A spectral detection method for crop plant segmentation in a plant factory

Abstract

The present application relates to the field of spectral detection technology, and specifically to a spectral detection method for segmenting crop plants in a plant factory. The method includes: collecting images before and after spraying; and obtaining before and after difference images and adjacent difference images based on image differences; obtaining the possibility that the plant area is a plant overlapping area through the difference between each connected domain in the before and after difference images and its minimum circumscribed circle area, and screening the plant overlapping area; obtaining the possibility that two absorption points are in the same area according to the pixel value difference and distance of the absorption points in all adjacent difference images; and determining the target class of each absorption point, and determining the absorption point corresponding to each plant according to the target class; dividing the plant overlapping area according to the absorption point of each plant to obtain the area of ​​each plant. The present application can divide the complex overlapping situation of plants more accurately.

Description

Technical Field

[0001] The present application relates to the field of spectral detection technology, and in particular to a spectral detection method for crop plant segmentation in a plant factory. Background Art

[0002] In plant factories, cameras are used to take images of crops to analyze how the crops are growing. However, because the planting density of several plants of the same variety of crops in plant factories is high and the distance between plants is small, in the process of photographing and evaluating the plants, plants that are close to each other often block each other, making it difficult to distinguish the two in the resulting image, resulting in uneven spraying of nutrient solution on the overlapping parts of overlapping plants, which leads to uneven growth of crops. The existing technology's solution to the above problem often divides the image into regions through a threshold segmentation algorithm, and divides the parts that do not belong to the same plant according to the overlapping boundaries between different plants and the different grayscale expression characteristics on both sides of the boundaries. However, some plants may have similar appearances in mutually blocked areas (such as leaves of different plants overlapping each other), and such a method is difficult to ensure segmentation accuracy. Summary of the invention

[0003] In order to solve the technical problem of poor segmentation accuracy, the present application provides a spectral detection method for crop plant segmentation in a plant factory. The technical solution adopted is as follows:

[0004] The present application proposes a spectral detection method for crop plant segmentation in a plant factory, the method comprising the following steps:

[0005] Collect images before spraying and several images after spraying, and the pixel value of each pixel in the image is the nitrogen content;

[0006] Subtracting the first image after spraying from the image before spraying to obtain a before-after difference image; segmenting the before-after difference image to obtain a plant region; obtaining a minimum circumscribed circle of the plant region, obtaining the possibility that the plant region is a plant overlapping region according to the difference between the area of ​​the plant region and its minimum circumscribed circle, the area of ​​the plant region and the average value of the nitrogen content of the plant region, and screening the plant overlapping region according to the possibility;

[0007] For adjacent images after spraying, two-by-two differences are made to obtain adjacent difference images; absorption points are selected according to the pixel values ​​of the pixels in the adjacent difference images; the possibility of the two absorption points being in the same region is obtained according to the difference in the pixel values ​​of the two absorption points in all adjacent difference images and the distance between the absorption points; the target class of each absorption point is determined according to the possibility of each absorption point being in the same region with all other absorption points; the target classes are merged according to the intersection and union between the target classes to obtain the absorption points corresponding to each plant;

[0008] The overlapping area of ​​the plants is divided by performing regional growth according to the absorption points of each plant in the overlapping area of ​​the plants, and the area corresponding to each plant is obtained.

[0009] In the above scheme, the spectral detection method is combined to carry out spectral detection on the plants, and the spectral detection results are combined with the images. The parts belonging to different plants are accurately divided through the difference characteristics of the absorption of various nutrients in each area in the local overlapping part over time. The complex overlapping situation of plants can be divided more accurately, and the areas belonging to each plant can be divided from each other. The division results are more accurate, which is convenient for subsequent judgment of the growth conditions of each plant.

[0010] In one embodiment, the method of obtaining the before-after difference image by subtracting the first image after spraying from the image before spraying is:

[0011] The pixel values ​​of each pixel point at the same position in the first image after spraying and the image before spraying are subtracted to obtain a difference image, and the difference image is recorded as the before-after difference image.

[0012] In one embodiment, the method of segmenting the front and back difference images to obtain the plant area is:

[0013] Threshold segmentation is performed on the pixel values ​​of all pixels in the before and after difference images, and the pixels are divided into two categories. One category of pixels is marked as 1, and the other category of pixels is marked as 0 for connected domain analysis. Each connected domain is regarded as a plant area.

