A method and apparatus for monitoring seed germination

Abstract

The present application relates to the technical field of image analysis, in particular to a seed germination monitoring method and device. The present application sets up multiple layers of three-dimensional culture space; collects images by a movable shooting device according to a preset path; carries out individual segmentation on each seed, constructs a four-dimensional morphological feature vector containing convexity, area, aspect ratio and physical length, and determines whether to germinate according to the four-dimensional morphological feature vector; in the case of no germination, carries out secondary determination by combining contour evolution analysis based on edge continuity, carries out sub-pixel level edge detection by a contour compression algorithm, and carries out convex hull operation on the contour point set to identify the change of concave filling rate to capture the tiny germination target; also drives the shooting device by a three-axis linkage mechanism, and realizes parallel processing of shooting and calculation; can monitor three-dimensional culture seeds in high-throughput and automation, distinguishes seed abnormalities and real germination by combining multi-dimensional morphological features and contour evolution analysis, captures the tiny germination target, and improves the accuracy and robustness of germination determination.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, and in particular to a method and apparatus for monitoring seed germination. Background Technology

[0002] In the fields of agricultural breeding and seed vigor testing, existing technologies typically involve placing seeds in petri dishes within a standard artificial climate chamber for germination experiments, relying primarily on manual observation and counting of seeds by removing the dishes at regular intervals each day. Some automated solutions use conveyor belts to transport seeds for photography, but mechanical vibrations can easily disturb the seed's position. Other solutions employ fixed camera arrays to photograph the culture trays, achieving some automation, but these solutions are extremely expensive and difficult to cover multi-layered, high-density, three-dimensional culture structures, making them unsuitable for large-scale germplasm resource screening or high-throughput vigor testing scenarios. Therefore, existing technologies generally suffer from low monitoring throughput, poor space utilization, delayed data acquisition, reliance on manual judgment leading to significant subjective errors, and a single dimension for germination assessment, making it difficult to distinguish between seed imbibition and true germination, and exhibiting weak detection capabilities for minute germination targets.

[0003] The present invention aims to overcome the defects of the prior art and provide a method and apparatus that can realize high-density three-dimensional culture, automated high-frequency monitoring, and accurate, real-time and automated determination of seed germination status based on image processing technology. Summary of the Invention

[0004] Therefore, the technical problem to be solved by the present invention is to overcome the problems of low monitoring throughput, poor efficiency, reliance on manual observation leading to subjective data lag, single germination judgment dimension, difficulty in distinguishing between absorption, abnormal and real germination, and low recognition rate of tiny germination targets in the prior art.

[0005] To solve the above-mentioned technical problems, the present invention provides a seed germination monitoring method, comprising: A multi-layered three-dimensional culture space is set up to hold multiple culture containers containing the seeds to be tested; Using a movable imaging device, images are sequentially captured above each culture container according to a preset path and time interval. Real-time image analysis is performed on the acquired images to identify and determine the germination status of the seeds; The real-time image analysis includes: segmenting each seed in the image; constructing a four-dimensional morphological feature vector based on the segmented seed image, which includes convexity features, area features, aspect ratio features, and physical length features; determining whether the seed has germinated based on the four-dimensional morphological feature vector; and making a secondary determination in cases where the seed is determined to be ungerminated, suspected to be slightly germinated, or has an abnormal contour, combined with contour evolution analysis based on edge continuity. The contour evolution analysis based on edge continuity includes: performing sub-pixel level edge detection using a contour compression algorithm; identifying changes in the concavity filling rate of the seed surface by performing convex hull operation on the contour point set to capture slightly germinated targets and distinguish between seed damage and image noise.

[0006] Preferably, the movable imaging device is driven by a three-axis linkage mechanism to move in the horizontal X and Y axes and the vertical Z axis to cover all culture containers in the multi-layer three-dimensional culture space.

[0007] Preferably, determining whether a seed has germinated based on the four-dimensional morphological feature vector includes: Calculate the convexity feature value of the seed image, wherein the convexity feature value is the ratio of the actual contour area of ​​the seed to its convex hull area; When the convexity feature value is greater than the first preset convexity threshold, it is determined that it has not sprouted, and a second determination is made by contour evolution analysis based on edge continuity. When the convexity feature value is less than or equal to the second preset convexity threshold, the physical length feature value is obtained: If the physical length characteristic value exceeds the preset length threshold, it is determined that it has sprouted. If the physical length feature value does not exceed the preset length threshold, it is determined to be a suspected micro-germination or contour anomaly, and then proceeds to contour evolution analysis based on edge continuity for secondary determination. Wherein, the first preset convexity threshold is greater than the second preset convexity threshold; When the convexity feature value is less than or equal to the first preset convexity threshold and greater than the second preset convexity threshold, a comprehensive judgment is made by combining area features, aspect ratio features, and physical length features: If the area characteristic value of the current seed increases within the preset imbibition range compared to the baseline area of ​​ungerminated seeds of the same variety and batch, and the aspect ratio and physical length characteristic values ​​do not change uniformly, then it is determined to be seed imbibition. If the area feature value of the current seed increases while the aspect ratio feature value changes and the physical length feature value exceeds the preset length threshold, it is determined that it has germinated. If none of the above conditions are met, it is determined that it has not sprouted, and a second determination is made based on the contour evolution analysis of edge continuity.

