A method for physical modal segmentation and physiological quality collaborative evaluation of seedling stems and leaves

Abstract

This application discloses a method for collaborative evaluation of the physical modal segmentation of seedling stems and leaves and their physiological quality, belonging to the field of agricultural robot perception and precision detection. The method includes the following steps: outputting a horizontal sinusoidal excitation signal and acquiring at least two images of the seedling under two morphological conditions within the leaf resonant frequency range; performing image difference on any two images to obtain a grayscale difference map; extracting a dynamic mask of the leaves from the grayscale difference map; segmenting the leaves and stems in one of the two images based on the dynamic mask; determining the measured amplitude of the leaves based on the dynamic mask; determining the physiological stiffness of the leaves based on the measured amplitude; determining the stiffness difference between the physiological stiffness and a stiffness threshold, and determining the ratio of the stiffness difference to the stiffness threshold; and determining the seedling quality based on the ratio. This application improves the accuracy and robustness of stem and leaf recognition and segmentation, and meets the requirements for early non-destructive recognition under low computational power and water stress.

Description

Technical Field

[0001] This application belongs to the field of agricultural robot perception and precision detection, specifically involving a method for the collaborative evaluation of physical modal segmentation and physiological quality of seedling stems and leaves. Background Technology

[0002] In the implementation of automated agricultural seedling production, accurate identification and separation of stems and leaves are key steps to achieve automated grafting, transplanting, and phenotypic analysis.

[0003] However, existing visual recognition technologies still face significant bottlenecks in practical applications: First, the chlorophyll content of plant stems and leaves is similar, resulting in highly overlapping spectral features in the visible light band, leading to highly similar colors and a "metachromatic" phenomenon. Traditional visual algorithms are prone to missegmentation in the area where the petiole and stem connect. Second, the greenhouse environment presents unstructured interference such as dynamic shadows, high light reflection, and foliage occlusion, severely affecting the stability of image texture information. Although deep learning models can partially alleviate these problems, they typically rely on high-performance hardware, making them difficult to deploy on low-power, resource-constrained agricultural embedded devices. Furthermore, their versatility in handling variable seedling conditions is poor, limiting their practical application in large-scale production.

[0004] Therefore, there is an urgent need to develop a stem and leaf recognition and segmentation method that balances accuracy, robustness, and low computational requirements to support the further development of agricultural automation. Summary of the Invention

[0005] The purpose of this application is to provide a method for the collaborative evaluation of physical modal segmentation of seedling stems and leaves and physiological quality, which can solve the problems of how to improve the accuracy and robustness of stem and leaf identification and segmentation, as well as meet the early non-destructive identification under low computing power requirements and water stress.

[0006] To solve the above-mentioned technical problems, this application is implemented as follows:

[0007] In a first aspect, embodiments of this application provide a method for the collaborative evaluation of physical modal segmentation of seedling stems and leaves and physiological quality, the method comprising:

[0008] Output a horizontal sinusoidal excitation signal and acquire at least two images of the seedling in two morphologies; wherein, the horizontal sinusoidal excitation signal is used to control the exciter to drive the push rod to perform linear reciprocating motion at a characteristic frequency, the characteristic frequency being within the resonant frequency range of the leaves of the seedling, and the resonant frequency ranges of the leaves and the stem of the seedling being different.

[0009] Perform image difference on any two images to obtain a grayscale difference image;

[0010] Extract the dynamic mask of the leaf from the grayscale difference image; wherein, the dynamic mask is a mask for the same leaf in two different forms in any two images;

[0011] Based on the dynamic mask, the leaves and the stem are segmented in one of the two images;

[0012] Based on the dynamic mask, the measured amplitude of the blade is determined;

[0013] Based on the measured amplitude, the physiological stiffness of the blade is determined;

[0014] Determine the stiffness difference between the physiological stiffness and the stiffness threshold, and determine the ratio of the stiffness difference to the stiffness threshold;

[0015] The quality of the seedlings is determined based on the ratio.

[0016] Secondly, embodiments of this application provide a seedling stem and leaf physical modal segmentation and physiological quality co-evaluation device, the seedling stem and leaf physical modal segmentation and physiological quality co-evaluation device comprising:

[0017] The acquisition module is used to output a horizontal sinusoidal excitation signal and acquire at least two images of the seedling in two different morphologies; wherein, the horizontal sinusoidal excitation signal is used to control the exciter to drive the push rod to perform linear reciprocating motion at a characteristic frequency, the characteristic frequency being within the resonant frequency range of the leaves of the seedling, and the resonant frequency ranges of the leaves and the stem of the seedling being different.

[0018] The difference module is used to perform image difference on any two images to obtain a grayscale difference image.

[0019] An extraction module is used to extract the dynamic mask of the blade from the grayscale difference image.

[0020] The segmentation module is used to segment the leaf and the stem in one of the two images based on the dynamic mask.

[0021] The first determining module is used to determine the measured amplitude of the blade based on the dynamic mask;

[0022] The second determining module is used to determine the physiological stiffness of the blade based on the measured amplitude.

[0023] The third determining module is used to determine the stiffness difference between the physiological stiffness and the stiffness threshold, and to determine the ratio of the stiffness difference to the stiffness threshold.

[0024] The fourth determining module is used to determine the quality of the seedlings based on the ratio.

[0025] Thirdly, embodiments of this application provide a computer device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method described in the first aspect.