[0014] In one embodiment, the method for obtaining the possibility that the plant region is a plant overlapping region according to the difference between the area of ​​the plant region and its minimum circumscribed circle, the area of ​​the plant region and the average value of the nitrogen content of the plant region is:

[0015] Calculate the mean value of nitrogen content of all pixels in the plant area as the mean value of nitrogen content in the plant area;

[0016] The possibility of overlapping plant regions was positively correlated with the area of ​​the plant region and the mean nitrogen content of the plant region, and negatively correlated with the area difference between the plant region and the minimum circumscribed circle.

[0017] In one embodiment, the method of performing pairwise subtraction of adjacent post-spraying images to obtain adjacent difference images is:

[0018] For all the images after spraying, the pixel values ​​of each pixel at the same position of two adjacent images after spraying are subtracted, and the absolute value of the difference is used as the pixel value of the pixel. A difference image is obtained for every two adjacent images after spraying, which is recorded as the adjacent difference image.

[0019] In one embodiment, the method for selecting absorption points according to the pixel values ​​of the pixels in the adjacent difference images is:

[0020] For each adjacent difference image, a maximum point is selected in the overlapping area of ​​the plant; if the pixel value of a pixel is greater than the pixel values ​​of all pixels in the 8-neighborhood, then the pixel is a maximum point; all the maximum points in the adjacent difference images are obtained and the maximum points are used as absorption points in the overlapping area of ​​the plant.

[0021] In one embodiment, the method for obtaining the possibility of the two absorption points being in the same region according to the difference in pixel values ​​of the two absorption points in all adjacent difference images and the distance between the absorption points is:

[0022] The probability of two absorption points being in the same region is positively correlated with the difference in pixel values ​​of the two absorption points and negatively correlated with the distance between the absorption points.

[0023] In one embodiment, the method for determining the target class of each absorption point according to the possibility of each absorption point being in the same region as all other absorption points is as follows:

[0024] Take each absorption point as the target point, calculate the probability of the target point and each absorption point being in the same region, if the probability of the same region is greater than the preset threshold of the same region, then the two points are considered to be the same plant, if the probability of the same region is less than the threshold of the same region, then the two points are considered not to be the same plant;

[0025] All absorption points of the same plant as the target point are taken as the target class of the target point.

[0026] In one embodiment, the method of merging the target classes according to the intersection and union between the target classes to obtain the absorption point corresponding to each plant is:

[0027] Get the target class of all absorption points; for each target class, count all the absorption points in it, calculate the intersection and union of the two target classes, and if the intersection and union ratio is greater than the preset merge threshold, merge the two target classes into one target class; until the target classes cannot be merged, each remaining target class corresponds to a plant.

[0028] In one embodiment, the method of dividing the overlapping area of ​​plants by performing regional growth according to the absorption point of each plant in the overlapping area of ​​plants is:

[0029] All absorption points are used as seed points, and the overlapping areas of plants are divided using the region growing method, where the range of all absorption points corresponding to each plant after region growth belongs to the same plant; the growth criterion in the region growing method is that the probability of the seed point and the pixel point being in the same area is greater than the preset same-area threshold.

[0030] The beneficial effects of this application are:

[0031] The present application combines the spectral detection method to perform spectral detection on the plants, combines the spectral detection results with the images, and accurately divides the parts belonging to different plants through the difference characteristics of the absorption of various nutrients in each area in the local overlapping part over time. The complex overlapping situation of plants can be divided more accurately, and the areas belonging to each plant can be divided from each other. The division results are more accurate, which is convenient for subsequent judgment of the growth conditions of each plant. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0033] Figure 1 A flow chart of a spectral detection method for crop plant segmentation in a plant factory provided in one embodiment of the present application;

[0034] Figure 2 This is a schematic diagram of the minimum circumscribed circle. DETAILED DESCRIPTION

[0035] In order to further explain the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following is a detailed description of the spectral detection method for crop plant segmentation in a plant factory proposed by the present application, its specific implementation method, structure, characteristics and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0036] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0037] An embodiment of a spectral detection method for segmenting crop plants in a plant factory:

[0038] The specific scheme of the spectral detection method for segmenting crop plants in a plant factory provided by the present application is described in detail below with reference to the accompanying drawings.

[0039] See also Figure 1 , which shows a flow chart of a spectral detection method for crop plant segmentation in a plant factory provided by an embodiment of the present application, the method comprising the following steps:

[0040] Step S001, collecting an image before spraying and several images after spraying.