[0008] Preferably, in cases where the bud is determined to be non-germinating, suspected to be micro-germinating, or has an abnormal contour, a secondary determination is performed using contour evolution analysis based on edge continuity, specifically including: For seeds that are preliminarily determined by the four-dimensional morphological feature vector to be ungerminated, suspected to be slightly germinated, or have abnormal contours, the Alpha channel of their binary mask is extracted, and a contour compression algorithm is used to perform sub-pixel level edge detection to obtain the seed contour point set, and noise contours with an area smaller than a preset area threshold are filtered out. Perform convex hull operation on the contour point set to calculate the concavity filling rate of the seed surface. The concavity filling rate is the ratio of the actual contour area of ​​the seed to its convex hull area. When the detected depression filling rate is less than the first preset depression threshold, it is determined to be a micro-germination target; When the detected depression filling rate is greater than or equal to the first preset depression threshold and less than the second preset depression threshold, the area ratio of the depression region and the rate of change of curvature of the depression region edge are further calculated: If the area ratio of the recessed region is less than a preset area ratio threshold, and the rate of change of curvature is less than a preset curvature threshold, then it is determined to be image noise or segmentation error, and the "no sprouting" determination is maintained. Preferably, after determining that the seed has germinated, the germination outline of the germinated seed is extracted, and based on the ratio of image pixels to physical size, at least one of the following physical indicators of the germinated seed is calculated: germination length, germination perimeter, and germination area; wherein, the germination length is calculated by extracting the distance of the farthest point of the radicle protrusion in the germination outline.

[0009] Preferably, the method employs a parallel processing flow: while the imaging device moves to the next acquisition position, the real-time image analysis is performed on the image acquired at the previous acquisition position.

[0010] The present invention also provides a seed germination monitoring device, comprising: A three-dimensional culture rack, which has multiple layers of culture space inside for placing culture containers; A camera device used to capture images of seeds in a culture container; A motion mechanism, connected to the imaging device, is used to drive the imaging device to move, so as to traverse the culture containers in each layer of the three-dimensional culture rack; A control and processing system is electrically connected to the shooting device and the motion mechanism, respectively, and the control and processing system is configured to control the motion mechanism and the shooting device to perform the steps of the method described above.

[0011] Preferably, the motion mechanism includes X-axis and Y-axis moving components for driving the shooting device to move in the horizontal plane, and Z-axis moving components for driving the shooting device to move in the vertical direction.

[0012] Preferably, the three-dimensional culture rack includes at least five layers of culture space, and each layer of culture space is provided with at least one row of placement positions for placing culture containers.

[0013] Preferably, the control and processing system includes a memory storing a computer program and a processor, which, when executed by the processor, implements the method steps described above.

[0014] The technical solution of the present invention has the following advantages compared with the prior art: The seed germination monitoring method described in this invention integrates a multi-layered, high-density, three-dimensional culture rack with a three-axis linkage automated imaging mechanism, enabling high-throughput parallel culture and monitoring of thousands of seed samples within a limited experimental space, significantly improving space utilization and experimental scale. Employing a "shoot-and-calculate" parallel processing mode, image acquisition and algorithm analysis are performed simultaneously, significantly shortening the single-round detection cycle and achieving high-frequency (e.g., every 5 minutes) automatic inspection of the entire sample, ensuring data timeliness. Furthermore, this invention constructs a four-dimensional morphological feature vector including convexity, area, aspect ratio, and physical length, achieving multi-dimensional and highly robust determination of seed germination status, effectively distinguishing between seed abnormalities, imbibition, and true germination. In particular, by introducing contour evolution analysis based on edge continuity for secondary determination, and utilizing changes in depression filling rate to accurately capture tiny germination targets (such as small seeds like rapeseed and tobacco), it significantly improves the detection sensitivity and accuracy of early germination "whitening" phenomena, completely eliminating subjective errors from manual observation, and providing traceable and quantifiable precise data support for breeding research and vigor detection. The entire system is compact and highly automated, which significantly reduces labor costs and improves the reliability and repeatability of experimental results. Attached Figure Description

[0015] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart illustrating the implementation of a seed germination monitoring method provided by the present invention; Figure 2 This is a schematic diagram of a movable shooting device provided in one embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a seed germination monitoring device provided by the present invention. Detailed Implementation

[0016] The core of this invention is to provide a seed germination monitoring method and device, which effectively realizes high-throughput, automated, real-time, and accurate monitoring and germination status determination of high-density three-dimensional cultured seeds.

