tea growth score of the node to be selected according to the node to be selected; determining the initial access node based on the tea growth score of the node to be selected, and updating the shielding state of each node and the tea growth score of each node;

[0020] Step S204: Determine the tea growth score of the point to be selected and the corresponding node again, select the next visited node in combination with the edge weight of the currently visited node, and update the shielding state of each node and the tea growth score of each node;

[0021] Step S205: loop step S204 to dynamically determine the tea picking path.

[0022] Further, the illumination image is preprocessed to obtain a preprocessed illumination image, including:

[0023] De-noising the illumination image to obtain a denoised illumination image;

[0024] Processing the grayscale of the denoised illumination image to obtain a grayscale illumination image;

[0025] Perform atmospheric correction on the grayscale illumination image to obtain a preprocessed illumination image.

[0026] Furthermore, in the process of shielding a node, a shielding mark is added to the feature vector of the node to be shielded, and the mark is used to indicate that the node does not participate in the tea picking path planning during the shielding period; after the shielding time is over, the shielding mark of the node is removed.

[0027] In a second aspect, a tea garden tea picking planning method based on visual recognition is applied to the system, comprising:

[0028] Collect illumination images and day-night temperature difference data of M mu of tea fields in the tea garden every day, and pre-process the illumination images and day-night temperature difference data to obtain pre-processed illumination images and day-night temperature difference data; identify the direct illumination time of each mu of tea field during the day according to the pre-processed illumination images, and judge whether the tea field is in the illumination area or the shadow area at the preset time;

[0029] Construct a topological map, where each node of the topological map is mapped to an acre of tea field. During the construction process, the direct sunlight time and the temperature difference between day and night of each acre of tea field are used as the initial features of the corresponding node;

[0030] Based on the illumination image and the picking time of the tea field, some nodes of the topological map are shielded to obtain an updated topological map;

[0031] The tea picking path in the tea garden is dynamically determined based on the updated topological graph.

[0032] According to a third aspect, a computing device includes:

[0033] one or more processors;

[0034] The storage device is used to store one or more programs. When the one or more programs are executed by one or more processors, the one or more processors implement the system.

[0035] In a fourth aspect, a computer-readable storage medium stores a program, and the program implements the system when executed by a processor.

[0036] The beneficial effects of the present invention are:

[0037] 1. The system collects and analyzes the lighting images and day-night temperature difference data of the tea garden, combined with the characteristics of the tea growing environment, to determine the tea leaves that may have a higher content of tea polyphenols. Based on these environmental factors, such as direct light time, day-night temperature difference, etc., the system dynamically adjusts the picking plan, so as to selectively select the tea leaves that have just matured and have a higher content of tea polyphenols for picking.

[0038] 2. The system determines the moisture content of tea leaves when they are picked by analyzing the lighting images of the tea garden. By picking the tea leaves when the moisture content is low, it avoids the tea leaves being too fresh and tender, which would cause damage to the appearance of the tea leaves during transportation and processing, thereby improving the quality and integrity of the tea leaves. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is an architecture diagram of a tea garden tea picking planning system based on visual recognition of the present invention;

[0040] Figure 2 It is a flow chart of tea picking path planning based on topological graph of the present invention.

[0041] In the figure: 101, data acquisition module; 102, graph structure module; 103, graph shielding module; 104, path planning module. DETAILED DESCRIPTION

[0042] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0043] like Figure 1-2 As shown, a tea garden tea picking planning system based on visual recognition includes:

[0044] Data collection module 101: used to collect illumination images and day-night temperature difference data of M mu of tea fields in the tea garden every day, and pre-process the illumination images and day-night temperature difference data to obtain pre-processed illumination images and day-night temperature difference data; according to the pre-processed illumination images, identify the direct illumination time of each mu of tea fields during the day, and determine whether the tea fields are in the illumination area or the shadow area at the preset time;

[0045] Graph structure module 102: used to construct a topological graph, wherein each node of the topological graph is mapped to an acre of tea field. During the construction process, the direct illumination time and the temperature difference between day and night of each acre of tea field are used as the initial features of the corresponding node; wherein the edge between nodes is constructed based on the feature similarity between the nodes, specifically including: setting a feature similarity threshold, when the feature similarity between any two nodes is greater than the preset feature similarity threshold, establishing an edge between the two nodes; determining the weight of the edge between the two nodes according to the minimum path length between the two acres of tea fields mapped by the two nodes;

[0046] The graph shielding module 103 is used to shield some nodes of the topological graph based on the illumination image and the picking time point of the tea field to obtain an updated topological graph;

[0047] The path planning module 104 is used to determine the path for picking tea leaves in the tea garden based on the updated topological map.