[0041] In plant factories, crops are sprayed with nutrient solution regularly. The nutrient solution contains a variety of elements that are beneficial to crop growth, among which nitrogen content is relatively high and can be detected by spectral imaging devices. The hyperspectral imaging device continuously shoots plants of the same type of crop, and collects spectral images of the plants before and after spraying the nutrient solution. The collected spectral images are sorted in time sequence. When spraying the nutrient solution, it is not sprayed directly over a large area, but the machine controls the nozzle to move and spray in a small range, and only sprays the branches and leaves of the plants.

[0042] The spectral image collected before spraying the nutrient solution is recorded as the pre-spraying image, and the spectral image collected after spraying the nutrient solution is recorded as the post-spraying image.

[0043] The hyperspectral imager used in this embodiment is LeafSpec-Dicot, the crop is soybean, and the parameters of the hyperspectral imager are: the central wavelength of blue light is 435nm, the band is 420-450nm; the central wavelength of near infrared is 750nm, the band is 720-780nm. The data is collected once before spraying the nutrient solution, once after spraying the nutrient solution, and then once every half an hour, for a total of 3 hours, and a total of eight spectral images are obtained.

[0044] So far, the pre-spraying image and several post-spraying images have been acquired.

[0045] Step S002, obtaining the before-after difference image; segmenting the before-after difference image to obtain the plant area; obtaining the possibility that the plant area is a plant overlapping area based on the difference between the plant area and its minimum circumscribed circle area, the area of ​​the plant area and the mean value of the nitrogen content of the plant area, and screening the plant overlapping area.

[0046] After spraying the nutrient solution, a large amount of nutrient solution adheres to the surface of the plant, and the nitrogen content also increases. Therefore, the existing method generally sets a nitrogen content threshold, segments the spectral image, and determines the location of each plant. However, static threshold segmentation is often too one-sided. The situation of spraying the nutrient solution is different each time. Simply setting a static nitrogen content threshold cannot guarantee that the segmentation result is accurate every time.

[0047] However, since there is no nutrient solution on the surface of each plant before spraying the nutrient solution, the nitrogen concentration is relatively low, and the nitrogen concentration in the plant area is quite different before and after spraying the nutrient solution. Therefore, the location of each plant can be accurately located based on the change in nitrogen concentration before and after spraying the nutrient solution.

[0048] Subtract the pixel values ​​of each pixel point at the same position in the first image after spraying from that in the image before spraying to obtain a difference image recorded as the before-after difference image, wherein the pixel value of each pixel point in the spectral image is the nitrogen content, and the pixel value of each pixel point in the before-after difference image is the nitrogen difference value.

[0049] After spraying the nutrient solution, the nitrogen concentration in the plant area should increase, while the nitrogen content in the non-plant area changes less. Therefore, for the before-after difference image, the pixel value of the pixel point in the plant area should be significantly greater than the pixel value of the pixel point in the non-plant area. Therefore, all the pixel points of the before-after difference image are threshold segmented, and the pixel points are divided into two categories. Through the connected domain analysis, multiple connected domains in the before-after difference image are obtained, and several connected domains composed of pixels with large pixel values ​​are used as plant areas. The algorithm used for threshold segmentation in this embodiment is the Otsu threshold method, which is a well-known technology and will not be described in detail here.

[0050] In the obtained plant area, there may be areas where multiple plants overlap, so it is necessary to distinguish the plant areas to determine whether the plant area is a single plant or multiple plants. When plants overlap, the area they occupy is larger than that of a single plant; and when plants overlap, the amount of nutrient solution attached to them is greater than that of a single plant, and the nitrogen concentration is higher; secondly, after the plants overlap, the two plant areas are connected together, and compared to the uniform growth of a single plant that is not affected, the shape of the overlapping area is quite different from that of a single plant. Therefore, the single plant area and the overlapping plant area can be distinguished based on the area occupied by each plant area, the nitrogen concentration, and the uniformity of the area shape.

[0051] Based on the before and after difference images, the area of ​​each plant region is obtained, and the nitrogen mean of all pixels in each plant region is calculated. Since the growth of a single plant will not be affected by other plants, its growth trend in all directions is relatively uniform, and the resulting regional shape will be closer to a quasi-circular shape; when the plants overlap, their branches and leaves intersect with each other and interfere with their growth trends, and multiple plant regions are connected together, and their regional shapes will not be similar to a quasi-circular shape; the minimum circumscribed circle of the overlapping and non-overlapping plant regions is as follows Figure 2 shown.

