CN-122002673-A - Illumination control method and system based on plant state

CN122002673ACN 122002673 ACN122002673 ACN 122002673ACN-122002673-A

Abstract

The invention relates to the field of plant state image recognition, in particular to a plant state-based illumination control method and a plant state-based illumination control system, wherein the method comprises the steps of periodically shooting a target plant, and obtaining images of multiple angles of the target plant in each period; and for each blade, calculating a state evaluation value based on all the optimal blade areas reserved in a preset period, judging that the blade is burnt when the state evaluation value is greater than a preset threshold value, and controlling the illumination intensity to be adjusted to safe light intensity. The invention can simultaneously capture two key signals of abnormal color and surface structure deterioration of the leaf blade possibly caused by light stress, and has higher detection sensitivity and comprehensive judgment on early and weak burn signs.

Inventors

DONG QIANG
CHEN KEMIN
HUO YINGJIE
Shu Yanping

Assignees

浙江普旷农业科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260408

Claims (8)

1. A plant state-based lighting control method, comprising: Periodically shooting the target plant, and obtaining images of a plurality of angles of the target plant in each period; For each period, extracting a blade area from each frame of image by using a trained convolutional neural network, matching and numbering the blade areas belonging to the same blade in continuous frame images, and reserving an optimal blade area corresponding to each blade; Calculating a state evaluation value based on all the optimal blade areas reserved in a preset period for each blade, judging that the blade is burnt when the state evaluation value is larger than a preset threshold value, and controlling the illumination intensity to be adjusted to be safe light intensity; The state evaluation value calculating method comprises the steps of obtaining entropy values of gray images of all optimal blade areas of a single blade, which are reserved in a preset period, according to a shooting time sequence, arranging the entropy values to form an entropy value sequence, obtaining a slope of the entropy value sequence by using a least square method, calculating an average RGB value of each optimal blade area, calculating a first average value of the average RGB values of all the optimal blade areas, calculating Euclidean distances between the average RGB value of each optimal blade area and the first average value, calculating a second average value of all the obtained Euclidean distances, normalizing the second average value, and multiplying the slope of the entropy value sequence to obtain the state evaluation value.
2. The plant state-based illumination control method according to claim 1, wherein the matching and numbering of the leaf areas belonging to the same leaf in the continuous frame images includes, for each two adjacent frames, taking any leaf area in the previous frame image as a target leaf area, taking any leaf area in the next frame image as a candidate leaf area, constructing a similarity matrix of the target leaf area and each candidate leaf area based on the number of pixel points of each leaf area and RGB values, calculating a similarity value between the target leaf area and each candidate leaf area based on the similarity matrix, taking a candidate leaf area corresponding to the largest similarity value as a matching leaf of the target leaf area, and assigning unique numbers to all leaf areas corresponding to the same leaf.
3. The illumination control method based on plant states according to claim 2, wherein the construction method of the similarity matrix of the target blade area and each candidate blade area comprises the steps of constructing a spatial neighborhood grid with a preset size by taking the candidate blade area as a center, calculating a matching index of the candidate blade area and the target blade area as an effective element value of the center position for the position of the spatial neighborhood grid center, acquiring adjacent blade areas of the candidate blade area at the position for each position except the center of the spatial neighborhood grid, calculating the matching index of all the adjacent blade areas of the adjacent blade area and the target blade area, taking the largest matching index as the effective element value of the position, and taking a preset special value as an ineffective element value of the position if the candidate blade is not provided with the adjacent blade area at the position, so that the similarity matrix is obtained.
4. A plant state-based illumination control method according to claim 3, wherein the calculation method of the matching index of the candidate blade region and the target blade region comprises calculating average RGB values of the candidate blade region and the target blade region, respectively, and calculating cosine similarity between the average RGB values of the candidate blade region and the target blade region, calculating a ratio of the number of pixels of the target blade region to the number of pixels of the candidate blade region, calculating an absolute value of a difference between the ratio and 1, calculating a difference between 1 and the absolute value, and taking a product of the difference and the cosine similarity as the matching index.
5. The method for controlling illumination based on plant status according to claim 4, wherein the calculating the similarity value between the target leaf area and the candidate leaf area comprises obtaining the effective element values and the total number of effective element values of the similarity matrix between the target leaf area and the candidate leaf area, calculating the average value of all the effective element values, normalizing the total number of effective element values, and multiplying the normalized total number by the average value of the effective element values to obtain the similarity value.
6. A plant-state-based illumination control method according to claim 1, characterized in that after matching and numbering of the leaf areas in all images in one cycle is completed, for each leaf, the corresponding leaf area with the largest number of pixels in all images is reserved as the optimal leaf area for that leaf.
7. The plant state-based illumination control method according to claim 4, wherein the calculation method of the average RGB value of the target blade area is that the RGB value of each pixel point in the target blade area is obtained, the values of the red channel, the green channel and the blue channel of all the pixel points are respectively averaged to obtain a three-dimensional vector, and the three-dimensional vector is used as the average RGB value.
8. A plant-state based lighting control system, comprising a processor and a memory storing computer program instructions that, when executed by the processor, implement the plant-state based lighting control method of any one of claims 1-7.

