A method and system for intelligent optimization of lighting for plant growth

Abstract

The present invention relates to the technical field of plant growth control, and in particular to a lighting intelligent optimization method and system for plant growth. The method comprises the following steps: obtaining a plant image through an image acquisition device and inputting the image into a plant life cycle recognition model to obtain a plant growth stage; obtaining a target lighting part category through comparison with a preset key part lighting table according to the plant growth stage; obtaining target lighting part position data of the plant through a plant key part positioning model according to the plant image and the target lighting part category; obtaining a lighting path through a lighting path iteration model according to the target lighting part position data of the plant; retrieving a lighting strategy database according to the plant growth stage to obtain lighting strategy parameters; and performing directionally illuminating the target lighting part of the plant according to the lighting strategy parameters and the lighting path, thereby achieving efficient plant growth and reducing light energy waste.

Description

Technical Field

[0001] The present invention relates to the technical field of plant growth control, and in particular to a lighting intelligent optimization method and system for plant growth. Background Art

[0002] In traditional plant cultivation methods, plant growth mainly depends on natural light. Due to the uncertainty of climatic conditions, relying on natural light can no longer meet the efficient and stable production needs in modern agriculture. With the development of science and technology, artificial light has gradually been used in plant cultivation to make up for the lack of natural light. However, since traditional artificial lighting technology mainly uses a unified lighting mode and a fixed lighting duration, it is difficult to cope with the growth of different types of plants, resulting in low plant growth efficiency.

[0003] At present, some research teams have focused on studying intelligent lighting methods and systems, using sensors to monitor plant growth data and dynamically adjust environmental parameters to optimize plant growth. However, these intelligent lighting methods and systems focus on adjusting the overall environment of the plant. The lighting often covers the entire plant area, ignoring the differences in the lighting requirements of different parts of the plant at different growth stages. For example, during the growth stage of the plant, the leaves need to be irradiated to promote photosynthesis, while during the fruit ripening stage, the fruit needs to be irradiated to enhance sweetness and pigment deposition. At this time, the leaves do not need too much light. Therefore, how to accurately illuminate specific parts according to the growth process of the plant and reduce unnecessary light energy consumption has become an urgent problem to be solved. Summary of the invention

[0004] (1) Technical issues to be resolved

[0005] The object of the present invention is to provide a method and system for intelligent optimization of lighting for plant growth, so as to solve the problem of light energy waste caused by intelligent lighting regulation for plant growth.

[0006] (2) Technical solution

[0007] To achieve the above objectives, on the one hand, the present invention provides a method for intelligent optimization of lighting for plant growth, the method comprising:

[0008] The plant image is acquired through an image acquisition device, and the plant image is input into a plant life cycle recognition model to obtain the plant growth stage; the life cycle recognition model is a mapping relationship between the plant image and the plant growth stage established according to a convolutional neural network algorithm; the plant growth stage includes the seedling stage, the growth stage, the flowering stage, and the fruit ripening stage.

[0009] According to the plant growth stage, the target illumination part category is obtained by comparing with the preset key part illumination table, and the target illumination part position data of the plant is obtained through the plant key part positioning model according to the plant image and the target illumination part category; the plant key part positioning model is a mapping relationship between the plant image, the target illumination part category and the target illumination part position data of the plant established according to the target detection algorithm combined with three-dimensional positioning.

[0010] According to the target illumination part position data of the plant, the illumination path is obtained through the illumination path iteration model. According to the plant growth stage, the illumination strategy database is retrieved to obtain the illumination strategy parameters. According to the illumination strategy parameters and the illumination path, the target illumination part of the plant is directed illuminated. The illumination strategy parameters include spectral wavelength, illumination intensity, and unit illumination duration.

[0011] Furthermore, the method of obtaining the target illumination part position data of the plant through the key part positioning model of the plant according to the plant image and the target illumination part category includes:

[0012] Acquire multi-angle plant images at a preset shooting position, use the YOLO target detection algorithm to identify the features of key parts of the plant images to obtain the categories and position bounding boxes of the key parts of the plant, filter the categories of the key parts of the plant according to the categories of the target illuminated parts, and extract local images according to the position bounding boxes to obtain the images of the target illuminated parts.

[0013] According to the target illumination part image, the local feature points of the target illumination part are extracted by the feature point detection algorithm to obtain the target illumination part descriptor, the target illumination part descriptors are matched to obtain the same points of the target illumination part descriptors and the same points are calibrated to obtain the target illumination part same point calibration data.

[0014] According to the same-point calibration data of the target illumination part, the two-dimensional point data is converted into three-dimensional point data through triangulation to obtain the three-dimensional coordinates of the target illumination part, and the three-dimensional coordinates of all target illumination parts are summarized as the target illumination part position data of the plant.

[0015] Furthermore, the method of extracting local feature points of the target illumination part through a feature point detection algorithm according to the target illumination part image to obtain a target illumination part descriptor, matching the target illumination part descriptors to obtain the same points of the target illumination part descriptors, and performing same-point calibration to obtain the same-point calibration data of the target illumination part includes:

[0016] The local feature points of the target illumination part are extracted through the ORB feature point detection algorithm to obtain the target illumination part descriptor.

[0017] The Hamming distance between the target illumination part descriptors in the target illumination part image at each angle is calculated as:

[0018] .

[0019] Among them is It is The mth object illumination part descriptor in the angular object illumination part image, is the first object illumination part descriptor, is the target illumination part descriptor and the target illumination part descriptor The Hamming distance between is the number of bits in the descriptor binary vector, is the target illumination part descriptor No. Bit, is the target illumination part descriptor No. Bit, is an exponential function, which means that if The result is 1 if they are not equal, otherwise 0.

[0020] For The minimum Hamming distance between each target illumination part descriptor in the j-th target illumination part image and each target illumination part descriptor in the j-th target illumination part image is used to obtain the first The nearest neighbor descriptor of each target illumination part descriptor in the angle target illumination part image in the jth angle target illumination part image for:

[0021] .

[0022] in is the index that minimizes the Hamming distance, It is Target Illumination Part Descriptor in Angular Target Illumination Part Image The nearest neighbor descriptor in the image of the object illuminated at the jth angle.