[0017] To enable those skilled in the art to better understand the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please refer to Figure 1. Figure 1 The flowchart illustrates the implementation of a seed germination monitoring method provided by this invention; the specific operation steps are as follows: S101: A multi-layered three-dimensional culture space is set up to hold multiple culture containers containing the seeds to be tested; S102: Using a movable imaging device, images are sequentially captured above each culture container according to a preset path and time interval. S103: Perform real-time image analysis on the acquired images to identify and determine the germination status of the seeds; The real-time image analysis includes: segmenting each seed in the image; constructing a four-dimensional morphological feature vector based on the segmented seed image, which includes convexity features, area features, aspect ratio features, and physical length features; determining whether the seed has germinated based on the four-dimensional morphological feature vector; and making a secondary determination in cases where the seed is determined to be ungerminated, suspected to be slightly germinated, or has an abnormal contour, combined with contour evolution analysis based on edge continuity. The contour evolution analysis based on edge continuity includes: performing sub-pixel level edge detection using a contour compression algorithm; identifying changes in the concavity filling rate of the seed surface by performing convex hull operation on the contour point set to capture slightly germinated targets and distinguish between seed damage and image noise.

[0019] Based on the above embodiments, this embodiment will provide a detailed description of step S101: In some embodiments, the multi-layer three-dimensional culture space is composed of a multi-layer three-dimensional culture frame, with each layer having at least one row of guide rails or slots for placing standard culture trays. This modular design facilitates the placement and handling of culture containers and batch operations, and the number of layers can be flexibly adjusted according to experimental needs.

[0020] In other embodiments, the culture container is not limited to a standard germination tray, but may also be a petri dish, a multi-well plate or other culture carrier suitable for seed germination, and its material may be transparent or translucent to facilitate image acquisition from above.

[0021] In one specific embodiment, the three-dimensional culture rack is constructed with an aluminum alloy frame, making it lightweight yet robust. The rack has five layers of culture space, with two rows of placement positions arranged in parallel on each layer. Each row can accommodate 13 standard germination trays, and each germination tray can hold 30 seeds. Therefore, this device can monitor up to 3900 seeds simultaneously in a single experiment, achieving high-density three-dimensional culture and high-throughput monitoring.

[0022] Specifically, the height of each layer of the cultivation space is specially designed to ensure that when the imaging device moves vertically, its lens maintains the optimal imaging distance with the surface of the seeds in the cultivation container, thereby ensuring the consistency and clarity of the acquired images.

[0023] It should be noted that the term "multi-layered" typically refers to three or more layers to significantly improve space utilization. Through this three-dimensional layout, the present invention substantially increases the seed monitoring capacity per unit area within a limited laboratory footprint, overcoming the low throughput limitations of traditional planar or single-layer culture methods. This design is a crucial structural foundation for the high-throughput monitoring capabilities of this device, and its beneficial effects include, but are not limited to: significantly improved experimental efficiency, and a feasible hardware platform for large-scale germplasm resource screening, genetic research, or seed quality assessment.

[0024] Based on the above embodiments, this embodiment will provide a detailed description of step S102: like Figure 2 In one embodiment, the movable imaging device is driven by a three-axis linkage mechanism to move in the horizontal X and Y axes and the vertical Z axis to cover all culture containers in the multi-layer three-dimensional culture space.

[0025] In some embodiments, the three-axis linkage mechanism includes: The combination of horizontal guide rails and drive motors on the main body of the device is used to drive the imaging device to move along the X-axis (lateral) to cover multiple culture trays in a single row within a single layer; The interlayer switching mechanism, which is linked to the horizontal guide rail, is used to drive the imaging device to move along the Z-axis (longitudinal direction) to achieve switching between different culture layers; The vertical lifting module mounted on the horizontal moving mechanism is used to drive the imaging device to move along the Y-axis to adjust the shooting focal length or cover multiple rows of culture containers in a single layer.

[0026] In other embodiments, the driving method may also be implemented in other ways, such as linear motor, synchronous belt drive or ball screw mechanism, as long as the shooting device can be accurately and stably positioned in three-dimensional space.

[0027] In one specific embodiment, the vertical lifting module employs a chain drive structure to drive a "photo-taking box" integrating a camera and lighting unit to move up and down. This design is compact, operates smoothly, and is easy to maintain.