[0026] Fourthly, embodiments of this application provide a computer-readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0027] Fifthly, embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0028] This application proposes a method for the collaborative evaluation of physical modal segmentation of seedling stems and leaves and physiological quality. This method utilizes the different resonant frequency ranges of leaves and stems. First, it induces leaf resonance and acquires at least two images of the leaf in two different morphologies during resonance. Second, it performs image difference analysis on any two images to obtain a grayscale difference image, and extracts the dynamic mask of the leaf from the grayscale difference image. Finally, based on the dynamic mask, it segments the leaf and stem in one of the two images. This achieves extremely high stem-leaf displacement difference, i.e., extremely high physical signal-to-noise ratio. Since it eliminates the need for visual algorithms for segmentation, it fundamentally solves the problems faced by traditional visual algorithms, including the "metachromatic" phenomenon leading to easy missegmentation in the petiole-stem connection area, and unstructured interference such as dynamic shadows, high light reflection, and foliage occlusion in greenhouse environments, which severely affect the stability of image texture information. Furthermore, because this method has a simple computational process and does not rely on high-performance hardware, it can be deployed on agricultural embedded devices. Furthermore, this method also determines the measured amplitude of the leaf based on the dynamic mask; determines the physiological stiffness of the leaf based on the measured amplitude; determines the stiffness difference between the physiological stiffness and the stiffness threshold, and determines the ratio of the stiffness difference to the stiffness threshold; and determines the seedling quality based on the ratio. In other words, the quality of the seedling is ultimately determined through measured amplitude, achieving early, non-destructive perception of physiological quality. Compared to traditional visual monitoring, which requires waiting for geometric deformation (wilt) in the leaves to identify stress, this method, through dynamic modal perception, can capture signals at an early stage when physiological stiffness undergoes a slight shift, providing a decision-making basis that precedes visual features for precise horticultural regulation. In summary, this method improves the accuracy and robustness of stem and leaf identification and segmentation, and meets the requirements for low computational power and early, non-destructive identification of water stress. Attached Figure Description

[0029] Figure 1This is a system architecture that can be applied to the embodiments of this application;

[0030] Figure 2 This is a flowchart illustrating the method for the collaborative evaluation of physical modal segmentation of seedling stems and leaves and physiological quality provided in some embodiments of this application;

[0031] Figure 3 This is a grayscale difference image provided in some embodiments of this application;

[0032] Figure 4 This is a structural block diagram of a seedling stem and leaf physical modality segmentation and physiological quality collaborative evaluation device provided in some embodiments of this application;

[0033] Figure 5 These are internal structural diagrams of a computer device provided in some embodiments of this application. Detailed Implementation

[0034] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0035] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0036] The following, in conjunction with the accompanying drawings, provides a detailed description of the seedling stem and leaf physical modality segmentation and physiological quality collaborative evaluation method provided in this application through specific embodiments and application scenarios.

[0037] Figure 1 This is a system architecture that can be applied to the embodiments of this application, such as... Figure 1 As shown, it may include an excitation subsystem, an imaging subsystem, and a control and processing subsystem.

[0038] The excitation subsystem includes a vibrator, monitoring sensors, and a seedling fixing base. The vibrator can be a high-frequency response voice coil motor or a piezoelectric ceramic excitation table, used to output a horizontal sinusoidal excitation signal based on the control signal output from the control and processing subsystem. The monitoring sensor can be a laser displacement sensor, used to monitor the reference amplitude input to the seedling fixing base in real time, serving as a reference for physiological parameter inversion. The seedling fixing base is rigidly connected to the vibrator's push rod and can use flexible shock-absorbing clamps to hold standard plug seedlings.

[0039] The imaging subsystem includes an industrial camera, a lens, and a light source. The industrial camera can be a global shutter area scan camera, preferably with a resolution of 5 megapixels or higher, and has an external hardware trigger interface to connect to the control and processing subsystem. The lens can be a low-distortion, large-depth-of-field macro lens, and the light source can be a high-frequency strobe source to ensure that no motion blur occurs during imaging under resonant high-frequency motion.

[0040] The connection control and processing subsystem includes a main control unit and a synchronization controller. The main control unit can be an embedded computing platform used for physiological parameter inversion and image difference operations. The synchronization controller can be a signal generation circuit based on a field-programmable gate array (FPGA) for synchronously outputting sinusoidal excitation signals and camera trigger pulses, with a phase synchronization error less than [value missing]. .

[0041] In one exemplary embodiment, a method for applying Figure 1 A systematic method for the joint evaluation of physical modal segmentation of seedling stems and leaves and physiological quality. (Refer to...) Figure 2 , Figure 2 This is a flowchart illustrating a method for the coordinated evaluation of seedling stem and leaf physical modality segmentation and physiological quality, provided in some embodiments of this application. The method includes steps 202-216. Wherein:

[0042] Step 202: Output a horizontal sinusoidal excitation signal and acquire at least two images of the seedling in two different morphologies; wherein, the horizontal sinusoidal excitation signal is used to control the vibrator to drive the push rod to perform linear reciprocating motion at a characteristic frequency, the characteristic frequency being within the resonant frequency range of the leaves of the seedling, and the resonant frequency ranges of the leaves and the stem of the seedling being different.

[0043] Different objects have different stiffnesses, and stiffness is positively correlated with natural frequency. When the vibration frequency is close to the object's natural frequency, the object will resonate; the vicinity of the natural frequency is the resonant frequency range. In the embodiments of this application, the stiffness of the leaves and the stiffness of the stem differ significantly; therefore, their resonant frequency ranges differ considerably. When the vibration frequency causes the leaves to resonate, this vibration frequency will not cause the stem to resonate.