[0023] According to the For each target illumination part descriptor in the j-th angle target illumination part image, the second smallest Hamming distance between it and each target illumination part descriptor in the j-th angle target illumination part image is obtained. Each target illumination part descriptor in the angle target illumination part image is the second nearest neighbor descriptor in the jth angle target illumination part image. The matching reliability ratio obtained based on the nearest neighbor descriptor and the second nearest neighbor descriptor is:

[0024] .

[0025] in It is Target Illumination Part Descriptor in Angular Target Illumination Part Image The second nearest neighbor descriptor in the image of the target illumination part at the jth angle, is the matching reliability ratio.

[0026] The matching reliability ratio is compared with the preset reliability filtering threshold. When the matching reliability ratio is less than the preset reliability filtering threshold, the nearest neighbor descriptor is used as its same point for calibration, and all calibration points are summarized to obtain the same-point calibration data of the target illumination part.

[0027] Furthermore, the method of converting two-dimensional point data into three-dimensional point data by triangulation according to the same-point calibration data of the target illumination part to obtain the three-dimensional coordinates of the target illumination part includes:

[0028] The intrinsic parameter matrix and extrinsic parameter matrix of the image acquisition device are obtained through camera calibration.

[0029] The two-dimensional coordinates of the calibration data of the same point of the target illumination part at each angle are normalized by calculating the inverse matrix of the internal parameter matrix to obtain the two-dimensional normalized coordinates of the target illumination part.

[0030] The two-dimensional normalized coordinates of the target illumination part at each angle are converted into three-dimensional space rays by calculating the inverse matrix of the external parameter matrix to obtain the three-dimensional rays of the same point of the target illumination part at each angle.

[0031] The least square method is used to solve the point closest to the three-dimensional ray of the target illumination part at each angle and obtain its coordinates to obtain the three-dimensional coordinates of the target illumination part.

[0032] Furthermore, the method of obtaining a lighting path through a lighting path iteration model according to the target lighting position data of the plant, retrieving a lighting strategy database according to the plant growth stage to obtain lighting strategy parameters, and performing directional lighting on the target lighting position of the plant according to the lighting strategy parameters and the lighting path includes:

[0033] The target illumination part position data of the plant are clustered by a clustering algorithm to obtain a target illumination part group, and the center point coordinates of each target illumination part group are obtained as the illumination path target point coordinates.

[0034] According to the coordinates of the target points of the lighting path, the shortest path of the light source passing through each lighting path target point is obtained through the ant colony algorithm as the lighting path.

[0035] According to the plant growth stage, the lighting strategy database is retrieved to obtain the spectral wavelength, light intensity, and unit lighting duration, so that the light source illumination point moves along the lighting path according to the spectral wavelength and light intensity, and stays at each lighting path target point in the lighting path according to the unit lighting duration to provide directional lighting to the target lighting parts of the plant.

[0036] Furthermore, the method of clustering the target illumination part position data of the plant by a clustering algorithm to obtain a target illumination part group, and obtaining the center point coordinates of each target illumination part group as the illumination path target point coordinates includes:

[0037] The target illumination position data of the plant is averaged to obtain the average distance of the target illumination position data points. for:

[0038] .

[0039] in is the number of three-dimensional coordinate points of the target illumination part, It is The three-dimensional coordinate points of the target illumination part, is the three-dimensional coordinate point of the jth target illumination part, is the average distance of the data points in the target illumination area.

[0040] Get the diameter of the light source irradiation point aperture, and get the initial cluster number based on the diameter of the light source irradiation point aperture and the average distance of the target illumination part data points for:

[0041] .

[0042] in is the diameter of the aperture at the point of illumination of the light source, is the initial number of clusters.

[0043] The initial clustering is performed by the K-means clustering algorithm according to the initial cluster number to obtain the initial target illumination part group.

[0044] Determine whether the maximum distance between the points in each cluster group and the center point in the initial target illumination part group is greater than the diameter of the aperture of the light source illumination point. When the maximum distance between the points in the cluster group and the center point is greater than the diameter of the aperture of the light source illumination point, increase the number of clusters according to the preset amplitude value and perform clustering again until the maximum distance between the points in all cluster groups and the center point is less than the diameter of the aperture of the light source illumination point. Get the center point coordinates of each target illumination part group when clustering stops as the target point coordinates of the illumination path.

[0045] Furthermore, the method of obtaining the shortest path of the light source passing through each illumination path target point as the illumination path by using an ant colony algorithm according to the illumination path target point coordinates includes:

[0046] Initialize the ant colony algorithm parameters, including the number of ants, pheromone concentration, and pheromone volatilization rate.

[0047] The Euclidean distance between each point is calculated according to the coordinates of the target point on the illumination path and the inverse is used as the heuristic information. for:

[0048]

[0049] in It is The coordinates of the target point of the lighting path, is the coordinate of the target point of the j-th illumination path, is the light path target point Heuristic information to the target point j of the lighting path.

[0050] The ant colony starts from a random illumination path target point and selects the next target point according to the illumination path selection probability until all illumination path target points are visited. The illumination path selection probability is:

[0051]

[0052] in is the light path target point The pheromone concentration at the target point j on the illumination path is is the light path target point To a light path destination point that has not been visited yet The pheromone concentration on is the light path target point To a light path destination point that has not been visited yet Heuristic information on is the pheromone concentration importance factor, is the heuristic information importance factor, is the light path target point The path selection probability to the illumination path target point j.

[0053] When all ants have completed a visit to all the target points of the illumination path, the complete path constructed by each ant is obtained, and the length of the complete path constructed by each ant is calculated. for:

[0054] .

[0055] in It is The complete path of the ant, It is The light path target point in the complete path of the ant and j, It is The length of the complete path constructed by the ants.

[0056] The pheromone concentration is updated according to the heuristic information and pheromone volatility rate on the complete path constructed by each ant to obtain the updated pheromone concentration. for:

[0057] .

[0058] in is the pheromone volatilization rate, is the target point of the illumination path The path segment consisting of the first Just an ant, is the target point of the illumination path The total number of ants on the path segment to the target point j of the illumination path, is the target point of the illumination path The path segment consisting of the illumination path j Heuristic information of ants, is the updated pheromone concentration.