[0028] Specifically, the control system generates a sequence of target coordinates for the X, Y, and Z axes according to a preset path planning algorithm, drives the motor to perform corresponding actions, and enables the imaging device to efficiently and completely traverse all monitoring points.

[0029] It should be noted that "covering all culture containers" in this invention means that the field of view of the imaging device can clearly capture the seeds in each designated culture container. It does not require a single shot to cover all containers; full coverage can be achieved through multiple repositioning movements. "Three-axis linkage" refers to the ability to control movement in three directions independently or to coordinate actions according to program instructions. This embodiment, through a three-dimensional movement scheme, replaces the traditional fixed multi-camera array with a single high-performance imaging device. This significantly reduces hardware costs and complexity while achieving flexible and efficient monitoring of three-dimensional, high-density culture layouts, which is key to improving the system's economy and reliability.

[0030] Based on the above embodiments, in some embodiments, the shooting device integrates a ring-shaped LED fill light unit, which is embedded in the "photo-taking box" and arranged around the camera lens to provide uniform and controllable lighting during shooting, so as to reduce shadows and ensure the consistency of image quality.

[0031] Furthermore, the control and processing system can automatically adjust the brightness and color temperature of the supplementary lighting unit according to the ambient light intensity to adapt to the imaging needs of different experimental conditions or seed types.

[0032] Based on the above embodiments, this embodiment will provide a detailed description of step S103: In one embodiment, a trained semantic segmentation model (such as a variant of U-Net) is used to perform background segmentation and single-seed localization on the acquired image, which can accurately separate the seed region in the culture tray image from the background (such as the culture medium and the tray body).

[0033] In one embodiment, determining whether a seed has germinated based on the four-dimensional morphological feature vector includes: Calculate the convexity feature value of the seed image, wherein the convexity feature value is the ratio of the actual contour area of ​​the seed to its convex hull area; Case 1: When the convexity feature value is greater than the first preset convexity threshold, it is determined that it has not sprouted, and then proceeds to the subsequent contour evolution analysis based on edge continuity for secondary determination.

[0034] Scenario 2: When the convexity feature value is less than or equal to the second preset convexity threshold, it indicates that the seed contour has undergone significant distortion. In this case, the physical length feature value is obtained: If the physical length characteristic value exceeds the preset length threshold, it is directly determined that it has sprouted; If the physical length feature value does not exceed the preset length threshold, it is determined to be a suspected micro-germination or contour abnormality. In this case, it will also enter the contour evolution analysis based on edge continuity for secondary determination to distinguish micro-germination, seed damage or image noise. The first preset convexity threshold is greater than the second preset convexity threshold.

[0035] Scenario 3: When the convexity feature value is less than or equal to the first preset convexity threshold and greater than the second preset convexity threshold, a comprehensive judgment is made by combining the area feature, aspect ratio feature, and physical length feature of the current image: If the area characteristic value of the current seed increases within the preset imbibition range compared to the baseline area of ​​the same variety and batch of ungerminated seeds, and the aspect ratio characteristic value and physical length characteristic value do not change significantly (i.e. the seed still maintains an approximately elliptical or circular outline with no obvious local protrusion), then it is determined to be an increase in volume caused by seed imbibition rather than germination. If the area characteristic value of the current seed increases while the aspect ratio characteristic value changes significantly (such as the ratio of the major axis to the minor axis increases) and the physical length characteristic value exceeds the preset length threshold, then it is determined that it has germinated. If none of the above conditions are met, it is determined that it has not sprouted, and a second determination is made based on the contour evolution analysis of edge continuity.

[0036] It should be noted that seeds undergo an imbibition process before germination, during which their overall volume increases while their shape remains regular. Although their convexity value decreases slightly, it is still usually higher than a first preset convexity threshold. This invention, by introducing physical length (representing the maximum protrusion distance) and aspect ratio features, and combining them with baseline morphological data of ungerminated seeds from the same batch, can effectively distinguish between "overall volume increase caused by imbibition" and "non-uniform geometric growth caused by radicle breakthrough," thereby avoiding misjudgment in this scenario using a traditional single convexity threshold.

[0037] In some embodiments, the first preset convexity threshold and the second preset convexity threshold can be set and optimized through training with historical data or empirical values. For example, for common crop seeds, the first preset convexity threshold can be set to 0.92, and the second preset convexity threshold can be set to 0.90. The preset length threshold is set according to the characteristics of the seed variety; for example, it can be set to 0.5 mm for rapeseed seeds and 1.0 mm for rice seeds.

[0038] It should be further explained that the setting of the second preset convexity threshold of 0.90 has clear biological benchmarks and experimental basis: (1) Biological benchmark: Before germination, the seed coat is in a highly regular state, and its outline area almost coincides with the convex area, with a convexity value close to 1.0. When the radicle breaks through the seed coat, the seed outline shows obvious local distortion, which causes its "solid area" to be unable to fill its "envelope area", and the convexity value decreases accordingly.