[0044] Based on this, this application proposes an excitation-response mechanism that utilizes the differences in the biomechanical properties of various organs of the seedling to achieve the physical segmentation of the seedling's organs. Specifically, by utilizing the characteristic frequency within the resonant frequency range of the leaf, the leaf can resonate while the stem does not, thus physically segmenting the leaf and stem.

[0045] Specifically, by applying a specific frequency excitation within the first-order resonance band (i.e., the resonance frequency range) of the blade, the high-stiffness stem is placed in a "quasi-rigid body following state" far from the resonance region, while the low-stiffness blade enters a "local resonance state," generating a nonlinear displacement amplification response, thus realizing a physical-level modal filter. When the blade resonates, a drastic displacement difference occurs, i.e., violent oscillation, which can therefore present different morphologies in different images.

[0046] Experimental data shows that the amplitude ratio of leaves to stems can reach 40:1, with the leaf amplitude approaching 40 mm and the stem amplitude around 1 mm. This clearly distinguishes between leaves and stems, significantly improving the signal-to-noise ratio at the physical level. This allows for direct filtering of complex dynamic shadows and highlights in the original image, avoiding the impact of these interferences on the physical modal segmentation of stems and leaves.

[0047] This characteristic frequency can be obtained through calibration, and the calibration process can be as follows:

[0048] Seedlings in different water physiological states are prepared in advance, including normal seedlings, water-stressed seedlings, and stunted seedlings. For each type of seedling, the vibrator is controlled to drive the push rod to perform linear reciprocating motion at different frequencies within a preset frequency range, thereby determining the corresponding resonance frequency range for each type of seedling. Based on all corresponding resonance frequency ranges, the characteristic frequency is determined. For example, the intersection of all corresponding resonance frequency ranges is first taken, and then the midpoint of the intersection is taken as the characteristic frequency.

[0049] Among them, at least two images in both forms refer to the seedlings not completely overlapping in the images.

[0050] Step 204: Perform image difference on any two images to obtain a grayscale difference image.

[0051] Image difference is generally applied to grayscale images. Then, pixel-level absolute difference is performed on two grayscale images. That is, the pixel values ​​at the same pixel position in the image are subtracted and the absolute value is calculated. This absolute value is used as the new pixel value at the same pixel position, thus obtaining a new image, namely a grayscale difference image.

[0052] Step 206: Extract the dynamic mask of the blade from the grayscale difference image.

[0053] It's understandable that for any two images, because the leaves resonate while the stems do not, the absolute value of the area containing the leaves will be significantly greater than that of the area containing the stems. Similarly, the absolute value of the area outside the seedlings in the two grayscale images will also be significantly smaller than that of the area containing the leaves, even reaching 0. That is, apart from the area containing the leaves, the grayscale values ​​of other areas in the two grayscale images remain essentially unchanged. In other words, in the grayscale difference image, the pixel values ​​of the area containing the leaves will be significantly greater than those of the areas outside the leaves.

[0054] Based on this, a high-precision binary dynamic mask of the blade can be extracted using the Otsu method, that is, the mask of the blade in two morphologies.

[0055] However, because the images of the leaves in the two morphologies may overlap after subtraction, such as... Figure 3 As shown, Figure 3 This is a grayscale difference image provided in some embodiments of this application. Figure 3 The seedlings in the specimens are "two-leaf" seedlings. The "dark area" is the area where the leaves are not located, and the "bright area" is the area where the leaves are located in the two forms. In other words, the "bright area" is a dynamic mask of the two leaves.

[0056] It should be noted that seedlings generally have more than one leaf, such as 2 leaves, 4 leaves, or 6 leaves. Therefore, the number of dynamic masks is also greater than one, the same as the number of leaves. Figure 3 The image includes two "bright area" connected components, meaning the number of dynamic masks is two. In this embodiment, each connected component can be labeled, thereby defining the corresponding pixel blocks as different leaf objects.

[0057] Step 208: Based on the static mask, segment the leaf and the stem in one of the two images.

[0058] In any two images, the image region corresponding to the static mask can be determined as the region where the leaf is located, thus completing the segmentation of the leaf and stem.

[0059] In some embodiments, segmenting the leaf and the stem in one of the two images based on the dynamic mask includes:

[0060] Extract the full-plant mask of the seedling from one of the two images.

[0061] The intersection of the dynamic mask and the whole-plant mask is determined to be the static mask of the leaf.

[0062] Based on the static mask, the leaves and the stem are segmented in one of the two images.

[0063] Specifically, a dynamic mask is obtained using a grayscale difference image obtained through the difference method. This dynamic mask only contains the leaf positions at time T1 and time T2, successfully eliminating stationary stems. A plant mask (referred to as a full-plant mask in this embodiment) is generated using the color index or grayscale threshold of the single-frame image at time T1. The full-plant mask only contains the leaves and stems at time T1, automatically excluding the background in the image, including the empty spaces where the leaves are located at time T2. Finally, a logical AND operation is performed on the dynamic mask and the plant mask to obtain their intersection, which is the static mask for the leaves. After obtaining the static mask, it is used to cover the single-frame image at time T1, resulting in an image containing only the leaves, i.e., the area covered by the static mask in the single-frame image at time T1.