[0059] The minimum value of the complete path length constructed by all ants is selected and recorded. After updating the pheromone concentration, the ant colony starts from the random illumination path target point again and selects the path node. The iteration is repeated until the convergence termination iteration is reached. The minimum value of the complete path length at the termination is output as the optimal path.

[0060] Furthermore, after the target illumination part of the plant is illuminated in a directional manner according to the illumination strategy parameters and the illumination path, the method further comprises:

[0061] The growth efficiency data and light energy consumption of the plant growth stages are collected and the ratio is calculated to obtain the actual light energy efficiency of each stage.

[0062] The actual light energy efficiency is compared with a preset target light energy efficiency range threshold, wherein the preset target light energy efficiency range threshold is set according to the historically measured growth efficiency data of the plant growth stage and the light energy consumption of the plant growth stage to ensure that the light energy utilization efficiency of the plant at each stage under normal growth conditions is within a numerical range. When the actual light energy efficiency is outside the preset target light energy efficiency range threshold, the light strategy parameters in the light strategy database are adjusted, including:

[0063] When the actual light energy efficiency is greater than the maximum value of the preset target light energy efficiency range threshold, the light intensity or the unit light duration is increased.

[0064] When the actual light energy efficiency is less than the minimum value of the preset target light energy efficiency range threshold, the light intensity or the unit light duration is reduced.

[0065] Based on the same inventive concept, the present invention provides an intelligent lighting optimization system for plant growth, the system comprising: a life cycle identification module, a target lighting positioning module, and a lighting path iteration module, and the modules are connected in sequence.

[0066] The life cycle recognition module is used to obtain plant images through an image acquisition device, and input the plant images into a plant life cycle recognition model to obtain the plant growth stage; the life cycle recognition model is a mapping relationship between plant images and plant growth stages established according to a convolutional neural network algorithm; the plant growth stage includes a seedling stage, a growth stage, a flowering stage, and a fruit ripening stage.

[0067] The target lighting positioning module is used to obtain the target lighting part category by comparing with a preset key part lighting table according to the plant growth stage, and obtain the target lighting part position data of the plant through a plant key part positioning model according to the plant image and the target lighting part category; the plant key part positioning model is a mapping relationship between the plant image, the target lighting part category and the target lighting part position data of the plant established according to the target detection algorithm combined with three-dimensional positioning.

[0068] The lighting path iteration module is used to obtain the lighting path through the lighting path iteration model according to the target lighting part position data of the plant, retrieve the lighting strategy database according to the plant growth stage to obtain the lighting strategy parameters, and perform directionally illuminated lighting on the target lighting part of the plant according to the lighting strategy parameters and the lighting path, wherein the lighting strategy parameters include spectral wavelength, light intensity, and unit lighting duration.

[0069] (3) Beneficial effects

[0070] Compared with the prior art, the present invention has the following beneficial effects: accurate positioning of key lighting parts of plants is achieved through target detection combined with triangulation, and intelligent lighting path optimization is further achieved through clustering and ant colony algorithms, thereby achieving directional lighting of specific parts of the plant growth stage as a whole to promote efficient plant growth and reduce light energy waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the embodiments of the present application 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 some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0072] Figure 1 This is a flowchart of a method for intelligent optimization of lighting for plant growth according to Embodiment 1 of the present invention;

[0073] Figure 2 This is a schematic diagram of the module composition of an intelligent lighting optimization system for plant growth according to Example 2 of the present invention. DETAILED DESCRIPTION

[0074] 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.

[0075] Example 1: Figure 1 As shown, this embodiment provides a method for intelligent optimization of lighting for plant growth, the method comprising:

[0076] S1. Acquire plant images through image acquisition equipment, and input the plant images into a plant life cycle recognition model to obtain plant growth stages; the life cycle recognition model is a mapping relationship between plant images and plant growth stages established according to a convolutional neural network algorithm; the plant growth stages include seedling stage, growth stage, flowering stage, and fruit maturity stage; the life cycle recognition model training process is to obtain a group of plant images from four angles, namely, front, back, left, and right, corresponding to each growth stage of the plant as the first training image, extract growth stage features from the first training image through a convolutional neural network, perform feature fusion to generate a global feature image corresponding to each stage as the second training image, annotate the first training image, the second training image and the corresponding growth stage to perform model training to obtain the plant life cycle recognition model, and the growth stage features include plant trunk height, leaves, flowers, and fruits.

[0077] S2. According to the plant growth stage, the target illumination part category is obtained by comparing with the preset key part illumination table, and the target illumination part position data of the plant is obtained through the plant key part positioning model according to the plant image and the target illumination part category; the plant key part positioning model is a mapping relationship between the plant image, the target illumination part category and the target illumination part position data of the plant established according to the target detection algorithm combined with three-dimensional positioning; the preset key part illumination table is a mapping relationship between each growth stage of the plant and the key part of the plant that needs to obtain illumination, i.e., the target illumination part category, which is set according to historical experience. For example, in the growth stage, the leaves need to be irradiated with enhanced blue light to promote photosynthesis. In this stage, the target illumination part category is the leaves, while in the fruit ripening stage, the target illumination part category is the fruit, i.e., the fruit part needs to be illuminated to enhance sweetness and pigment deposition. At this time, the leaves do not need too much light to reduce energy consumption.

[0078] S3. Obtain the lighting path through the lighting path iteration model according to the target lighting part position data of the plant, retrieve the lighting strategy database according to the plant growth stage to obtain the lighting strategy parameters, and carry out directional lighting on the target lighting part of the plant according to the lighting strategy parameters and the lighting path. The lighting strategy parameters include spectral wavelength, light intensity, and unit lighting duration. The lighting strategy database is set according to historical experience so that the plants can meet the lighting conditions required for normal growth at each stage; the unit lighting duration is the lighting duration of the light source to a single target lighting part position in the lighting path.

[0079] It should be noted that the method of obtaining the target illumination part position data of the plant through the key part positioning model of the plant according to the plant image and the target illumination part category includes:

[0080] S21. Acquire multi-angle plant images at preset shooting positions, perform plant key part feature recognition on the plant images through the YOLO target detection algorithm to obtain the plant key part category and position boundary box, filter the plant key part category according to the target illumination part category, and perform local image extraction according to the position boundary box to obtain the target illumination part image; the preset shooting positions are four angles of the plant, namely the front, back, left and right, and the plant key parts include the trunk, leaves, flowers and fruits. For example, perform plant key part feature recognition on the plant images through the YOLO algorithm to obtain the trunk, leaves, flowers and fruit data. In the growth stage of the plant, if the target illumination part category is leaves, the image with the category of leaves is filtered out from the YOLO recognition result and local image extraction is performed according to the boundary box as the target illumination part image of this stage.