[0039] (2) Basis for threshold setting: Through experimental verification on various crop seeds, 0.90 is the critical point for distinguishing between "imbibition morphology" and "germination morphology". During the imbibition process, the volume of ungerminated seeds increases as a whole, but the shape remains regular, and the convexity value is usually maintained above 0.92; while when the radicle breaks through the seed coat, the convexity value will drop below 0.90. Therefore, when the convexity value drops from 0.98 to below 0.90, it marks the completion of the qualitative change from physiological dormancy to growth and development of the seed.

[0040] (3) Robustness Guarantee: In one embodiment, the system employs a deep learning background removal algorithm (such as rembg) for foreground segmentation during the image preprocessing stage, effectively eliminating the interference of illumination and shadows on contour extraction before determining convexity. Simultaneously, by introducing baseline morphological data of ungerminated seeds from the same batch for normalization comparison, the impact of different imaging conditions on threshold judgment is further reduced. Therefore, the threshold setting exhibits high robustness under different illumination intensities and different experimental batches, requiring no frequent manual intervention.

[0041] In one embodiment, the secondary determination, based on contour evolution analysis according to edge continuity, is performed when the germination is determined to be non-germination, suspected micro-germination, or contour anomaly. This specifically includes: For seeds that are preliminarily determined to be ungerminated based on the four-dimensional morphological feature vector, the Alpha channel of their binary mask is extracted, and a contour compression algorithm (such as the Douglas-Peucker algorithm or cv2.CHAIN_APPROX_SIMPLE) is used to perform sub-pixel level edge detection to obtain the seed contour point set, and noise contours with an area smaller than a preset area threshold (such as 10 pixels) are filtered out.

[0042] Perform convex hull operation on the contour point set to calculate the concavity filling rate of the seed surface. The concavity filling rate is the ratio of the actual contour area of ​​the seed to the convex hull area. The microscopic characterization of this ratio in the local concave region is as follows: when there are tiny protrusions on the seed surface (such as radicle breakthrough), the actual contour cannot fill the convex hull, resulting in a significant decrease in this ratio. When the detected depression filling rate is less than the first preset depression threshold (e.g., 0.85), it indicates that there are significant micro-protrusions or local distortions on the seed surface, which are identified as micro-germination targets. Then, the germination index measurement process (extracting germination contours, calculating germination length, etc.) is initiated. When the detected depression filling rate is greater than or equal to the first preset depression threshold and less than the second preset depression threshold (e.g., 0.85~0.92), it indicates that the seed contour has slight irregularities and further quantitative judgment is required. First, extract the concave region (i.e., the area of ​​difference between the convex hull contour and the actual contour) and calculate the area ratio of the concave region. That is, the ratio of the area of ​​the concave region to the actual outline area of ​​the seed; Secondly, the rate of change of curvature at the edge of the depression is calculated to characterize the smoothness of the depression edge. The formula for calculating the rate of change of curvature is:

[0043] in, Let N be the average curvature of the edge, and N be the total number of pixels on the edge. The greater the rate of change of curvature, the more irregular the edge is, indicating the presence of jagged breaks or sharp corners.

[0044] The judgment is based on the above quantitative indicators: like <Preset area percentage threshold (e.g., 5%) (the recessed area has a small area percentage) and If the pre-set curvature threshold (e.g., 0.3) is low (low curvature change rate, smooth edges), it indicates that the depression is a small noise or edge fluctuation generated during image acquisition or segmentation, rather than a biological change in the seed itself. It is judged as image noise or segmentation error, the non-germination judgment is maintained, and the seed is marked as "requires manual review". like ≥Preset area percentage threshold (e.g., 5%) and If the indentation rate is greater than or equal to the preset curvature threshold (e.g., 0.3) (the area of ​​the indentation is relatively large and the edges show jagged breaks or sharp corners), it indicates that there is non-physiological structural damage on the seed surface, which is judged as seed damage or mechanical damage, and marked as "abnormal seed", and is not included in the germination statistics. When the indentation filling rate is detected to be greater than or equal to the second preset indentation threshold (e.g., 0.92), it indicates that the seed outline is regular and there is no obvious local indentation, and the non-germination judgment is maintained.

[0045] It should be noted that the first preset concavity threshold, the second preset concavity threshold, the preset area ratio threshold, and the preset curvature threshold can be calibrated according to the seed variety and imaging resolution. For small seeds such as rapeseed and tobacco, the first preset concavity threshold can be appropriately relaxed to 0.82; for large seeds such as rice and corn, it can be tightened to 0.88. The calculation of the concavity filling rate complements the overall convexity feature—convexity reflects the regularity of the global contour, while the concavity filling rate focuses on local minute deformations. The combination of the two can effectively distinguish between "minor germination," "seed damage," and "image noise."