[0052] Therefore, for each plant area, its minimum circumscribed circle is obtained, and the shape of the plant area is represented by the difference between the area of ​​the plant area and the area of ​​the minimum circumscribed circle. The possibility that the plant area is a plant overlapping area is obtained based on the area of ​​the plant area, the area difference between the plant area and the minimum circumscribed circle, and the mean nitrogen content of the plant area.

[0053] The possibility of overlapping plant regions was positively correlated with the area of ​​the plant region and the mean nitrogen content of the plant region, and negatively correlated with the area difference between the plant region and the minimum circumscribed circle.

[0054] It should be noted that positive correlation means that when one variable increases, the other variable also increases, and the two variables change in the same direction. When one variable changes from large to small or from small to large, the other variable also changes from large to small or from small to large. The specific relationship is determined by actual application and this application does not impose any special restrictions.

[0055] It should be noted that negative correlation means that when one variable increases, the other variable decreases accordingly, and the two variables change in opposite directions. When one variable changes from large to small or from small to large, the other variable also changes from small to large or from large to small. The specific relationship is determined by actual application and this application does not impose any special restrictions.

[0056] Preferably, in this embodiment, the possibility that the plant region is a plant overlapping region is:

[0057] , represents the area of ​​the ith plant region, represents the area of ​​the minimum circumscribed circle of the ith plant region, represents the mean nitrogen content of the ith plant region, represents the normalization function, Indicates the possibility that the i-th plant region is a plant overlapping region. It indicates the ratio of plant area to the minimum circumscribed circle area. The smaller the value, the more irregular the shape is and the more likely it is that there are overlapping plants.

[0058] Therefore, an overlapping threshold is set. If the possibility that a plant region is a plant overlapping region is greater than the overlapping threshold, the plant region is considered to be a plant overlapping region. Thus, all plant regions are calculated to screen out all plant overlapping regions. In this embodiment, the overlapping threshold is 0.75.

[0059] At this point, all overlapping areas of the plants have been screened out.

[0060] Step S003, obtain adjacent difference images; and screen the absorption points; obtain the possibility of the two absorption points being in the same area based on the difference in pixel values ​​of the two absorption points in all adjacent difference images and the distance between the absorption points; determine the target class of each absorption point based on the possibility of each absorption point being in the same area with all other absorption points; merge the target classes according to the intersection and union between the target classes to obtain the absorption points corresponding to each plant.

[0061] There are structures such as stomata on the leaves of plants that can absorb nitrogen into the plant body. The nitrogen absorption rate is higher around these structures, which can represent the range of the structure and the nitrogen absorption rate of the plant. Therefore, the main nitrogen absorption point of the plant can be found according to the nitrogen absorption significance.

[0062] For all the images after spraying, the pixel values ​​of each pixel at the same position of two adjacent images after spraying are subtracted, and the absolute value of the difference is used as the pixel value of the pixel. A difference image is obtained for every two adjacent images after spraying, which is recorded as the adjacent difference image.

[0063] For each overlapping area of ​​plants, the absorption rate of nitrogen at its internal absorption point is relatively large, so the absorption point can be determined based on the peak point of nitrogen change in the area. For each adjacent difference image, select its maximum point in the overlapping area of ​​plants. If the pixel value of a pixel point is larger than the pixel value of all pixels in the 8-neighborhood, the pixel point is considered to be a maximum point. Obtain all the maximum points in the adjacent difference images and use the maximum value as the absorption point of the overlapping area of ​​plants.

[0064] To determine whether two absorption points belong to the same plant, the distance between the two points and the nitrogen absorption rate of the two points can be used to evaluate. The closer the distance is and the more different the nitrogen absorption rate is, the less likely it is that the two absorption points belong to the same plant. Therefore, the absorption points can be distinguished based on the distance between the two absorption points and the consistency of the absorption rate. For the adjacent difference images, the pixel value of each pixel is the nitrogen change of the pixel over a period of time. The nitrogen that disappears is the nitrogen absorbed by the plant. Therefore, the pixel value of each pixel in the adjacent difference images is the nitrogen absorption rate.

[0065] For each overlapping area of ​​each plant, two absorption points are randomly selected and the distance between the two absorption points is calculated by Euclidean distance. The possibility of the two absorption points being in the same area is obtained based on the difference in nitrogen absorption rate of the two absorption points in all adjacent difference images and the distance between the absorption points.