[0081] S22. Extract local feature points of the target illumination part through a feature point detection algorithm according to the target illumination part image to obtain a target illumination part descriptor, match the target illumination part descriptors to obtain the same points of the target illumination part descriptors, and perform same-point calibration to obtain the same-point calibration data of the target illumination part.

[0082] S23. According to the same-point calibration data of the target illumination part, the two-dimensional point data is converted into three-dimensional point data by triangulation to obtain the three-dimensional coordinates of the target illumination part, and all the three-dimensional coordinates of the target illumination part are summarized as the target illumination part position data of the plant.

[0083] It should be noted that the method of extracting local feature points of the target illumination part by a feature point detection algorithm according to the target illumination part image to obtain a target illumination part descriptor, matching the target illumination part descriptors to obtain the same points of the target illumination part descriptors, and performing same-point calibration to obtain the same-point calibration data of the target illumination part includes:

[0084] S221. Extract local feature points of the target illumination part through the ORB feature point detection algorithm to obtain a target illumination part descriptor; the ORB algorithm generates binary coded descriptors, each descriptor encoding a specific feature point of the target part. For the four angle images, four descriptor sets can be obtained by extracting local feature points of the target illumination part through the ORB feature point detection algorithm.

[0085] S222, calculate the Hamming distance between the target illumination part descriptors in the target illumination part image at each angle as:

[0086] .

[0087] Among them is It is The first image in the angle target illumination area object illumination part descriptor, is the first object illumination part descriptor, is the target illumination part descriptor and the target illumination part descriptor The Hamming distance between is the number of bits of the descriptor binary vector, is the target illumination part descriptor No. Bit, is the target illumination part descriptor No. Bit, is an exponential function, which means that if The result is 1 if they are not equal, otherwise 0.

[0088] S223, for The minimum Hamming distance between each target illumination part descriptor in the j-th angle target illumination part image and each target illumination part descriptor in the j-th angle target illumination part image is used to obtain the first Each target illumination part descriptor in the angle target illumination part image is the nearest neighbor descriptor in the jth angle target illumination part image for:

[0089] .

[0090] in is the index that minimizes the Hamming distance, It is Target Illumination Part Descriptor in Angular Target Illumination Part Image The nearest neighbor descriptor in the image of the object illuminated at the jth angle.

[0091] S224, according to For each target illumination part descriptor in the j-th angle target illumination part image, the second smallest Hamming distance between it and each target illumination part descriptor in the j-th angle target illumination part image is obtained. Each target illumination part descriptor in the angle target illumination part image is the second nearest neighbor descriptor in the j-th angle target illumination part image. The matching reliability ratio obtained based on the nearest neighbor descriptor and the second nearest neighbor descriptor is:

[0092] .

[0093] in It is Target Illumination Part Descriptor in Angular Target Illumination Part Image The second nearest neighbor descriptor in the image of the target illumination part at the jth angle, is the matching reliability ratio.

[0094] S225. Compare the matching reliability ratio with the preset reliability filtering threshold. When the matching reliability ratio is less than the preset reliability filtering threshold, the nearest neighbor descriptor is used as the same point for calibration, and all calibration points are summarized to obtain the calibration data of the same point of the target illumination part. The preset reliability filtering threshold is a value that can filter out most mismatched points based on experience. When the value is set to [0.6, 0.75], most mismatched points can be effectively filtered out while retaining high-quality matching points. That is, when the distance of the nearest neighbor is significantly less than the distance of the second nearest neighbor, the match is considered reliable. By comparing the matching reliability ratio with the preset reliability filtering threshold, all matching target illumination part descriptors are filtered to remove points that may cause mismatches and retain reliable matching points.

[0095] It should be noted that the method of converting two-dimensional point data into three-dimensional point data by triangulation according to the same-point calibration data of the target illumination part to obtain the three-dimensional coordinates of the target illumination part includes:

[0096] S231. Obtain the intrinsic parameter matrix and extrinsic parameter matrix of the image acquisition device through camera calibration. The camera calibration is a commonly used technology in the industry, which is used to obtain the intrinsic parameter matrix and extrinsic parameter matrix of the image acquisition device. The intrinsic parameter matrix uses a 3×3 matrix to represent the focal length and optical center position of the camera, reflecting the internal optical properties of the image acquisition device. The extrinsic parameter matrix represents the position and direction of the camera in the world coordinate system, reflecting the spatial relationship of the image acquisition device. The extrinsic parameter matrix is ​​composed of a rotation matrix and translation vectors To indicate that yes The rotation matrix of the orientation camera, yes The camera's translation vector, the rotation matrix represents the camera's direction relative to the world coordinate system, and the translation vector represents the camera's position relative to the world coordinate system. For example: when the horizontal focal length and vertical focal length of the image acquisition device are Pixel, optical center position , then its internal parameter matrix can be expressed as , the forward image acquisition device is taken as the coordinate origin, and the image acquisition devices in the other three directions are translated 2 meters backward, left, and right relative to the forward device, respectively. Then the forward external parameter matrix Represented as the forward rotation matrix , that is, the coordinate system of the forward camera is aligned with the coordinate axis of the world coordinate system without rotation, and the position of the camera is located at the origin of the world coordinate system; the backward external parameter matrix Represented as the backward rotation matrix , that is, the backward camera direction is rotated 180 degrees relative to the world coordinate system and translated 2 meters backward; the left external parameter matrix Represented as a left rotation matrix , , that is, the left camera direction is rotated 90 degrees relative to the world coordinate system and translated 2 meters to the left; the right external parameter matrix Represented as a right rotation matrix , , that is, the right camera direction is rotated 90 degrees relative to the world coordinate system and translated 2 meters to the right.

[0097] S232, normalizing the two-dimensional coordinates of the same-point calibration data of the target illumination part at each angle by calculating the inverse matrix of the internal parameter matrix to obtain the two-dimensional normalized coordinates of the target illumination part.