[0046] In one embodiment, the method further includes: after determining that the seed has germinated, extracting the germination outline of the germinated seed, and calculating at least one of the following physical indicators of the germinated seed based on the ratio of image pixels to physical size: germination length, germination perimeter, and germination area; wherein the germination length is calculated by extracting the distance of the farthest point of the radicle protrusion in the germination outline.

[0047] Specifically, after determining that a seed has germinated, the germination contour of the germinated seed is extracted, and a sequence of all continuous boundary pixels on the contour is obtained. The total pixel perimeter of the seed contour is obtained by summing the Euclidean distances between adjacent pixels in the sequence. Further, according to preset calibration parameters (i.e., the actual physical size corresponding to a unit pixel in the image), the total pixel perimeter is converted into the actual physical perimeter of the germinated seed, thereby achieving automated measurement of its morphological scale. For calculating the germination length, the maximum Euclidean distance between the farthest point of the radicle protrusion and the main contour of the seed is calculated and converted into a physical length by identifying the farthest point of the radicle protrusion.

[0048] Specifically, the calibration parameters are obtained by taking images of a calibration plate of known size to ensure the accuracy of length measurement.

[0049] It should be noted that "convex hull" refers to the smallest convex polygon containing all points of the seed outline; "convexity" is a dimensionless index describing how closely the outline approximates a convex polygon, with a value closer to 1 indicating a more convex and regular outline; and "depression filling rate" is a refined representation of convexity in local micro-regions. This embodiment, by constructing a four-dimensional morphological feature vector and combining it with outline evolution analysis, achieves multi-scale germination determination from macro to micro levels, effectively overcoming the misjudgment problems caused by seed imbibition, uneven light exposure, or low contrast between the radicle and seed coat in traditional methods, and significantly improving the robustness and accuracy of state classification.

[0050] Based on the above embodiments, in some embodiments, the method further includes a system calibration step when running for the first time or when changing the type of culture container. This step involves driving an imaging device to photograph a calibration plate (placed at the location of the culture container) with regularly arranged known-size marks, automatically calculating and storing the conversion relationship between image pixels and actual physical dimensions (such as the PIXELS_PER_MM parameter), for subsequent calculation of the physical length of the shoot.

[0051] In other embodiments, the initialization step further includes a mechanical homing operation, that is, driving the three-axis motion mechanism to move to a preset mechanical origin, which serves as the reference for all subsequent movement coordinates to ensure positioning accuracy.

[0052] Based on the above embodiments, in one embodiment, the result data generated by the real-time image analysis step includes seed ID (which can be derived from location coordinates), germination status, convexity value, physical length value, depression filling rate, and corresponding timestamp, which are written in real time to a structured log file (such as CSV format) and synchronously stored in a local or cloud database.

[0053] In one specific embodiment, the control and processing system is also equipped with a visual user interface for real-time display of the current shooting location, the germination rate statistical curve of the analyzed seeds, and the analysis image with contour and convex hull annotations, so as to facilitate users to monitor the experimental process.

[0054] Based on the above embodiments, in some embodiments, the method further includes an exception handling process: if the image analysis algorithm fails to successfully segment any seed contour at a certain position, the system will record the position as "recognition failure" in the log, and can choose to drive the shooting device to re-shoot at the position; if the motion mechanism encounters resistance or exceeds the preset position during movement, the system will trigger an emergency stop and alarm to prevent equipment damage.

[0055] Based on the above embodiments, the method adopts a parallel processing flow: while the shooting device moves to the next acquisition position, the real-time image analysis is performed on the image acquired at the previous acquisition position.

[0056] In some embodiments, the parallel processing flow is implemented through multi-threading or multi-processing technology, wherein the first thread or process is responsible for controlling the coordinated movement and image acquisition of the shooting device and the motion mechanism, the second thread or process is responsible for executing the image analysis algorithm, and the two threads or processes exchange and synchronize data through shared memory or message queues.

[0057] In other embodiments, the parallel processing flow can also be implemented using asynchronous I / O combined with an event-driven model. After the image acquisition event is triggered, the system immediately sends the image data into the processing queue and then controls the shooting device to move again. The image analysis task is retrieved from the queue and processed by the background worker thread, thereby realizing the pipelined operation of acquisition and calculation.

[0058] In one specific embodiment, after system initialization, the control thread drives the imaging device to move directly above the first culture tray and take a picture. At the moment the picture is taken, the control thread immediately issues a command to make the motion mechanism start moving to the next tray position, and at the same time sends the acquired image data packet to the processing queue. The analysis thread continuously listens to the queue, and once a new image is acquired, it immediately calls the image recognition algorithm to process it and writes the result to the database. This process is repeated until all tray positions are traversed, completing one round of panoramic monitoring.