[0048] It should be noted that under the modern agricultural management level, after eliminating the differences between individual plants during the growth of tea trees, the water, fertilizer and pest control are kept at the same level; therefore, the difference in tea leaves is reflected in the content of tea polyphenols; tea gardens are usually built beside mountains, and the objective difference of tea trees is reflected in the planting height. The differences caused by different heights include the temperature difference between day and night, the light blocking by clouds and fog, and the light blocking by the mountain shape itself, which ultimately leads to significant differences in the tea polyphenol content of tea leaves from different tea trees;

[0049] Preferably, the preprocessing method of the illumination image includes denoising, grayscale conversion and atmospheric correction of the illumination image.

[0050] It should be noted that the CNN-GRU neural network model is used to identify the direct illumination time of the M mu tea garden during the day, and to determine whether the tea field is in the illumination area or the shadow area at the preset time; for example, the historical illumination image pictures of the M mu tea field are collected and input into the CNN, and the CNN outputs the feature map of the tea field, and then the feature map of the obtained illumination image is flattened into a vector by row and input into the GRU, and the GRU output indicates that the tea field is in the illumination area or the shadow area at the next time point; the output of the GRU is used as the historical data input, and the data of T days can be used to predict whether the M mu tea field is in the illumination area or the shadow area at the preset time of the T+1 day, and the direct illumination time of the M mu tea garden during the day is calculated based on the prediction result; wherein, T is greater than 1; for example, one mu of tea field is in the illumination area for 6 hours in the 24 hours of the T+1 day, and the direct illumination time of one mu of tea field is 6 hours; preferably, in order to alleviate the gradient problem, the fusion of GRU and self-attention mechanism is selected to improve the GRU neural network model;

[0051] For example, the vector of the feature map flattened by row is represented as X, and X is embedded in the embedding layer of the self-attention mechanism as the embedding vector ;Will Input into GRU, the calculation formula is as follows:

[0052] The calculation formula of the reset gate of GRU is as follows:

[0053] ;in, Represents the embedding vector The output after iteration t times, represents the output of the reset gate, and Respectively represent the weight parameters of the reset gate, represents the bias of the reset gate, express The hidden state of the iteration, express Activation function;

[0054] The calculation formula of GRU's update gate is as follows:

[0055] ;in, represents the output of the update gate, and represents the weight parameter of the update gate, represents the bias of the update gate;

[0056] The calculation formula for the candidate hidden door of GRU is as follows:

[0057] ;in, represents the candidate hidden state of the candidate hidden gate, and represents the weight parameter of the candidate hidden door, represents the bias of the candidate hidden door, represents the hyperbolic tangent function;

[0058] The final hidden state of GRU is calculated as follows:

[0059] ;in, represents the final hidden state;

[0060] Will As a query for the self-attention mechanism, Queries and values ​​as self-attention mechanisms;

[0061] The calculation formula of the self-attention mechanism is as follows:

[0062] ;in, Represents the query matrix of the self-attention mechanism, with dimension , express The weight matrix of

[0063] ;in, Represents the key matrix of the self-attention mechanism, with dimension , express The weight matrix of

[0064] ;in, Represents the key matrix of the self-attention mechanism, with dimension , express The weight matrix of

[0065] ;in, represents the weight of the self-attention mechanism, express Activation function, Represents a transpose operation;

[0066] ;in, represents the residual, Represents the dimension of vector X;

[0067] + ;in, Express + Classification calculation of

[0068] GRU captures the short-term dependencies of vector X, while the self-attention mechanism can directly model long-distance dependencies, reducing the need for deep GRU layers; the use of residual connections enables the output of each hidden layer of GRU to be directly passed to subsequent layers, helping to maintain stable propagation of gradients, thereby alleviating the common gradient problem of GRU.