[0064] This embodiment utilizes the different resonant frequency ranges of leaves and stems. First, the leaves are made to resonate, and at least two images of the leaves in two different morphologies are acquired during the resonance. Second, image difference is performed on any two images to obtain a grayscale difference image, and the dynamic mask of the leaves is extracted from the grayscale difference image. Finally, based on the dynamic mask, the leaves and stems are segmented in one of the two images. This achieves extremely high differences in stem and leaf displacement, i.e., extremely high physical signal-to-noise ratio. Since segmentation does not require the use of visual algorithms, it fundamentally solves the problems faced by traditional visual algorithms. These problems include the "metachromatic" phenomenon, which easily leads to missegmentation in the area where the petiole and stem connect, and unstructured interference such as dynamic shadows, high light reflection, and foliage occlusion in greenhouse environments, which seriously affect the stability of image texture information. In addition, because the method has a simple calculation process and does not rely on high-performance hardware, it can be deployed on agricultural embedded devices. Furthermore, this method also determines the measured amplitude of the leaf based on the dynamic mask; determines the physiological stiffness of the leaf based on the measured amplitude; determines the stiffness difference between the physiological stiffness and the stiffness threshold, and determines the ratio of the stiffness difference to the stiffness threshold; and determines the seedling quality based on the ratio. In other words, the quality of the seedling is ultimately determined through measured amplitude, achieving early, non-destructive perception of physiological quality. Compared to traditional visual monitoring, which requires waiting for geometric deformation (wilt) in the leaves to identify stress, this method, through dynamic modal perception, can capture signals at an early stage when physiological stiffness undergoes a slight shift, providing a decision-making basis that precedes visual features for precise horticultural regulation. In summary, this method improves the accuracy and robustness of stem and leaf identification and segmentation, and meets the requirements for low computational power and early, non-destructive identification of water stress.

[0065] In some embodiments, acquiring at least two images of the seedlings in two morphologies includes:

[0066] When the horizontal sinusoidal excitation signal is output, a camera trigger pulse signal is output synchronously; wherein, the camera trigger pulse signal is used to control the industrial camera to image at different times.

[0067] Acquire at least two images of the seedlings in two different morphologies provided by the industrial camera.

[0068] In this embodiment, synchronous sensing is achieved by simultaneously outputting a camera trigger pulse signal when outputting the horizontal sinusoidal excitation signal. That is, while the excitation subsystem drives the seedling fixing base to perform horizontal sinusoidal excitation, the synchronous controller can drive the camera to image at the moment corresponding to the phase of the excitation signal, thus facilitating the acquisition of at least two images of the seedling in two different states. For example, based on synchronous sensing, only two images can be captured to obtain images of the seedling in two different states, rather than obtaining images with identical states.

[0069] In some embodiments, the exposure time is strictly controlled within 1ms, and ultra-short exposure combined with subpixel displacement tracking technology is used to "freeze" the high-speed movement trajectory of the blades in order to eliminate motion blur.

[0070] In some embodiments, the camera trigger pulse signal is used to control the phase of the industrial camera relative to the horizontal sinusoidal excitation signal. and Time-lapse imaging. This means that an image frame can be acquired at the moment corresponding to the phase peak and trough of the horizontal sinusoidal excitation signal, thus obtaining two images under the condition of maximum blade displacement difference.

[0071] In some embodiments, the method further includes:

[0072] Based on the static mask, the area and outline of the blade are determined;

[0073] If the area is smaller than the preset area or the outline is deformed, the seedling is determined to be a weak seedling or a deformed seedling.

[0074] The area of ​​the leaf can be determined by the number of pixels in the static mask and the system magnification (i.e., the conversion ratio between real-world coordinates and camera coordinates), and the outline of the leaf is the boundary of the static mask.

[0075] It should be noted that the preset area and contour distortion can be set according to the specific application scenario, and this application embodiment does not limit them.

[0076] In some embodiments, if the area is greater than or equal to a preset area and the outline is normal, then the step of determining the measured amplitude of the leaf based on the dynamic mask is performed. The quality of the seedling is then further assessed using the measured amplitude.

[0077] It can be understood that the above judgment logic is a layered judgment logic. The first layer is morphological screening, which uses two indicators, area and contour, for judgment. The second layer is physiological screening, which, based on the morphological qualification, can use two indicators, amplitude and / or phase, for judgment.

[0078] In some embodiments, based on the excitation response mechanism and combined with dynamic characteristics, this application embodiment constructs a dynamic detection model with a two-way perception-evaluation mapping. These dynamic characteristics may include amplitude and phase lag angle, thereby utilizing the differences in biomechanical properties of various organs of the seedling to achieve modal separation at the physical level while simultaneously completing the quantitative inversion of the physiological quality of the seedling.

[0079] Specifically, the dynamic response characteristics of leaves under resonant conditions (such as amplitude and phase lag angle) are defined as "physiological quality fingerprints." By establishing a monotonic mapping model between dynamic response characteristics and plant physiological stiffness (Young's modulus), physiological quality evaluation results reflecting seedling moisture content and tissue density are simultaneously output while the segmentation process is completed. Water loss in seedlings leads to decreased moisture content, which in turn reduces tissue density and consequently decreases physiological stiffness. Seedling moisture content is strongly correlated with physiological quality; standard quality seedlings are well-hydrated with high moisture content, while decreased moisture content leads to decreased physiological quality. Seedlings with a significant decrease in moisture content are classified as water-stressed seedlings.

[0080] The amplitude can be used to determine whether the seedlings are short of water; the phase lag angle can be used for auxiliary verification. It should be noted that water shortage not only affects the amplitude but also changes the vibration lag time. If the amplitude exceeds the standard and the phase angle is abnormal, the seedlings can be identified as water-stressed seedlings.