[0059] Specifically, to ensure consistency between processing timing and data, the system can attach metadata to each acquired image, including but not limited to the capture timestamp and the spatial coordinates of the culture tray (such as layer number, row number, and tray position number). After processing, the analysis thread will associate the judgment results with the same metadata with the measurement data and store them together. This design allows for accurate data alignment and traceability in subsequent processing even if the processing speed is slightly slower than the acquisition speed.

[0060] It should be noted that the "parallel processing flow" emphasizes the overlapping use of "movement time" and "computation time," with the core being to reduce system idle waiting, thereby significantly improving overall monitoring efficiency under conditions of limited hardware speed. The terms "threads," "processes," "asynchronous I / O," "event-driven," and "queues" are all common techniques for achieving concurrency in the computer field, and those skilled in the art can select and adapt them according to specific hardware architectures and software platforms.

[0061] This embodiment introduces a parallel processing flow, which can make full use of the time inevitably consumed by mechanical movement for image calculation. This makes the total time of a single-wheel monitoring of the system approximately equal to the time it takes for the mechanical movement to traverse all points, without the need to add the processing time of all images. This provides a key technical guarantee for achieving high-frequency (e.g., once every 5 minutes) continuous automatic monitoring, and significantly improves the system's throughput and time resolution.

[0062] like Figure 3 The present invention also provides a seed germination monitoring device, comprising: A three-dimensional culture rack, which has multiple layers of culture space inside for placing culture containers; A camera device used to capture images of seeds in a culture container; A motion mechanism, connected to the imaging device, is used to drive the imaging device to move, so as to traverse the culture containers in each layer of the three-dimensional culture rack; A control and processing system is electrically connected to the shooting device and the motion mechanism, respectively, and the control and processing system is configured to control the motion mechanism and the shooting device to perform the steps of the method described above.

[0063] In some embodiments, the motion mechanism includes X-axis and Y-axis motion components for driving the shooting device to move in a horizontal plane, and Z-axis motion components for driving the shooting device to move in a vertical direction.

[0064] In other embodiments, the three-dimensional culture rack includes at least five layers of culture space, each layer of culture space having at least one row of placement positions for placing culture containers.

[0065] In one specific embodiment, the control and processing system includes a memory storing a computer program and a processor, which, when executed by the processor, implements the method steps described above.

[0066] Specifically, the Z-axis moving component can be implemented using any of the following methods: chain drive, lead screw drive, or linear motor, to precisely adjust the focusing height of the imaging device relative to the culture container.

[0067] Based on the above embodiments, in one specific embodiment, the three-dimensional culture rack can be placed entirely in a controllable environment (such as an artificial climate chamber), and the control and processing system can also communicate with the temperature, humidity, and light controller of the environment to realize integrated automatic management of culture conditions and monitoring processes, ensuring the consistency of experimental conditions.

[0068] Based on the above embodiments, each layer of the three-dimensional culture rack has a limiting structure for each guide rail or slot, ensuring that the culture tray will not shift position under slight vibrations caused by multiple shooting movements.

[0069] In other embodiments, a transparent observation window or door may be provided on the main frame to facilitate manual spot checks or maintenance without interrupting automated monitoring.