[0069] In one embodiment of the present invention, based on the judgment result that each acre of tea field is in the illumination area or the shadow area at the preset moment, the continuous illumination time and the continuous shielding time of each acre of tea field at the preset moment are calculated.

[0070] It should be noted that, through the output of the CNN-GRU neural network model, it is obtained that one acre of tea field is in the illuminated area at the i-th moment on the T+1 day, and in the shadow area at the j-th moment, then the continuous illumination time of the tea field is ji; if the tea field is in the shadow area at the j-th moment and in the illuminated area at the i-th moment, then the continuous shading time of the tea field is ij; at the k-th moment, if the tea field is in the illuminated area, the continuous illumination time is calculated, and if it is in the shadow area, the continuous shadow time is calculated. Based on this method, the continuous illumination time and continuous shadow time of each acre of tea garden are predicted.

[0071] It should be noted that a duration threshold is set. If the continuous illumination time and the continuous shading time are lower than the duration threshold, they are considered to be continuous and uninterrupted. The duration threshold indicates the shortest time for the tea leaves to reach moisture balance when alternating between the illumination area and the shadow area. For example, the tea leaves are in the illumination area from the i-th moment, and the continuous illumination time is calculated at the k-th moment. If ki is less than the duration threshold, the continuous shading time is calculated from the j-th moment when the tea field was in the shadow area from the previous moment.

[0072] It should be noted that the topological map has M nodes, and the feature of each node is a normalized vector representation constructed by the normalized direct illumination time and day and night temperature difference of the corresponding mapped tea field; the feature similarity of each node and other preset nodes is calculated, and the feature similarity is the cosine similarity , the formula is:

[0073] ;in, and They represent the feature vector of the i-th node and the feature vector of the j-th node respectively. represents the dot product, represents the modulus of the eigenvector;

[0074] The feature similarity threshold is set, preferably 0.95; for example, if the feature similarity between the i-th tea field and the j-th tea field is 0.96, an edge is constructed between the i-th node and the j-th node; if the feature similarity between the u-th tea field and the v-th tea field is 0.88, no edge is constructed between the u-th node and the v-th node.

[0075] It should be noted that the shortest path length between two acres of tea fields can be calculated by calculating the Euclidean distance between the two acres of tea fields. , the formula is:

[0076] ;

[0077] in, represents the coordinates of the i-th acre of tea field, represents the coordinates of the j-th acre of tea field;

[0078] Edge weight According to the shortest path length and feature similarity between two acres of tea fields, the formula is expressed as:

[0079] ;in, The weight parameter representing the feature similarity, The weight parameter representing the shortest path length, and are not 0, and .

[0080] In one embodiment of the present invention, shielding some nodes of the topological map based on the illumination image and the picking time point of the tea field to obtain an updated topological map includes:

[0081] Based on the picking time of each acre of tea field, combined with the day and night temperature difference of each acre of tea field and the daily direct sunlight time of each acre of tea field, the tea growth factor of each acre of tea field is obtained, and the first shielding time of the node mapped by each acre of tea field is determined according to the tea growth factor;

[0082] Based on the continuous illumination time and continuous shielding time of each acre of tea field at a preset time, the moisture factor of tea leaves in each acre of tea field is calculated, and the second shielding time of the node mapped to each acre of tea field is determined according to the moisture factor;

[0083] Part of the nodes in the topology graph are shielded based on the first shielding time and the second shielding time to obtain an updated topology graph.

[0084] It should be noted that if the j-th acre of tea field was picked at the i-th time point, then the picking time point of the j-th acre of tea field is i; the tea growth factor is expressed as the growth rate of tea leaves. Combined with the day and night temperature difference and direct light time of the j-th acre of tea field, the tea growth factor of the j-th acre of tea is calculated. , the formula is:

[0085] ;in, represents the temperature difference between day and night in the j-th acre of tea field, represents the direct illumination time of the j-th acre of tea field, Indicates standard parameters, represents the average temperature of the j-th acre of tea field, and They represent the weight parameter of the day-night temperature difference of the j-th acre of tea field and the weight parameter of the direct sunlight time of the j-th acre of tea field respectively;

[0086] Based on the tea growth factor, it can be obtained that the first shielding time is required for the tea leaves in the j-mu tea field to reach the minimum picking length ; The formula is:

[0087] ;in, Indicates the minimum picking length of tea leaves, which is a custom parameter;

[0088] During the first shielding time period after the i-th moment, the j-th node in the topology graph is actively shielded and does not participate in the calculation of the picking path planning task.