[0081] Step 210: Determine the measured amplitude of the blade based on the dynamic mask.

[0082] In this embodiment, it can be referred to Figure 3 The measured amplitude of the leaf can be half the distance between the leaf tips in the two morphologies, or half the distance between the centroids of the connected domains in the two morphologies.

[0083] When determining the measured amplitude of the blade based on a dynamic mask, since the measured amplitude of the blade is half the distance between the blade tips in both morphologies:

[0084] The first pixel coordinates of each leaf tip of the same leaf in the dynamic mask are determined, and the first coordinate difference is calculated; based on the system magnification, the first coordinate difference is magnified to the first distance of the leaf tip in the real world; wherein the system magnification is the conversion ratio between real world coordinates and camera coordinates; half of the first distance is determined as the measured amplitude of the leaf.

[0085] It can be understood that the first distance is the vertical distance from the crest to the trough of the sinusoidal motion at the blade tip when the blade resonates, which is twice the amplitude of the sinusoidal motion. Therefore, the measured amplitude is half of the first distance.

[0086] When determining the measured amplitude of the blade based on a dynamic mask, since the measured amplitude of the blade is half the distance between the centroids of the connected domains in both morphologies:

[0087] Determine the second pixel coordinates of the two centroids in the connected domain corresponding to the same blade in the dynamic mask, and calculate the second coordinate difference; based on the system magnification, magnify the second coordinate difference to the second distance of the blade centroid in the real world; wherein the system magnification is the conversion ratio between real world coordinates and camera coordinates; and determine half of the second distance as the measured amplitude of the blade.

[0088] It can be understood that the second distance is the vertical distance from the crest to the trough of the sinusoidal motion of the blade's center of mass when the blade resonates, which is twice the amplitude of the sinusoidal motion. Therefore, the measured amplitude is half of the second distance.

[0089] Specifically, for any single leaf of the seedling:

[0090] First, if overlap occurs, the dynamic mask is determined to be a connected mask. In this case, the connected mask's role is to lock the envelope of the leaf movement and exclude the stem, thus defining the connected mask as the Region of Interest (ROI). Centroids are extracted from single-frame images; that is, although the dynamic mask is connected, the leaf position is fixed in the original image, where the original image represents the leaf in two different morphologies. Based on this, the two centroids in the connected mask for the same leaf are determined through the following two steps:

[0091] Step A: Cover the connected mask with On the original image at that moment. Within the connected mask, leaf pixels are extracted using grayscale / color thresholding, and the second pixel coordinates of their centroids are calculated. .

[0092] Step B: Cover the connected mask with On the original image at that moment. Similarly, extract the leaf pixels and calculate the second pixel coordinates of its centroid. .

[0093] Step C: Calculation and The Euclidean distance is the second coordinate difference, which is also the displacement of the blade. Using the above method, even if the blade shapes at two different times are highly overlapping in space, forming a large connected region, the centroid positions at the two times can still be accurately separated, thus accurately calculating the measured amplitude without the need for complex geometric segmentation on the mask image.

[0094] Secondly, if no overlap occurs, the centroids of each connected domain can be extracted directly, and the second distance can be determined by the two centroids of the same leaf.

[0095] Step 212: Determine the physiological stiffness of the blade based on the measured amplitude.

[0096] In some embodiments, considering the biological nonhomogeneity of seedling organs and the sensitivity of physiological states to stiffness, this embodiment constructs an equivalent multi-degree-of-freedom (MDOF) dynamic discrete model based on Euler-Bernoulli beam theory, as shown in formula (1):

[0097] (1)

[0098] in, For amplitude vectors, Here is the stiffness matrix; This is the velocity vector, which is the first derivative of displacement with respect to time. Here is the damping matrix; This is the acceleration vector, i.e., the second derivative of displacement with respect to time. This is the quality matrix; For time.

[0099] For the elastic force term, by Decide, It depends directly on the physiological stiffness of the seedling, i.e., Young's modulus. , which is the item to be tested. The damping force term represents the viscoelasticity within the plant, i.e., its ability to dissipate vibrational energy. It is related to water content and corresponds to the quality factor. . Inertial force term The inertial force term is derived from the mass matrix. Decision made. In this plan, Depending on the density of the seedling's biological tissue and the proportion of air cavities, it reflects the inertial resistance generated by the seedling's own mass distribution during vibration. This is the excitation force term, corresponding to the known excitation force applied by the vibrator, which is converted from the reference amplitude.

[0100] Based on formula (1), physiological stiffness can be derived. The calculation formula, i.e., physiological stiffness Determined by the following formula (2):

[0101] (2)

[0102] Among them, the The measured amplitude; The reference amplitude; The characteristic frequency; The damping ratio factor is obtained through the formula. It is confirmed that the The quality factor; the reference amplitude and the It is obtained by standardizing the quality of seedlings.

[0103] In some embodiments, during calibration, the vibrator can be controlled to drive the seedling fixing base to perform a horizontal sinusoidal frequency sweep (e.g., 5Hz-200Hz) with a certain amplitude (e.g., 1.0mm). The frequency response functions for different water physiological states (e.g., normal seedlings, water-stressed seedlings, etc.) are measured and recorded. Through each frequency response function, not only can the characteristic frequencies used for morphological segmentation be locked, but also the reference amplitude of standard quality seedlings at the resonance point and the quality factor reflecting the viscoelastic damping characteristics of the system can be calibrated. .