[0098] For example, the two-dimensional coordinates of the same point of the calibration data of a certain target illumination part in the four directions of view are the forward coordinates: , backward coordinates: , left coordinate: , right coordinate: , then the two-dimensional normalized coordinates of the forward target illumination part are , and similarly, the two-dimensional normalized coordinates of the backward target illumination part can be obtained: , the two-dimensional normalized coordinates of the left target illumination position: , the two-dimensional normalized coordinates of the right-facing target illumination position: .

[0099] S233. Convert the two-dimensional normalized coordinates of the target illumination part at each angle into three-dimensional space rays by calculating the inverse matrix of the external parameter matrix to obtain the three-dimensional rays of the same point of the target illumination part at each angle.

[0100] For example: The direction of the three-dimensional ray at the same point in the forward target illumination area is Similarly, the three-dimensional ray direction of the same point in the target illumination area can be obtained , the three-dimensional ray direction of the same point in the left target illumination area , the direction of the three-dimensional ray at the same point of the right target illumination position Therefore, taking the position of the image acquisition device in each direction as the starting point, four three-dimensional rays at the same point of the target illumination position can be obtained: forward , backward , left , right .

[0101] S234, solving the point closest to the three-dimensional ray at the same point of the target illumination part at each angle by the least square method and obtaining its coordinates to obtain the three-dimensional coordinates of the target illumination part.

[0102] For example, by solving the above equation using the least squares method, the three-dimensional coordinates of the point closest to the four three-dimensional rays at the same point of the target illumination part are obtained as follows: , the specific iterative process of the least squares method can be solved by the least_squares function of the SciPy library.

[0103] It should be noted that the method of obtaining the lighting path through the lighting path iteration model according to the target lighting part position data of the plant, retrieving the lighting strategy database according to the plant growth stage to obtain the lighting strategy parameters, and performing directional lighting on the target lighting part of the plant according to the lighting strategy parameters and the lighting path includes:

[0104] S31. Clustering the target illumination part position data of the plant through a clustering algorithm to obtain a target illumination part group, and obtaining the center point coordinates of each target illumination part group as the illumination path target point coordinates.

[0105] S32. According to the coordinates of the illumination path target points, the shortest path of the light source passing through each illumination path target point is obtained by using an ant colony algorithm as the illumination path.

[0106] S33. According to the plant growth stage, the illumination strategy database is retrieved to obtain the spectral wavelength, illumination intensity, and unit illumination duration, so that the light source illumination point moves along the illumination path according to the spectral wavelength and illumination intensity, and stays at each illumination path target point in the illumination path according to the unit illumination duration to perform directional illumination on the target illumination part of the plant. For example: in the seedling stage, the leaf part moves along the illumination path according to the spectral wavelength of 450-470nm blue light with an illumination intensity of 200µmol / m² / s and stays at each illumination path target point in the illumination path for 14 hours to perform directional illumination on the target illumination part of the plant; in the fruit stage, the fruit part moves along the illumination path according to the spectral wavelength of 600-700nm red light with an illumination intensity of 400µmol / m² / s and stays at each illumination path target point in the illumination path for 8 hours to perform directional illumination on the target illumination part of the plant.

[0107] It should be noted that the method of clustering the target illumination part position data of the plant by a clustering algorithm to obtain a target illumination part group, and obtaining the center point coordinates of each target illumination part group as the illumination path target point coordinates includes:

[0108] S311, averaging the target illumination position data of the plant to obtain the average distance of the target illumination position data points for:

[0109] .

[0110] in is the number of three-dimensional coordinate points of the target illumination part, It is The three-dimensional coordinate points of the target illumination part, is the three-dimensional coordinate point of the jth target illumination part, is the average distance of the data points in the target illumination area.

[0111] S312, obtaining the diameter of the aperture of the light source illumination point, and obtaining the initial clustering number according to the diameter of the aperture of the light source illumination point and the average distance of the data points of the target illumination part for:

[0112] .

[0113] in is the diameter of the aperture at the point of illumination of the light source, is the initial number of clusters.

[0114] Since the size of the light source illumination point is fixed, more cluster groups are needed if the data points of the target illumination part are scattered, and fewer cluster groups are needed if the distribution is dense. The above method can quickly set a reasonable initial cluster number according to the overall distribution of the target points; for example, the diameter of the aperture of the light source illumination point is 0.2 meters, the average distance of the data points of the target illumination part is 0.1 meters, and the number of three-dimensional coordinate points of the target illumination part is 100, then the initial cluster number is If the average distance between the data points of the target illumination part is 0.02 meters, then the initial cluster number is indivual.

[0115] S313, performing initial clustering by using the K-means clustering algorithm according to the initial cluster number to obtain an initial target illumination part group; the K-means clustering goal is to minimize the square error between the points in each group and the group center, and the K-means clustering can be performed by the SciKit-learn library to obtain the coordinates of the center point after clustering.

[0116] S314, determine whether the maximum distance between each point in the cluster group and the center point in the initial target illumination part group is greater than the diameter of the light source illumination point aperture. When the maximum distance between the point in the cluster group and the center point is greater than the diameter of the light source illumination point aperture, increase the number of clusters according to the preset amplitude value and perform clustering again until the maximum distance between all points in the cluster group and the center point is less than the diameter of the light source illumination point aperture. The center point coordinates of each target illumination part group when clustering stops are obtained as the illumination path target point coordinates. The preset amplitude value is set according to historical experience. In this embodiment, the preset amplitude value is 10% of the initial cluster number, that is, the initial cluster number is gradually increased by 10% as the updated cluster number. For example, when When , ,when When , ,By dynamically adjusting the number of clusters to match the coverage of the aperture, efficient utilization of light is ensured.

[0117] It should be noted that the method of obtaining the shortest path of the light source passing through each lighting path target point as the lighting path by using an ant colony algorithm according to the lighting path target point coordinates includes:

[0118] S321, initialize the parameters of the ant colony algorithm, including the number of ants, pheromone concentration, and pheromone volatilization rate; the initialized number of ants, pheromone volatilization rate, and pheromone concentration are set according to historical experience. For example, in this embodiment, the number of ants is set to 20, the initialization value of the pheromone concentration is usually set to 0.1, and the pheromone volatilization rate is set to 0.5.