[0089] It should be noted that after picking, the tea leaves are stored in the storage bags of the picking machines. The higher the moisture content of the tea leaves, the tenderer the tea leaves are, and the higher the proportion of cargo damage during transportation to the factory for processing; therefore, the tea leaves need to be picked when the dynamic moisture content of the tea leaves is low.

[0090] For example, when the continuous illumination time of the i-th acre of tea field is When the total amount of tea leaves evaporates, , the formula is:

[0091] ;in, Indicates the evaporation rate of tea leaves per unit time. Indicates the temperature of the illuminated area;

[0092] When the continuous shading time of the i-th acre of tea field is When the total water absorption of tea leaves is , the formula is:

[0093] ;in, Indicates the amount of water absorbed by tea leaves per unit time. represents the temperature of the shaded area;

[0094] Setting the evaporation threshold If the total evaporation of tea leaves during continuous illumination exceeds the evaporation threshold, it means that the tea field can be picked; set the water absorption threshold If the total water absorption of tea leaves during the continuous shading time exceeds the water absorption threshold, it means that the tea field cannot be picked; the calculation results are The tea fields are ready for picking. The tea field cannot be picked during this time, the second shielding time The formula is:

[0095] ;in Represents the union operation of sets;

[0096] During the second shielding time, the i-th node in the topology graph is actively shielded and does not participate in the calculation of the picking path planning task.

[0097] In one embodiment of the present invention, shielding some nodes of the topological map based on the illumination image and the picking time point of the tea field to obtain an updated topological map also includes:

[0098] According to the direct sunlight time and the temperature difference between day and night of each acre of tea field on that day, the current features of the nodes mapped to each acre of tea field are updated to obtain an updated topological map; wherein the current features are obtained based on the initial feature update in this way. It should be noted that the feature update of node i is expressed as:

[0099] ;in, represents the feature vector of node i at time t+1, represents the feature vector of node i at time t, It represents the feature vector composed of the direct sunlight time and the day and night temperature difference collected by node i in the time period from t to t+1. Represents the weight parameter of the feature vector of node i at time t.

[0100] In one embodiment of the present invention, determining a tea picking path in a tea garden based on an updated topological map includes:

[0101] Step S201: Calculate the tea growth score of each node according to the current characteristics of each node in the topological map;

[0102] Step S202: Shielding some nodes in the graph network according to the first shielding time and the second shielding time, and determining the nodes to be selected;

[0103] Step S203: based on the tea growth score of the node to be selected, select the initial access node, and after accessing the initial node, update the shielding state of each node and the tea growth score of each node;

[0104] Step S204: Determine the point to be selected again, select the next visited node according to the tea growth score of the node to be selected and the weight of the edge of the currently visited node, and update the shielding state of each node and the tea growth score of each node;

[0105] Step S205: loop step S204 to dynamically determine the tea picking path.

[0106] It should be noted that the tea growth score of the jth node is calculated based on the current characteristics of the jth node in the topological graph. , the formula is:

[0107] ;in, represents the tea growth factor of the j-mu tea field The weight parameter, represents the weight parameter of the day-night temperature difference and light duration of the j-th acre of tea field, It represents the time interval between the j-th acre of tea field and the last picking time point;

[0108] Preferably, the tea growth score threshold of the j-th acre tea field is 60 points. When the tea growth score of the j-th acre tea field is lower than 60 points, it indicates that the j-th acre tea field is in the first shielding time; when the tea growth score of the j-th acre tea field is equal to 60 points, .