[0066] The probability of two absorption points being in the same area is positively correlated with the difference in nitrogen absorption rates between the two absorption points and negatively correlated with the distance between the absorption points.

[0067] Preferably, in this embodiment, the expression for the possibility of two absorption points being in the same region is:

[0068] , represents the nitrogen absorption rate of the jth absorption point in the adjacent difference image between the ath spraying image and the a+1th spraying image, represents the nitrogen absorption rate of the kth absorption point in the adjacent difference image between the ath image after spraying and the a+1th image after spraying, H represents the number of images after spraying, represents the distance between the jth absorption point and the kth absorption point, represents the normalization function, Indicates the possibility that the j-th absorption point and the k-th absorption point are in the same area.

[0069] Preferably, in another embodiment of the present application, the expression for the possibility of two absorption points being in the same region is:

[0070] , represents the nitrogen absorption rate of the jth absorption point in the adjacent difference image between the ath spraying image and the a+1th spraying image, represents the nitrogen absorption rate of the kth absorption point in the adjacent difference image between the ath image after spraying and the a+1th image after spraying, H represents the number of images after spraying, represents the distance between the jth absorption point and the kth absorption point, represents the normalization function, represents the possibility that the j-th absorption point and the k-th absorption point are in the same region, Represents an exponential function with a natural constant as its base.

[0071] The closer the distance between two absorption points is, the more likely it is that the two absorption points belong to the same plant. The greater the difference in nitrogen absorption rate between the two absorption points is, the less likely it is that the two absorption points belong to the same plant.

[0072] Take each absorption point as the target point and calculate the possibility of the target point and each absorption point being in the same area. If the possibility of being in the same area is greater than the same area threshold, the two points are considered to be the same plant. If the possibility of being in the same area is less than the same area threshold, the two points are considered not to be the same plant. In this embodiment, the same area threshold is 0.7.

[0073] Thus, all absorption points are divided into two categories. The absorption points of the same plant as the target point are classified as one category, and this category is recorded as the target class of the target point, and the target class of all absorption points is obtained. For each target class, all the absorption points in it are counted, and the intersection and union of the two target classes are calculated. If the intersection and union ratio is greater than the merging threshold, in this embodiment, the merging threshold is 0.8, then the two target classes are merged into one target class, and all target classes are merged two by two based on this. When merging is impossible, each remaining target class corresponds to an absorption point of a plant.

[0074] At this point, the absorption points corresponding to different plants in each plant overlapping area are determined.

[0075] Step S004, dividing the overlapping area of ​​plants by performing regional growth according to the absorption points of each plant in the overlapping area of ​​plants, and obtaining the area corresponding to each plant.

[0076] All absorption points are used as seed points, and the overlapping areas of plants are divided using the regional growing method. The range of all absorption point areas corresponding to each plant after regional growth belongs to the same plant, thereby distinguishing different plants, and uploading the divided images to the system. In order to ensure that each column of plants can obtain enough nutrient solution, the system controls the angle of the nozzle through the image of a single plant uploaded in the system, so that the nutrient solution sprayed can be evenly sprayed on each plant, ensuring that each column of plants absorbs the nutrient solution. The growth criterion in the regional growing method is that the probability of the seed point and the pixel point being in the same area is greater than the threshold of the same area.

[0077] It should be noted that the above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