Description

Illumination control method and system based on plant state Technical Field The invention relates to the field of plant state image recognition. And more particularly to a plant state based lighting control method and control system. Background Artificial illumination is a key environmental factor for regulating plant growth and improving yield and quality. However, improper illumination intensity or spectral distribution, particularly too strong illumination, is highly susceptible to light inhibition or even light burn phenomena in the plant leaves. Leaf burn not only directly damages photosynthetic organs of plants and reduces the growth efficiency of the plants, but also can cause secondary diseases, and finally leads to crop yield reduction and quality reduction. The plant leaf is used as the organ with the most direct sensing and responding to the illumination, and the change of the phenotypic characteristics such as color, texture and the like is an early and visual index reflecting the physiological state and the stress degree. With the development of computer vision and image processing technology, monitoring plant growth status using image analysis has become a research hotspot. In the prior art, there are methods for evaluating plant health or identifying plant diseases and insect pests by capturing plant images and extracting features such as color, texture, morphology and the like. However, in the actual shooting process, due to the reasons of unstable flying attitude of the unmanned aerial vehicle or overlapping shielding among the blades, the existing method cannot stably and accurately track and match the image areas of the same blade at different time and different visual angles, so that the existing method focuses on static feature analysis, lacks analysis on the evolution trend of the state of the blade along with time, and is difficult to find early burn of the blade caused by illumination, thereby leading to insensitive illumination control. Disclosure of Invention In order to solve the technical problem that the prior method is difficult to find early burn of the blade caused by illumination, and the illumination control is insensitive, the invention provides a scheme in the following aspects. In a first aspect, a plant state based lighting control method includes: Periodically shooting the target plant, and obtaining images of a plurality of angles of the target plant in each period; For each period, extracting a blade area from each frame of image by using a trained convolutional neural network, matching and numbering the blade areas belonging to the same blade in continuous frame images, and reserving an optimal blade area corresponding to each blade; Calculating a state evaluation value based on all the optimal blade areas reserved in a preset period for each blade, judging that the blade is burnt when the state evaluation value is larger than a preset threshold value, and controlling the illumination intensity to be adjusted to be safe light intensity; The state evaluation value calculating method comprises the steps of obtaining entropy values of gray images of all optimal blade areas of a single blade, which are reserved in a preset period, according to a shooting time sequence, arranging the entropy values to form an entropy value sequence, obtaining a slope of the entropy value sequence by using a least square method, calculating an average RGB value of each optimal blade area, calculating a first average value of the average RGB values of all the optimal blade areas, calculating Euclidean distances between the average RGB value of each optimal blade area and the first average value, calculating a second average value of all the obtained Euclidean distances, normalizing the second average value, and multiplying the slope of the entropy value sequence to obtain the state evaluation value. Preferably, matching and numbering blade areas belonging to the same blade in continuous frame images comprises the steps of regarding any blade area in a previous frame image as a target blade area, regarding any blade area in a subsequent frame image as a candidate blade area, constructing a similarity matrix of the target blade area and each candidate blade area based on the number of pixel points of each blade area and RGB values, calculating a similarity value between the target blade area and each candidate blade area based on the similarity matrix, regarding the candidate blade area corresponding to the largest similarity value as a matching blade of the target blade area, and assigning unique numbers to all blade areas corresponding to the same blade. The construction method of the similarity matrix of the target blade area and each candidate blade area comprises the steps of constructing a space neighborhood grid with a preset size by taking the candidate blade area as the center, calculating a matching index of the candidate blade area and the target blade area as an effective element