[0109] It should be noted that when the picking path planning is selected at the i-th moment, some nodes in the topological map are first shielded according to the first shielding time and the second shielding time, and then a node with the highest tea growth score is found in the unshielded nodes as the initial access node. After the initial access node is picked, the shielded nodes in the topological map are updated. If the i-th moment is on the T-th day and the i+1-th moment is on the T+1-th day, the feature vectors of the M nodes in the topological map are updated based on the feature vector composed of the day and night temperature difference and direct light time of the M tea fields collected on the T-th day;

[0110] For example, a simple graph network: the nodes include: , the edges include: ; Select B as the current node, where both node A and node C have direct edges with node B, and are first-degree nodes, and node D is a second-degree node of node E; where node A is shielded, and the node to be selected is node C;

[0111] After the topology map is updated, the node similarity is recalculated and the node edges and edge weights are constructed. When visiting the next tea field, based on the node mapped to the current tea field, select the node with an edge built with the node and not blocked, and select the node according to the edge weight and tea growth score. The formula is expressed as:

[0112] ;in, Indicates the node to be selected. represents the tea growth score of the node to be selected, Represents the weight of the edge between the node to be selected and the current node, and max represents the max function;

[0113] For example, a simple topology diagram: the nodes include: , the edges include: ; Select B as the current node, where both node A and node C have direct edges with node B and are first-degree nodes, and node D is a second-degree node of node E; where nodes A and C are shielded, and the nodes to be selected are nodes D and E;

[0114] If there is an edge with the current node, but there is no unshielded node; then based on the second-degree nodes of the current node, select the unshielded node among the second-degree nodes; the second-degree nodes include: ;

[0115] The nodes are selected based on the edge weight and tea growth score. The formula is:

[0116] ;in, Indicates the blocked first-degree node, Indicates the node to be selected. Represents the weight of the edge between the current node and the first-degree node, Indicates the tea growth score at the first node. Indicates the node to be selected. represents the tea growth score of the node to be selected, Represents the weight of the edge between the node to be selected and the first-degree node;

[0117] Based on the above process, the picking path selection formula is obtained:

[0118] ;in, represents the degree of the node currently considered; Represents the first-degree node, that is, the node directly connected to the current node; represents the i-th degree node, that is, the adjacent node of the i-1-th degree node; Represents the i-th degree node Tea growth scoring; Represents the i-1th degree node and the i-th degree node The weight of the edge between them; Indicates the product of the score in each path and the weight of the edge; Indicates that the path with the highest score is selected among all the paths.

[0119] The picking path is obtained by looping the above process.

[0120] It should be noted that this system plans the picking route for high-quality tea, rather than traversal route planning for the entire tea garden; therefore, some tea fields will not be included in the picking plan because their light exposure time and day-night temperature difference data cannot prove the good quality of their tea. The tea in these tea fields needs to be manually planned and picked by farmers based on their own experience.

[0121] In one embodiment of the present invention, preprocessing the illumination image to obtain the preprocessed illumination image includes:

[0122] De-noising the illumination image to obtain a denoised illumination image;

[0123] Processing the grayscale of the denoised illumination image to obtain a grayscale illumination image;

[0124] Perform atmospheric correction on the grayscale illumination image to obtain a preprocessed illumination image.

[0125] In one embodiment of the present invention, in the process of shielding a node, a shielding mark is added to the feature vector of the node to be shielded, and the mark is used to indicate that the node does not participate in the tea picking path planning during the shielding period; after the shielding time ends, the shielding mark of the node is removed.

[0126] A tea garden tea picking planning method based on visual recognition, applied to the system, comprises:

[0127] Collect illumination images and day-night temperature difference data of M mu of tea fields in the tea garden every day, and pre-process the illumination images and day-night temperature difference data to obtain pre-processed illumination images and day-night temperature difference data; identify the direct illumination time of each mu of tea field during the day according to the pre-processed illumination images, and judge whether the tea field is in the illumination area or the shadow area at the preset time;

[0128] Construct a topological map, where each node of the topological map is mapped to an acre of tea field. During the construction process, the direct sunlight time and the temperature difference between day and night of each acre of tea field are used as the initial features of the corresponding node;

[0129] Based on the illumination image and the picking time of the tea field, some nodes of the topological map are shielded to obtain an updated topological map;

[0130] Determine the tea picking path in the tea garden based on the updated topological map.

[0131] A computing device comprising:

[0132] one or more processors;

[0133] The storage device is used to store one or more programs. When the one or more programs are executed by one or more processors, the one or more processors implement the system.