[0119] S322, calculate the Euclidean distance between each point according to the coordinates of the target point on the illumination path and take the inverse as the heuristic information for:

[0120] .

[0121] in It is The coordinates of the target point of the lighting path, is the coordinate of the jth lighting path target point, is the light path target point Heuristic information about the lighting path to the target point j.

[0122] S323, the ant colony starts from the random illumination path target point and selects the next target point according to the illumination path selection probability until all illumination path target points are visited, and the illumination path selection probability is:

[0123] .

[0124] in is the light path target point The pheromone concentration at the target point j on the illumination path is is the light path target point To a light path destination point that has not been visited yet The pheromone concentration on is the light path target point To a light path destination point that has not been visited yet Heuristic information on is the pheromone concentration importance factor, is the heuristic information importance factor, is the light path target point The path selection probability to the illumination path target point j; and The influence of pheromone concentration and heuristic information on path selection is controlled respectively. In this embodiment, , Controls the ants' reliance on pheromone concentration during path selection. It will increase the diversity of path exploration, but may increase the convergence time of the algorithm. Controls the ants' reliance on heuristic information during path selection. It will speed up the convergence of the algorithm, but it may also lead to convergence to a local optimal solution.

[0125] S324: When all ants have completed a visit to all illumination path target points, the complete path constructed by each ant is obtained, and the length of the complete path constructed by each ant is calculated. for:

[0126] .

[0127] in It is The complete path of the ant, It is The light path target point in the complete path of the ant and j, It is The length of the complete path constructed by the ants.

[0128] S325, updating the pheromone concentration according to the heuristic information and pheromone volatilization rate on the complete path constructed by each ant to obtain an updated pheromone concentration for:

[0129] .

[0130] in is the pheromone volatilization rate, is the target point of the illumination path The path segment consisting of the first Just an ant, is the target point of the illumination path The total number of ants on the path segment to the target point j of the illumination path, is the target point of the illumination path The path segment consisting of the first Heuristic information of ants, is the updated pheromone concentration.

[0131] S326, select the minimum value of the complete path length constructed by all ants and record it, and after updating the pheromone concentration, start the ant colony from the random light path target point again and select the path node, repeat the iteration until the convergence termination iteration is reached, and output the minimum value of the complete path length at the termination as the best path. After multiple iterations, the pheromone concentration on the short path will continue to increase and more ants will pass through it. In the end, most ants will choose the path with high pheromone concentration, that is, the path with short access distance. When the number of visits of ants on the path gradually concentrates on the short path, the algorithm gradually converges, and gradually converges to a shortest path.

[0132] It should be noted that after the target illumination part of the plant is illuminated in a directional manner according to the illumination strategy parameters and the illumination path, the method further includes:

[0133] S301, collect the growth efficiency data of the plant growth stage and the light energy consumption of the plant growth stage and calculate the ratio to obtain the actual light energy efficiency of each stage; the growth efficiency data of the plant generation stage includes the plant height increase in the seedling stage, the chlorophyll content increase in the growth stage and the flowering stage, and the fruit size increase in the fruit maturity stage measured according to the preset measurement cycle, and the light energy consumption of the plant growth stage is obtained by querying the power consumption of the equipment. For example: the actual light energy efficiency of the seedling stage is: , the actual light energy efficiency during the growth and flowering stages is: , the actual light energy efficiency at the fruit ripening stage is: .

[0134] S302, comparing the actual light energy efficiency with a preset target light energy efficiency range threshold, wherein the preset target light energy efficiency range threshold is set based on historically measured growth efficiency data of plant growth stages and light energy consumption of plant growth stages to ensure that the light energy utilization efficiency of plants at each stage under normal growth conditions is within a numerical range, and when the actual light energy efficiency is outside the preset target light energy efficiency range threshold, adjusting the light strategy parameters in the light strategy database, including:

[0135] S303. When the actual light energy efficiency is greater than the maximum value of the preset target light energy efficiency range threshold, increase the light intensity or unit light duration; when the actual light energy efficiency is greater than the maximum value of the preset target light energy efficiency range threshold, the light energy efficiency is high, indicating that the plant maintains efficient growth. At this time, increasing the light intensity or extending the light duration is conducive to prompting the plant to absorb more light energy and accelerate plant growth.

[0136] S304. When the actual light energy efficiency is less than the minimum value of the preset target light energy efficiency range threshold, reduce the light intensity or the unit light duration. When the actual light energy efficiency is less than the minimum value of the preset target light energy efficiency range threshold, the light efficiency is low, indicating that the plant growth has reached a saturation state. Excessive light does not promote plant growth but leads to energy waste. Therefore, the light intensity should be appropriately reduced or the light duration should be shortened to avoid excessive energy consumption.

[0137] Due to individual differences among plants and the different growth responses they may exhibit under different environments, the lighting strategy database is adjusted by analyzing the actual light energy efficiency at each stage, thereby achieving dynamic optimization of the lighting strategy.

[0138] Embodiment 2: Based on the same inventive concept, Figure 2 As shown, this embodiment provides a lighting intelligent optimization system for plant growth, the system comprising: a life cycle identification module, a target lighting positioning module, and a lighting path iteration module, and the modules are connected in sequence.

[0139] The life cycle recognition module is used to obtain plant images through an image acquisition device, and input the plant images into a plant life cycle recognition model to obtain the plant growth stage; the life cycle recognition model is a mapping relationship between plant images and plant growth stages established according to a convolutional neural network algorithm; the plant growth stage includes a seedling stage, a growth stage, a flowering stage, and a fruit ripening stage.

[0140] The target lighting positioning module is used to obtain the target lighting part category by comparing with a preset key part lighting table according to the plant growth stage, and obtain the target lighting part position data of the plant through a plant key part positioning model according to the plant image and the target lighting part category; the plant key part positioning model is a mapping relationship between the plant image, the target lighting part category and the target lighting part position data of the plant established according to the target detection algorithm combined with three-dimensional positioning.

[0141] The lighting path iteration module is used to obtain the lighting path through the lighting path iteration model according to the target lighting part position data of the plant, retrieve the lighting strategy database according to the plant growth stage to obtain the lighting strategy parameters, and perform directionally illuminated lighting on the target lighting part of the plant according to the lighting strategy parameters and the lighting path, wherein the lighting strategy parameters include spectral wavelength, light intensity, and unit lighting duration.