[0104] It should be noted that formula (1) is a perception-evaluation mapping model. That is, the mathematical basis of formula (2). Formula (1) and formula (2) are the relationship between the original function and the inverse function. Formula (1) is the forward process: it describes the process if the stiffness is known. K How much amplitude will the seedlings produce? Formula (2) is the reverse process: using the solution of formula (1), a model is constructed because the amplitude has been measured. Therefore, the physiological stiffness can be calculated. The algorithm.

[0105] The process of deriving formula (2) from formula (1) is as follows:

[0106] Step 1: Visual Measurement Layer

[0107] Input data: Two frames of images captured by an industrial camera.

[0108] Calculation results: Obtain the blade's... .

[0109] Step 2: Dynamic Inversion Layer

[0110] Based on formula (1), through By reverse reasoning .

[0111] It should be noted that, although the calculation was... But at this time, it is still unknown How much of it comes from materials? (Based on the formulas of mechanics of materials.) It can be deduced that .and (Long) and (Thickness moment of inertia) is a constant for standard quality seedlings, therefore It depends directly on the physiological stiffness of the seedling, i.e., Young's modulus. .

[0112] Step 3: Microscopic Property Material Property Mapping Layer Status: Finally, formula (2) is obtained using the first and second steps. Solve the problem.

[0113] Among them, the stiffness matrix This method is used to establish a segmented model for the biomechanical properties of the stem-leaf junction in seedlings. Since the biomechanical properties of the stem and leaves differ significantly, a segmented model can be established using a stiffness matrix. Specifically, due to the higher degree of lignification in the stem, the corresponding stiffness is greater. For example, the Young's modulus of the stem of a standard quality seedling can be used... The pressure is set to 50 MPa; while the blades can be set to an extremely thin, flexible structure (e.g., thickness). Its stiffness is Young's modulus. As a core mechanical indicator reflecting water health status, it fluctuates with changes in cell turgor pressure caused by physiological water content, such as the Young's modulus of leaves of standard quality seedlings. It can be set to 5MPa.

[0114] Among them, for the mass matrix The tissue air cavity ratio parameter can be introduced based on the theory of plant porous media. The vascular bundles of the stem are dense and high in density. It can be set to The spongy tissue of the leaves is rich in air cavities and has a high density. It can be set to This differentiation setting ensures that the model can sensitively capture frequency response deviations caused by tissue dehydration.

[0115] Step 214: Determine the stiffness difference between the physiological stiffness and the stiffness threshold, and determine the ratio of the stiffness difference to the stiffness threshold.

[0116] The stiffness threshold is the calibrated value of standard quality seedlings, for example, the stiffness threshold is 5MPa.

[0117] Step 216: Determine the quality of the seedlings based on the ratio.

[0118] This ratio can be compared with a preset ratio, which is an empirical value. For example, if the preset ratio is 15%, and the ratio is greater than the preset ratio, the seedlings can be identified as Grade II seedlings (such as seedlings under water stress, weak seedlings, or deformed seedlings); if the ratio is less than or equal to the preset ratio, the seedlings can be identified as Grade I seedlings (standard quality seedlings).

[0119] Furthermore, the sign of the stiffness difference can be used to determine whether secondary seedlings are water-stressed, weak, or deformed. For example, a negative stiffness difference indicates that the secondary seedlings are water-stressed. This allows for early warning of seedling water health status 12-24 hours before visual wilting occurs.

[0120] Furthermore, seedling quality can be automatically marked on the segmentation diagram. This mark can be a grade or a mark indicating whether the seedlings are qualified.

[0121] This embodiment determines seedling quality through measured amplitude, achieving early, non-destructive perception of physiological quality. Compared to traditional visual monitoring, which requires waiting for geometric deformation (wilt) in leaves to identify stress, this embodiment, through dynamic modal perception, can capture signals at an early stage when physiological stiffness undergoes slight shifts, providing a decision-making basis that precedes visual characteristics for precise horticultural regulation.

[0122] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0123] For ease of understanding, a specific embodiment will be used as an example:

[0124] Example 1: Automated grading and screening of transplanted seedlings based on mechanical sensing collaboration.

[0125] 1. Implementation targets and parameter settings:

[0126] In this embodiment, tomato seedlings in plug trays at the 4-6 leaf stage were selected as the target organism. The system dynamics model presupposes the equivalent Young's modulus of healthy individuals. The pressure is 5 MPa. Based on the previous frequency sweep calibration results, the characteristic excitation frequency for the leaves of this variety was selected. The reference amplitude input to the exciter is 44.6 Hz. Set to 1.0 mm.

[0127] 2. Collaborative work process:

[0128] Synchronous Sensing: The excitation subsystem drives the seedling fixing base to perform horizontal sinusoidal excitation. The synchronous controller drives the industrial camera to acquire one frame of image at the peak and trough of the excitation signal phase, with the exposure time strictly controlled within 1ms to eliminate motion blur.

[0129] Physical segmentation: Pixel-level absolute difference is performed between two image frames. At this point, the high-stiffness stem appears as a dark area (signal suppression) in the image due to minimal displacement, while the leaves, entering a local resonant state, exhibit a significant displacement difference and appear as prominent bright areas. The system extracts a high-precision binarized mask of the leaves using the Otsu method.

[0130] Quality assessment: While extracting the mask, the system extracts the motion trajectory of the blade's centroid through connected component analysis, and calculates and obtains the measured amplitude of the blade. .