[0134] A computer-readable storage medium stores a program, and the program implements the system when executed by a processor.

[0135] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A tea garden tea picking planning system based on visual recognition, characterized in that: include: Data collection module (101): used to collect illumination images and day-night temperature difference data of M mu of tea fields in a tea garden every day, and pre-process the illumination images and day-night temperature difference data to obtain pre-processed illumination images and day-night temperature difference data; based on the pre-processed illumination images, identify the direct illumination time of each mu of tea field during the day, and determine whether the tea field is in an illuminated area or a shadowed area at a preset time; A graph structure module (102): used to construct a topological graph, wherein each node of the topological graph is mapped to an acre of tea field. During the construction process, the direct illumination time and the temperature difference between day and night of each acre of tea field are used as the initial features of the corresponding node; wherein the edge between the nodes is constructed based on the feature similarity between the nodes, specifically including: setting a feature similarity threshold, when the feature similarity between any two nodes is greater than the preset feature similarity threshold, establishing an edge between the two nodes; determining the weight of the edge between the two nodes according to the minimum path length between the two acres of tea fields mapped by the two nodes; wherein the minimum path length represents the Euclidean distance between the two acres of tea fields; Graph shielding module (103): used to shield some nodes of the topological graph based on the illumination image and the tea field picking time point to obtain an updated topological graph, as follows: Based on the picking time of each acre of tea field, combined with the temperature difference between day and night and the daily direct sunlight time of each acre of tea field, the tea growth factor of each acre of tea field is obtained. The tea growth factor specifically includes: ;in, represents the tea growth factor of the j-mu tea field, represents the temperature difference between day and night in the j-th acre of tea field, represents the direct illumination time of the j-th acre of tea field, Indicates standard parameters, represents the average temperature of the j-th acre of tea field, and They represent the weight parameter of the day-night temperature difference of the j-th acre of tea field and the weight parameter of the direct sunlight time of the j-th acre of tea field respectively; According to tea growth factor Determine the first shielding time of the node mapped to the j-th acre of tea field. The calculation formula for the first shielding time is as follows: ;in, Indicates the minimum picking length of tea leaves; During the first shielding time period, the jth node in the topology graph is actively shielded and does not participate in the calculation of the picking path planning task; When the continuous illumination time of the i-th acre of tea field is When the total evaporation of tea leaves is , the formula is as follows: ;in, Indicates the evaporation rate of tea leaves per unit time. Indicates the temperature of the illuminated area; When the continuous shading time of the i-th acre of tea field is The total water absorption of tea leaves is , the formula is as follows: ;in, Indicates the amount of water absorbed by tea leaves per unit time. represents the temperature of the shaded area; Setting the evaporation threshold and water absorption threshold , get the second shielding time The formula is as follows: ;in Represents the union operation of sets; During the second shielding time, the i-th node in the topology graph is actively shielded and does not participate in the calculation of the picking path planning task; Based on the first shielding time and the second shielding time, shield some nodes of the topological graph to obtain an updated topological graph; Path planning module (104): used to determine the path for picking tea leaves in the tea garden based on the updated topological map.

2. A tea garden tea picking planning system based on visual recognition according to claim 1, characterized in that: Based on the judgment result that each acre of tea field is in the illuminated area or the shadowed area at the preset moment, the continuous illumination time and the continuous shading time of each acre of tea field at the preset moment are calculated.

3. The tea garden tea picking planning system based on visual recognition according to claim 2 is characterized in that: Based on the illumination image and the tea field picking time point, some nodes of the topology map are shielded to obtain an updated topology map, which also includes: According to the direct illumination time and the temperature difference between day and night of each acre of tea field on that day, the current features of the nodes mapped to each acre of tea field are updated to obtain an updated topological map; wherein the current features are obtained in this way based on the initial feature update.

4. The tea garden tea picking planning system based on visual recognition according to claim 3 is characterized in that: Determine the tea picking path in the tea garden based on the updated topology map, including: Step S201: Calculate the tea growth score of each node according to the current characteristics of each node in the topological map; Step S202: masking some nodes in the graph network according to the first masking time and the second masking time to determine the nodes to be selected; Step S203: determining the tea growth score of the node to be selected according to the node to be selected; determining the initial access node based on the tea growth score of the node to be selected, and updating the shielding state of each node and the tea growth score of each node; Step S204: Determine the tea growth score of the point to be selected and the corresponding node again, select the next visited node in combination with the edge weight of the currently visited node, and update the shielding state of each node and the tea growth score of each node; Step S205: loop step S204 to dynamically determine the tea picking path.