[0142] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in Embodiment 1 of the method, and will not be elaborated here.

[0143] Finally, it should be noted that: Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for intelligent optimization of lighting for plant growth, characterized in that: The method comprises: Acquire plant images through an image acquisition device, and input the plant images into a plant life cycle recognition model to obtain plant growth stages; the life cycle recognition model is a mapping relationship between plant images and plant growth stages established according to a convolutional neural network algorithm; the plant growth stages include a seedling stage, a growth stage, a flowering stage, and a fruit ripening stage; According to the plant growth stage, the target illumination part category is obtained by comparing with the preset key part illumination table, and the target illumination part position data of the plant is obtained by the plant key part positioning model according to the plant image and the target illumination part category; the plant key part positioning model is a mapping relationship between the plant image, the target illumination part category and the target illumination part position data of the plant established according to the target detection algorithm combined with three-dimensional positioning; According to the target illumination part position data of the plant, a illumination path is obtained through an illumination path iteration model, and according to the growth stage of the plant, an illumination strategy database is retrieved to obtain illumination strategy parameters, and according to the illumination strategy parameters and the illumination path, a directional illumination is performed on the target illumination part of the plant, wherein the illumination strategy parameters include spectral wavelength, illumination intensity, and unit illumination duration; The method for obtaining the target illumination part position data of the plant through the key part positioning model of the plant according to the plant image and the target illumination part category comprises: Acquire multi-angle plant images at a preset shooting position, use the YOLO target detection algorithm to identify the key parts of the plant images to obtain the key parts category and position bounding box of the plant, filter the key parts category of the plant according to the target illumination part category, and extract the local image according to the position bounding box to obtain the target illumination part image; According to the target illumination part image, local feature points of the target illumination part are extracted by a feature point detection algorithm to obtain a target illumination part descriptor, the target illumination part descriptors are matched to obtain the same points of the target illumination part descriptors, and the same point calibration is performed to obtain the same point calibration data of the target illumination part; According to the same-point calibration data of the target illumination part, the two-dimensional point data is converted into three-dimensional point data through triangulation to obtain the three-dimensional coordinates of the target illumination part, and the three-dimensional coordinates of all target illumination parts are summarized as the target illumination part position data of the plant.

2. The method for intelligent optimization of lighting for plant growth according to claim 1, characterized in that: The method of extracting local feature points of the target illumination part by a feature point detection algorithm according to the target illumination part image to obtain a target illumination part descriptor, matching the target illumination part descriptors to obtain the same points of the target illumination part descriptors, and performing same point calibration to obtain the same point calibration data of the target illumination part comprises: The local feature points of the target illumination part are extracted by the ORB feature point detection algorithm to obtain the target illumination part descriptor; The Hamming distance between the target illumination part descriptors in the target illumination part image at each angle is calculated as: ； Among them is It is The first image in the angle target illumination area object illumination part descriptor, It is The first image in the angle target illumination area object illumination part descriptor, is the target illumination part descriptor and the target illumination part descriptor The Hamming distance between is the number of bits of the descriptor binary vector, is the target illumination part descriptor No. Bit, is the target illumination part descriptor No. Bit, is an exponential function, which means that if If they are not equal, the result is 1, otherwise it is 0; For Each target illumination part descriptor in the angle target illumination part image is divided into The minimum Hamming distance between each target illumination part descriptor in the angle target illumination part image is obtained. Each object illumination part descriptor in the angle object illumination part image is Nearest Neighbor Descriptors in Angular Target Illumination Part Images for: ； in is the index that minimizes the Hamming distance, It is Target Illumination Part Descriptor in Angular Target Illumination Part Image In the Nearest neighbor descriptors in angular target illumination part images; According to the Each target illumination part descriptor in the angle target illumination part image is divided into The Hamming distance between each target illumination part descriptor in the angle target illumination part image is taken as the second smallest value to obtain the Each object illumination part descriptor in the angle object illumination part image is The second nearest neighbor descriptor in the angle target illumination part image, the matching reliability ratio obtained based on the nearest neighbor descriptor and the second nearest neighbor descriptor is: ； in It is Target Illumination Part Descriptor in Angular Target Illumination Part Image In the The second nearest neighbor descriptor in the angular target illumination part image, is the matching reliability ratio; The matching reliability ratio is compared with the preset reliability filtering threshold. When the matching reliability ratio is less than the preset reliability filtering threshold, the nearest neighbor descriptor is used as its same point for calibration, and all calibration points are summarized to obtain the same-point calibration data of the target illumination part.

3. The method for intelligent optimization of lighting for plant growth according to claim 1, characterized in that: The method of converting two-dimensional point data into three-dimensional point data by triangulation according to the same-point calibration data of the target illumination part to obtain the three-dimensional coordinates of the target illumination part comprises: Obtain the intrinsic parameter matrix and extrinsic parameter matrix of the image acquisition device through camera calibration; The two-dimensional coordinates of the calibration data of the same point of the target illumination part at each angle are normalized by calculating the inverse matrix of the internal parameter matrix to obtain the two-dimensional normalized coordinates of the target illumination part; The two-dimensional normalized coordinates of the target illumination part at each angle are converted into three-dimensional space rays by calculating the inverse matrix of the external parameter matrix to obtain the three-dimensional rays of the same point of the target illumination part at each angle; The least square method is used to solve the point closest to the three-dimensional ray of the target illumination part at each angle and obtain its coordinates to obtain the three-dimensional coordinates of the target illumination part.

4. The method for intelligent optimization of lighting for plant growth according to claim 1, characterized in that: The method of obtaining a lighting path through a lighting path iteration model according to the target lighting position data of the plant, retrieving a lighting strategy database according to the plant growth stage to obtain lighting strategy parameters, and performing directional lighting on the target lighting position of the plant according to the lighting strategy parameters and the lighting path includes: The target illumination position data of the plant are clustered by a clustering algorithm to obtain a target illumination position group, and the center point coordinates of each target illumination position group are obtained as the illumination path target point coordinates; According to the coordinates of the target points of the lighting path, the shortest path of the light source passing through the target points of each lighting path is obtained by using the ant colony algorithm as the lighting path; According to the plant growth stage, the lighting strategy database is retrieved to obtain the spectral wavelength, light intensity, and unit lighting duration, so that the light source illumination point moves along the lighting path according to the spectral wavelength and light intensity, and stays at each lighting path target point in the lighting path according to the unit lighting duration to provide directional lighting to the target lighting parts of the plant.