[0070] It should be noted that the "culture container" described in this invention includes, but is not limited to, standard germination trays, petri dishes, or customized carriers; "real-time image analysis" covers the entire process from image acquisition to result output, and its time delay mainly depends on the processing time of a single image, achieving near real-time analysis under a parallel system architecture. This embodiment, through the synergy of mechanical structure and control algorithm, achieves high-density, automated, and non-invasive seed germination monitoring, significantly improving monitoring throughput, data objectivity, and experimental efficiency. Those skilled in the art should understand that the embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0071] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0072] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0073] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0074] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for monitoring seed germination, characterized in that, include: A multi-layered three-dimensional culture space is set up to hold multiple culture containers containing the seeds to be tested; Using a movable imaging device, images are sequentially captured above each culture container according to a preset path and time interval. Real-time image analysis is performed on the acquired images to identify and determine the germination status of the seeds; The real-time image analysis includes: Each seed in the image is segmented individually, and a four-dimensional morphological feature vector containing convexity features, area features, aspect ratio features, and physical length features is constructed based on the segmented seed images. Determining whether a seed has germinated based on the aforementioned four-dimensional morphological feature vector includes: Calculate the convexity feature value of the seed image, wherein the convexity feature value is the ratio of the actual contour area of ​​the seed to its convex hull area; When the convexity feature value is greater than the first preset convexity threshold, it is determined that it has not sprouted, and a second determination is made by contour evolution analysis based on edge continuity. When the convexity feature value is less than or equal to the second preset convexity threshold, the physical length feature value is obtained: If the physical length characteristic value exceeds the preset length threshold, it is determined that it has sprouted. If the physical length feature value does not exceed the preset length threshold, it is determined to be a suspected micro-germination or contour anomaly, and then proceeds to contour evolution analysis based on edge continuity for secondary determination. Wherein, the first preset convexity threshold is greater than the second preset convexity threshold; When the convexity feature value is less than or equal to the first preset convexity threshold and greater than the second preset convexity threshold, a comprehensive judgment is made by combining area features, aspect ratio features, and physical length features: If the area characteristic value of the current seed increases within the preset imbibition range compared to the baseline area of ​​ungerminated seeds of the same variety and batch, and the aspect ratio characteristic value and physical length characteristic value do not change non-uniformly, then it is determined to be seed imbibition. If the area feature value of the current seed increases while the aspect ratio feature value changes and the physical length feature value exceeds the preset length threshold, it is determined that it has germinated. If none of the above conditions are met, it is determined that it has not sprouted and enters the contour evolution analysis based on edge continuity for a second determination; In cases where the bud is determined to be non-germinating, suspected to be slightly germinating, or has an abnormal contour, a secondary determination is made based on contour evolution analysis according to edge continuity, specifically including: For seeds that are preliminarily determined by the four-dimensional morphological feature vector to be ungerminated, suspected to be slightly germinated, or have abnormal contours, the Alpha channel of their binary mask is extracted, and a contour compression algorithm is used to perform sub-pixel level edge detection to obtain the seed contour point set, and noise contours with an area smaller than a preset area threshold are filtered out. Perform convex hull operation on the contour point set to calculate the concavity filling rate of the seed surface. The concavity filling rate is the ratio of the actual contour area of ​​the seed to its convex hull area. When the detected depression filling rate is less than the first preset depression threshold, it is determined to be a micro-germination target; When the detected depression filling rate is greater than or equal to the first preset depression threshold and less than the second preset depression threshold, the area ratio of the depression region and the rate of change of curvature of the depression region edge are further calculated: If the area ratio of the recessed region is less than a preset area ratio threshold, and the rate of change of curvature is less than a preset curvature threshold, then it is determined to be image noise or segmentation error, and the "no sprouting" determination is maintained. If the area ratio of the recessed region is greater than or equal to a preset area ratio threshold, and the rate of change of curvature is greater than or equal to a preset curvature threshold, then it is determined to be seed damage or mechanical damage and marked as an abnormal seed. When the detected depression filling rate is greater than or equal to the second preset depression threshold, the no-germination determination is maintained. The contour evolution analysis based on edge continuity includes: using a contour compression algorithm to perform sub-pixel level edge detection, and performing convex hull operation on the contour point set to identify changes in the concavity filling rate of the seed surface in order to capture tiny germination targets and distinguish between seed damage or image noise.

2. The method according to claim 1, characterized in that, The movable imaging device is driven by a three-axis linkage mechanism to move in the horizontal X and Y axes and the vertical Z axis to cover all culture containers in the multi-layer three-dimensional culture space.

3. The method according to claim 1, characterized in that, The method further includes: after determining that the seed has germinated, extracting the germination outline of the germinated seed, and calculating at least one of the following physical indicators of the germinated seed based on the ratio of image pixels to physical size: germination length, germination perimeter, and germination area; wherein the germination length is calculated by extracting the distance of the farthest point of the radicle protrusion in the germination outline.

4. The method according to claim 1, characterized in that, The method employs a parallel processing flow: while the imaging device moves to the next acquisition position, the real-time image analysis is performed on the image acquired at the previous acquisition position.

5. A seed germination monitoring device, characterized in that, include: A three-dimensional culture rack, which has multiple layers of culture space inside for placing culture containers; A camera device used to capture images of seeds in a culture container; A motion mechanism, connected to the imaging device, is used to drive the imaging device to move, so as to traverse the culture containers in each layer of the three-dimensional culture rack; A control and processing system is electrically connected to the shooting device and the motion mechanism, respectively, and the control and processing system is configured to control the motion mechanism and the shooting device to perform the steps of the method as described in any one of claims 1 to 4.

6. The apparatus according to claim 5, characterized in that, The motion mechanism includes X-axis and Y-axis moving components for driving the shooting device to move in the horizontal plane, and Z-axis moving components for driving the shooting device to move in the vertical direction.

7. The apparatus according to claim 5, characterized in that, The three-dimensional culture rack includes at least five layers of culture space, and each layer of culture space is provided with at least one row of placement positions for placing culture containers.

Images