5. The tea garden tea picking planning system based on visual recognition according to claim 1 is characterized in that: Preprocessing the illumination image to obtain a preprocessed illumination image, including: De-noising the illumination image to obtain a denoised illumination image; Processing the grayscale of the denoised illumination image to obtain a grayscale illumination image; Perform atmospheric correction on the grayscale illumination image to obtain a preprocessed illumination image.

6. The tea garden tea picking planning system based on visual recognition according to claim 1 is characterized in that: In the process of shielding a node, a shielding mark is added to the feature vector of the node to be shielded, and the mark is used to indicate that the node does not participate in the tea picking path planning during the shielding period; after the shielding time is over, the shielding mark of the node is removed.

7. A tea garden tea picking planning method based on visual recognition, characterized in that: Applicable to the system according to any one of claims 1 to 6, comprising: Step 701, collecting illumination images and day-night temperature difference data of M mu of tea fields in the tea garden every day, and preprocessing the illumination images and day-night temperature difference data to obtain preprocessed illumination images and day-night temperature difference data; identifying the direct illumination time of each mu of tea field during the day according to the preprocessed illumination images, and determining whether the tea field is in an illuminated area or a shadowed area at a preset time; Step 702, constructing a topological map, wherein each node of the topological map is mapped to an acre of tea field, and during the construction process, the direct illumination time and the temperature difference between day and night of each acre of tea field are used as the initial features of the corresponding node; wherein the edge between the nodes is constructed based on the feature similarity between the nodes, specifically including: setting a feature similarity threshold, when the feature similarity between any two nodes is greater than the preset feature similarity threshold, establishing an edge between the two nodes; determining the weight of the edge between the two nodes according to the minimum path length between the two acres of tea fields mapped by the two nodes; wherein the minimum path length represents the Euclidean distance of the two acres of tea fields; Step 703, based on the illumination image and the tea field picking time point, some nodes of the topological map are shielded to obtain an updated topological map, as follows: Based on the picking time of each acre of tea field, combined with the temperature difference between day and night and the daily direct sunlight time of each acre of tea field, the tea growth factor of each acre of tea field is obtained. The calculation formula of tea growth factor is as follows: ;in, represents the tea growth factor of the j-mu tea field, represents the temperature difference between day and night in the j-th acre of tea field, represents the direct illumination time of the j-th acre of tea field, Indicates standard parameters, represents the average temperature of the j-th acre of tea field, and They represent the weight parameter of the day-night temperature difference of the j-th acre of tea field and the weight parameter of the direct sunlight time of the j-th acre of tea field respectively; According to tea growth factor Determine the first shielding time of the node mapped to the j-th acre of tea field. The calculation formula for the first shielding time is as follows: ;in, Indicates the minimum picking length of tea leaves; During the first shielding time period, the jth node in the topology graph is actively shielded and does not participate in the calculation of the picking path planning task; When the continuous illumination time of the i-th acre of tea field is When the total evaporation of tea leaves is , the formula is as follows: ;in, Indicates the evaporation rate of tea leaves per unit time. Indicates the temperature of the illuminated area; When the continuous shading time of the i-th acre of tea field is The total water absorption of tea leaves is , the formula is as follows: ;in, Indicates the amount of water absorbed by tea leaves per unit time. represents the temperature of the shaded area; Setting the evaporation threshold and water absorption threshold , get the second shielding time The formula is as follows: ;in Represents the union operation of sets; During the second shielding time, the i-th node in the topology graph is actively shielded and does not participate in the calculation of the picking path planning task; Based on the first shielding time and the second shielding time, shield some nodes of the topological graph to obtain an updated topological graph; Step 704, determining the tea picking path in the tea garden based on the updated topological map.

8. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by one or more processors, the one or more processors implement the system according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: A program is stored in the computer-readable storage medium, and when the program is executed by a processor, the system according to any one of claims 1 to 6 is implemented.

Images