[0131] 3. Implementation Results and Data Evaluation:

[0132] Grade 1 seedlings (high-quality): If the leaves are measured... According to the mapping model, its physiological stiffness (Young's modulus) is close to 5MPa, indicating that the tissue is full of water and has high rigidity, which meets the mechanical strength requirements of high-speed grasping by automated transplanting robots.

[0133] Second-grade seedlings (water stress / physiologically weak seedlings): If the leaf thickness is measured... An offset of more than 15% (e.g.) The system determined through mapping model inversion that the Young's modulus had dropped to around 3 MPa. This shift revealed that the seedlings suffered from insufficient cell turgor pressure and inadequate tissue uprightness due to early water shortage. The system marked them as unqualified, and the sorting robot removed them.

[0134] Example 2: Early non-contact monitoring of water stress in seedlings under smart greenhouse conditions.

[0135] 1. Implementation Background:

[0136] Greenhouse tomatoes are extremely sensitive to water conditions during the seedling stage. Traditional visual monitoring methods rely on visible changes in leaf shape, such as wilting or curling, to identify water stress, which has a significant time lag.

[0137] 2. Collaborative evaluation mechanism:

[0138] This embodiment utilizes the high sensitivity of seedlings' "physiological stiffness" to water loss to lock the excitation frequency at the first-order resonant frequency of the leaves. The system obtains the dynamic response characteristics of the leaves by periodically (e.g., every 2 hours) performing a 44.6Hz micro-vibration probe on the plug seedlings.

[0139] 3. Judgment Logic and Technical Advantages:

[0140] Multidimensional indicator perception: While simultaneously acquiring blade area and profile, the system focuses on analyzing... With phase lag angle .

[0141] Early monitoring results: Experimental data show that before the leaves show visible deformation, the vibration damping ratio and physiological stiffness of the tissue will change due to the slight change in intracellular water content, resulting in a significant shift in its resonance response at the characteristic frequency.

[0142] Collaborative Application: This embodiment can detect an abnormal decrease in the physiological stiffness of seedlings 12-24 hours before visual wilting occurs, without damaging the seedlings. This detection result can be directly fed back to the greenhouse IoT system, enabling differentiated water replenishment for stressed seedlings through coordinated precision irrigation equipment. This avoids indiscriminate irrigation and improves the precision of seedling production.

[0143] Based on the same inventive concept, this application also provides a device for implementing the above-mentioned method for the coordinated evaluation of physical modal segmentation and physiological quality of seedling stems and leaves. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the device for the coordinated evaluation of physical modal segmentation and physiological quality of seedling stems and leaves provided below can be found in the limitations of the method for the coordinated evaluation of physical modal segmentation and physiological quality of seedling stems and leaves described above, and will not be repeated here.

[0144] In one exemplary embodiment, such as Figure 4 As shown, a device for the joint evaluation of physical modal segmentation and physiological quality of seedling stems and leaves is provided, comprising: an acquisition module 100, a difference module 200, an extraction module 300, a segmentation module 400, a first determination module 500, a second determination module 600, a third determination module 700, and a fourth determination module 800, wherein:

[0145] The acquisition module 100 is used to output a horizontal sinusoidal excitation signal and acquire at least two images of the seedling in two different morphologies; wherein, the horizontal sinusoidal excitation signal is used to control the exciter to drive the push rod to perform linear reciprocating motion at a characteristic frequency, the characteristic frequency being within the resonant frequency range of the leaves of the seedling, and the resonant frequency ranges of the leaves and the stem of the seedling being different.

[0146] The difference module 200 is used to perform image difference on any two images to obtain a grayscale difference image.

[0147] Extraction module 300 is used to extract the static mask of the blade from the grayscale difference image.

[0148] The segmentation module 400 is used to segment the leaf and the stem in one of the two images based on the static mask.

[0149] The first determining module 500 is used to determine the measured amplitude of the blade based on the dynamic mask.

[0150] The second determining module 600 is used to determine the physiological stiffness of the blade based on the measured amplitude.

[0151] The third determining module 700 is used to determine the stiffness difference between the physiological stiffness and the stiffness threshold, and to determine the ratio of the stiffness difference to the stiffness threshold.

[0152] The fourth determining module 800 is used to determine the quality of the seedlings based on the ratio.

[0153] In some embodiments, the acquisition module 100 is specifically used for:

[0154] When the horizontal sinusoidal excitation signal is output, a camera trigger pulse signal is output synchronously; wherein, the camera trigger pulse signal is used to control the industrial camera to image at different times.

[0155] Acquire at least two images of the seedlings in two different morphologies provided by the industrial camera.

[0156] In some embodiments, the camera trigger pulse signal is used to control the phase of the industrial camera relative to the horizontal sinusoidal excitation signal. and Time-lapse imaging.

[0157] In some embodiments, the segmentation module 400 is specifically used for:

[0158] Extract the full-plant mask of the seedling from one of the two images.

[0159] The intersection of the dynamic mask and the whole-plant mask is determined to be the static mask of the leaf.

[0160] Based on the static mask, the leaves and the stem are segmented in one of the two images.

[0161] In some embodiments, the device further includes:

[0162] The fifth determining module is used to determine the area and outline of the blade based on the static mask.

[0163] The determination module is used to determine that the seedling is a weak seedling or a deformed seedling if the area is smaller than a preset area or the outline is deformed.

[0164] In some embodiments, the first determining module 500 is specifically used for:

[0165] Determine the first pixel coordinates of each leaf tip of the same leaf in the dynamic mask, and calculate the first coordinate difference.