5. The method for intelligent optimization of lighting for plant growth according to claim 4, characterized in that: The method of clustering the target illumination position data of the plant by a clustering algorithm to obtain a target illumination position group, and obtaining the center point coordinates of each target illumination position group as the illumination path target point coordinates includes: The target illumination position data of the plant is averaged to obtain the average distance of the target illumination position data points. for: ； in is the number of three-dimensional coordinate points of the target illumination part, It is The three-dimensional coordinate points of the target illumination part, It is The three-dimensional coordinate points of the target illumination part, is the average distance of the data points at the target illumination part; Get the diameter of the light source irradiation point aperture, and get the initial cluster number based on the diameter of the light source irradiation point aperture and the average distance of the target illumination part data points for: ； in is the diameter of the aperture at the point of illumination of the light source, is the initial number of clusters; Performing initial clustering by using the K-means clustering algorithm according to the initial cluster number to obtain an initial target illumination part group; Determine whether the maximum distance between the points in each cluster group and the center point in the initial target illumination part group is greater than the diameter of the aperture of the light source illumination point. When the maximum distance between the points in the cluster group and the center point is greater than the diameter of the aperture of the light source illumination point, increase the number of clusters according to the preset amplitude value and perform clustering again until the maximum distance between the points in all cluster groups and the center point is less than the diameter of the aperture of the light source illumination point. Get the center point coordinates of each target illumination part group when clustering stops as the target point coordinates of the illumination path.

6. The method for intelligent optimization of lighting for plant growth according to claim 5, characterized in that: The method of obtaining the shortest path of the light source passing through each light path target point as the light path by using an ant colony algorithm according to the light path target point coordinates includes: Initialize the ant colony algorithm parameters, including the number of ants, pheromone concentration, and pheromone volatilization rate; The Euclidean distance between each point is calculated according to the coordinates of the target point on the illumination path and the inverse is used as the heuristic information. for: ; in It is The coordinates of the target point of the lighting path, It is The coordinates of the target point in the lighting path, is the light path target point To the light path target point Heuristic information; The ant colony starts from a random illumination path target point and selects the next target point according to the illumination path selection probability until all illumination path target points are visited. The illumination path selection probability is: ; in is the light path target point To the light path target point The pheromone concentration on is the light path target point To a light path destination point that has not been visited yet The pheromone concentration on is the light path target point To a light path destination point that has not been visited yet Heuristic information on is the pheromone concentration importance factor, is the heuristic information importance factor, is the light path target point To the light path target point The path selection probability of When all ants have completed a visit to all the target points of the illumination path, the complete path constructed by each ant is obtained, and the length of the complete path constructed by each ant is calculated. for: ; in It is The complete path of the ant, It is The light path target point in the complete path of the ant and , It is The length of the complete path constructed by the ants; The pheromone concentration is updated according to the heuristic information and pheromone volatility rate on the complete path constructed by each ant to obtain the updated pheromone concentration. for: ; in is the pheromone volatilization rate, is the target point of the illumination path To the light path target point The path segment Just an ant, is the target point of the illumination path To the light path target point The total number of ants that make up the path segment, is the target point of the illumination path To the light path target point The path segment Heuristic information of ants, is the updated pheromone concentration; The minimum value of the complete path length constructed by all ants is selected and recorded. After updating the pheromone concentration, the ant colony starts from the random illumination path target point again and selects the path node. The iteration is repeated until the convergence termination iteration is reached. The minimum value of the complete path length at the termination is output as the optimal path.

7. The method for intelligent optimization of lighting for plant growth according to claim 1, characterized in that: After the target illumination part of the plant is illuminated in a directional manner according to the illumination strategy parameters and the illumination path, the method further comprises: Collect the growth efficiency data and light energy consumption of the plant growth stage and calculate the ratio to obtain the actual light energy efficiency of each stage; The actual light energy efficiency is compared with a preset target light energy efficiency range threshold, wherein the preset target light energy efficiency range threshold is set according to the historically measured growth efficiency data of the plant growth stage and the light energy consumption of the plant growth stage to ensure that the light energy utilization efficiency of the plant at each stage under normal growth conditions is within a numerical range. When the actual light energy efficiency is outside the preset target light energy efficiency range threshold, the light strategy parameters in the light strategy database are adjusted, including: When the actual light energy efficiency is greater than the maximum value of the preset target light energy efficiency range threshold, the light intensity or unit light duration is increased; When the actual light energy efficiency is less than the minimum value of the preset target light energy efficiency range threshold, the light intensity or the unit light duration is reduced.

8. An intelligent lighting optimization system for plant growth, used to execute the method according to any one of claims 1 to 7, characterized in that: The system comprises: a life cycle identification module, a target illumination positioning module, and an illumination path iteration module, and the modules are connected in sequence; The life cycle recognition module is used to obtain plant images through an image acquisition device, and input the plant images into a plant life cycle recognition model to obtain plant growth stages; the life cycle recognition model is a mapping relationship between plant images and plant growth stages established according to a convolutional neural network algorithm; the plant growth stages include seedling stage, growth stage, flowering stage, and fruit maturity stage; The target illumination positioning module is used to obtain the target illumination part category by comparing with a preset key part illumination table according to the plant growth stage, and obtain the target illumination part position data of the plant through the plant key part positioning model according to the plant image and the target illumination part category; the plant key part positioning model is a mapping relationship between the plant image, the target illumination part category and the target illumination part position data of the plant established according to the target detection algorithm combined with three-dimensional positioning; The lighting path iteration module is used to obtain the lighting path through the lighting path iteration model according to the target lighting part position data of the plant, retrieve the lighting strategy database according to the plant growth stage to obtain the lighting strategy parameters, and perform directionally illuminated lighting on the target lighting part of the plant according to the lighting strategy parameters and the lighting path, wherein the lighting strategy parameters include spectral wavelength, light intensity, and unit lighting duration.

Images