[0078] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A spectral detection method for crop plant segmentation in a plant factory, characterized in that: The method comprises the following steps: Collecting a pre-spraying image and several post-spraying images, wherein the pixel value of each pixel in the image is the nitrogen content, the post-spraying image is an image of the crop collected after spraying a nutrient solution with a high nitrogen content, and the pre-spraying image is an image of the crop before spraying the nutrient solution; Subtracting the first image after spraying from the image before spraying to obtain a before-after difference image; segmenting the before-after difference image to obtain a plant region; obtaining a minimum circumscribed circle of the plant region, obtaining the possibility that the plant region is a plant overlapping region based on the difference between the area of ​​the plant region and its minimum circumscribed circle, the area of ​​the plant region and the average value of the nitrogen content of the plant region, and screening the plant overlapping region based on the possibility; For adjacent images after spraying, two-by-two differences are made to obtain adjacent difference images; absorption points are selected according to the pixel values ​​of the pixels in the adjacent difference images; the possibility of the two absorption points being in the same region is obtained according to the difference in the pixel values ​​of the two absorption points in all adjacent difference images and the distance between the absorption points; the target class of each absorption point is determined according to the possibility of each absorption point being in the same region with all other absorption points; the target classes are merged according to the intersection and union between the target classes to obtain the absorption points corresponding to each plant; The overlapping area of ​​the plants is divided by performing regional growth according to the absorption points of each plant in the overlapping area of ​​the plants, and the area corresponding to each plant is obtained; The method for screening absorption points according to the pixel values ​​of the pixels in the adjacent difference images is: For each adjacent difference image, select the maximum point in the overlapping area of ​​the plant; if the pixel value of a pixel is greater than the pixel values ​​of all pixels in the 8-neighborhood, then the pixel is the maximum point; obtain all the maximum points in the adjacent difference images and use the maximum points as the absorption points of the overlapping area of ​​the plant; The method for determining the target class of each absorption point according to the possibility of each absorption point being in the same region as all other absorption points is: Take each absorption point as the target point, calculate the probability of the target point and each absorption point being in the same region, if the probability of the same region is greater than the preset threshold of the same region, then the two points are considered to be the same plant, if the probability of the same region is less than the threshold of the same region, then the two points are considered not to be the same plant; All absorption points of the same plant as the target point are taken as the target class of the target point.

2. A spectral detection method for crop plant segmentation in a plant factory as claimed in claim 1, characterized in that: The method of obtaining the before-after difference image by subtracting the first image after spraying from the image before spraying is: The pixel values ​​of each pixel point at the same position in the first image after spraying and the image before spraying are subtracted to obtain a difference image, and the difference image is recorded as the before-after difference image.

3. The spectral detection method for crop plant segmentation in a plant factory according to claim 1, characterized in that: The method for segmenting the front and back difference images to obtain the plant area is: Threshold segmentation is performed on the pixel values ​​of all pixels in the before and after difference images, and the pixels are divided into two categories. One category of pixels is marked as 1, and the other category of pixels is marked as 0 for connected domain analysis. Each connected domain is regarded as a plant area.

4. The spectral detection method for crop plant segmentation in a plant factory according to claim 1, characterized in that: The method for obtaining the possibility that the plant region is a plant overlapping region according to the difference between the plant region and the area of ​​its minimum circumscribed circle, the area of ​​the plant region and the average value of the nitrogen content in the plant region is: Calculate the mean value of nitrogen content of all pixels in the plant area as the mean value of nitrogen content in the plant area; The possibility of overlapping plant regions was positively correlated with the area of ​​the plant region and the mean nitrogen content of the plant region, and negatively correlated with the area difference between the plant region and the minimum circumscribed circle.

5. The spectral detection method for crop plant segmentation in a plant factory according to claim 1, characterized in that: The method for obtaining adjacent difference images by performing pairwise subtraction on adjacent sprayed images is: For all the images after spraying, the pixel values ​​of each pixel at the same position of two adjacent images after spraying are subtracted, and the absolute value of the difference is used as the pixel value of the pixel. A difference image is obtained for every two adjacent images after spraying, which is recorded as the adjacent difference image.

6. The spectral detection method for crop plant segmentation in a plant factory according to claim 1, characterized in that: The method for obtaining the possibility of the two absorption points being in the same region according to the difference in pixel values ​​of the two absorption points in all adjacent difference images and the distance between the absorption points is: The probability of two absorption points being in the same region is positively correlated with the difference in pixel values ​​of the two absorption points and negatively correlated with the distance between the absorption points.

7. The spectral detection method for crop plant segmentation in a plant factory according to claim 1, characterized in that: The method of merging the target classes according to the intersection and union between the target classes to obtain the absorption points corresponding to each plant is: Get the target class of all absorption points; for each target class, count all the absorption points in it, calculate the intersection and union of the two target classes, and if the intersection and union ratio is greater than the preset merge threshold, merge the two target classes into one target class; until the target classes cannot be merged, each remaining target class corresponds to a plant.

8. The spectral detection method for crop plant segmentation in a plant factory as claimed in claim 1, characterized in that: The method for dividing the overlapping area of ​​plants by performing regional growth according to the absorption point of each plant in the overlapping area of ​​plants is: All absorption points are used as seed points, and the overlapping areas of plants are divided using the region growing method, where the range of all absorption points corresponding to each plant after region growth belongs to the same plant; the growth criterion in the region growing method is that the probability of the seed point and the pixel point being in the same area is greater than the preset same-area threshold.

Images