[0166] Based on the system magnification, the first coordinate difference is magnified to the first distance of the leaf tip in the real world; wherein, the system magnification is the conversion ratio between real-world coordinates and camera coordinates.

[0167] Half of the first distance is determined as the measured amplitude of the blade.

[0168] In some embodiments, the first determining module 500 is specifically used for:

[0169] Determine the second pixel coordinates of the two centroids in the connected domain corresponding to the same leaf in the dynamic mask, and calculate the second coordinate difference.

[0170] Based on the system magnification, the second coordinate difference is magnified to the second distance of the blade's centroid in the real world; wherein, the system magnification is the conversion ratio between real-world coordinates and camera coordinates.

[0171] Half of the second distance is determined as the measured amplitude of the blade.

[0172] In some embodiments, the physiological stiffness Determined by the following formula:

[0173]

[0174] Among them, the The measured amplitude; The reference amplitude; The characteristic frequency; The damping ratio factor is obtained through the formula. The calculation shows that the The quality factor; the reference amplitude and the It is obtained by standardizing the quality of seedlings.

[0175] Each module in the aforementioned seedling stem and leaf physical modality segmentation and physiological quality collaborative evaluation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0176] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a method for the collaborative evaluation of seedling stem and leaf physical modal segmentation and physiological quality. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0177] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0178] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0179] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0180] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for the collaborative evaluation of physical modal segmentation of seedling stems and leaves and physiological quality, characterized in that, The method for co-evaluating the physical modality segmentation of seedling stems and leaves and the physiological quality includes: Output a horizontal sinusoidal excitation signal and acquire at least two images of the seedling in two morphologies; wherein, the horizontal sinusoidal excitation signal is used to control the exciter to drive the push rod to perform linear reciprocating motion at a characteristic frequency, the characteristic frequency being within the resonant frequency range of the leaves of the seedling, and the resonant frequency ranges of the leaves and the stem of the seedling being different. Perform image difference on any two images to obtain a grayscale difference image; Extract the dynamic mask of the leaf from the grayscale difference image; wherein, the dynamic mask is a mask for the same leaf in two different forms in any two images; Extract the whole-plant mask of the seedling from one of the two images; The intersection of the dynamic mask and the whole-plant mask is determined to be the static mask of the leaf; Based on the static mask, the leaf and the stem are segmented in one of the two images; Based on the dynamic mask, the measured amplitude of the blade is determined; Based on the measured amplitude, the physiological stiffness of the blade is determined; Among them, the physiological stiffness Determined by the following formula: Among them, the The measured amplitude; The reference amplitude; The characteristic frequency; The damping ratio factor is obtained through the formula. The calculation shows that the The quality factor; the reference amplitude and the Obtained by standardizing the quality of seedlings; Determine the stiffness difference between the physiological stiffness and the stiffness threshold, and determine the ratio of the stiffness difference to the stiffness threshold; The quality of the seedlings is determined based on the ratio.

2. The method for synergistic evaluation of physical modal segmentation and physiological quality of seedling stems and leaves according to claim 1, characterized in that, The acquisition of at least two images of the seedlings in two morphologies includes: When the horizontal sinusoidal excitation signal is output, a camera trigger pulse signal is output synchronously; wherein, the camera trigger pulse signal is used to control the industrial camera to image at different times; Acquire at least two images of the seedlings in two different morphologies provided by the industrial camera.

3. The method for synergistic evaluation of physical modal segmentation and physiological quality of seedling stems and leaves according to claim 2, characterized in that, The camera trigger pulse signal is used to control the phase of the industrial camera in response to the horizontal sinusoidal excitation signal. and Time-lapse imaging.

4. The method for synergistic evaluation of physical modal segmentation and physiological quality of seedling stems and leaves according to claim 1, characterized in that, The method further includes: Based on the static mask, the area and outline of the blade are determined; If the area is smaller than the preset area or the outline is deformed, the seedling is determined to be a weak seedling or a deformed seedling.

5. The method for synergistic evaluation of physical modal segmentation and physiological quality of seedling stems and leaves according to claim 1, characterized in that, Determining the measured amplitude of the blade based on the dynamic mask includes: Determine the first pixel coordinates of each leaf tip of the same leaf in the dynamic mask, and calculate the first coordinate difference; Based on the system magnification, the first coordinate difference is magnified to the first distance of the leaf tip in the real world; wherein, the system magnification is the conversion ratio between real-world coordinates and camera coordinates; Half of the first distance is determined as the measured amplitude of the blade.

6. The method for synergistic evaluation of physical modal segmentation and physiological quality of seedling stems and leaves according to claim 1, characterized in that, Determining the measured amplitude of the blade based on the dynamic mask includes: Determine the second pixel coordinates of the two centroids in the connected domain corresponding to the same leaf in the dynamic mask, and calculate the second coordinate difference; Based on the system magnification, the second coordinate difference is magnified to the second distance of the blade's centroid in the real world; wherein, the system magnification is the conversion ratio between real-world coordinates and camera coordinates; Half of the second distance is determined as the measured amplitude of the blade.

7. A computer device, characterized in that, The method includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the seedling stem and leaf physical modality segmentation and physiological quality co-evaluation method as described in any one of claims 1-6.

8. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions, which, when executed by a processor, implement the steps of the seedling stem and leaf physical modality segmentation and physiological quality co-evaluation method as described in any one of claims 1-6.

Images