Pollen viability detection method and device, computer equipment, medium and program product

Abstract

The invention relates to the technical field of electronics, and discloses a pollen viability detection method and device, computer equipment, a medium and a program product. The pollen viability detection method comprises the following steps: identifying viability type information of each single pollen in a first target image of to-be-detected pollen through a pre-constructed pollen viability detection model; according to the method, the activity information of the to-be-detected pollen is determined according to the activity type information of the different single-grain pollen, so that rapid and accurate detection of the pollen activity is realized, and the problems of relatively low detection efficiency and precision of pollen activity detection through a traditional pollen activity detection method in related technologies are solved.

Description

Technical Field

[0001] The present invention relates to the field of electronic technology, and in particular to a pollen vitality detection method, device, computer equipment, medium and program product. Background Art

[0002] Pollen viability, a key indicator of reproductive capacity and genetic potential, is directly related to a crop's pollination success rate, seed set rate, and ultimately yield. In modern agricultural practice, the widespread use of advanced pollination technologies, such as drone pollination and rope pollination, has improved pollination efficiency to a certain extent. However, insufficient pollen viability hinders the full potential of these modern pollination techniques, resulting in suboptimal seed set rates and, consequently, impacting overall corn yield and quality. Therefore, strengthening pollen viability monitoring and assessment is crucial for improving breeding capabilities, fully utilizing modern pollination technologies, and ensuring stable and efficient corn production.

[0003] Related technologies use traditional pollen testing methods, such as triphenyltetrazolium chloride (TTC) staining, acetic acid carmine staining, I2-KI staining, and pollen tube germination, to process pollen. This processed pollen is then manually analyzed to determine pollen viability. While these traditional pollen viability testing methods can reflect pollen activity to a certain extent, they are generally difficult to interpret due to cumbersome procedures, time-consuming processes, and high subjectivity. These methods make it difficult to meet the urgent need for efficient and precise management in modern agriculture. Summary of the Invention

[0004] In view of this, the present invention provides a pollen viability detection method, apparatus, computer equipment, medium and program product to solve the problem of low efficiency and accuracy in manually interpreting the results of traditional pollen viability detection in the related art.

[0005] In a first aspect, the present invention provides a pollen vitality detection method, which includes: obtaining a first target image of the pollen to be detected; inputting the first target image of the pollen to be detected into a pre-constructed pollen vitality detection model so that the pollen vitality detection model outputs a second target image, wherein the second target image contains vitality type information of each single pollen grain in the first target image; and determining the vitality information of the pollen to be detected based on the type information of each single pollen grain.

[0006] The pollen viability detection method provided by the present invention uses a pre-constructed pollen viability detection model to identify the viability type information of each single pollen grain in the first target image of the pollen to be detected, and determines the viability information of the pollen to be detected based on the viability type information of different single pollen grains, thereby realizing rapid and accurate detection of pollen viability, and solving the problems of low detection efficiency and accuracy in pollen viability detection performed by traditional pollen viability detection methods in related technologies.

[0007] In an optional embodiment, a pollen viability detection model is constructed by the following steps: obtaining a first image of multiple pollen samples and a second image of each first image, the second image containing the viability type information of each single pollen grain in the corresponding first image; associating the first image and the second image of each pollen sample to obtain an associated data set; using the associated data set to train a preset target detection model until the model accuracy meets the preset conditions, thereby obtaining a pollen viability detection model.

[0008] The method provided in this optional embodiment associates the first image and the second image of each pollen sample to construct an associated data set, and uses the associated data set to train a preset target detection model to obtain a pollen viability detection model. The pollen viability detection model can obtain a first image based on the pollen to be detected and output a second image containing the viability type information of each single pollen grain, thereby realizing rapid and accurate detection of pollen viability.

[0009] In an optional embodiment, the step of determining the vitality information of the pollen to be tested based on the type information of each single pollen grain includes: determining the number of single pollen grains of each vitality type based on the vitality type information corresponding to different single pollen grains; substituting the number of single pollen grains corresponding to different vitality types into a pre-constructed pollen population vitality evaluation model for calculation to obtain the vitality information of the pollen to be tested.

[0010] The method provided by this optional embodiment obtains the vitality information of the pollen to be tested by substituting the number of single pollen grains corresponding to different vitality types in the pollen to be tested into a pre-built pollen population vitality assessment model for calculation, thereby achieving accurate assessment of the vitality of the pollen population.

[0011] In an optional embodiment, the step of constructing the pollen vitality detection model also includes: obtaining a test set, which contains multiple image data to be tested; inputting each image data to be tested into the pollen vitality detection model for testing, and obtaining the vitality type information of each single pollen grain in the corresponding image data to be tested; and evaluating the detection accuracy of the pollen vitality detection model based on the vitality type information of each single pollen grain in each image data to be tested.

[0012] In an optional embodiment, the detection accuracy of the pollen vitality detection model is evaluated based on the vitality type information of each single pollen grain in each image data to be tested, including: calculating the index values ​​corresponding to different preset evaluation indicators based on the vitality type information of each single pollen grain in each image data to be tested; and evaluating the detection accuracy of the pollen vitality detection model based on the index values ​​corresponding to different preset evaluation indicators.

[0013] In an optional embodiment, the preset target detection model includes a backbone network, a neck network and a detection head; the backbone network is used to extract features from the input first image to obtain feature maps of different levels, the neck network is used to perform feature fusion on feature maps of different levels to obtain a fused feature map, and the detection head is used to output detection results based on the fused feature map.

[0014] In a second aspect, the present invention provides a pollen vitality detection device, which includes: an acquisition module for acquiring a first target image of pollen to be detected; a detection module for inputting the first target image of pollen to be detected into a pre-built pollen vitality detection model, so that the pollen vitality detection model outputs a second target image, wherein the second target image contains vitality type information of each single pollen grain in the first target image; and a determination module for determining the vitality information of the pollen to be detected based on the type information of each single pollen grain.

[0015] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the pollen vitality detection method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0016] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the pollen viability detection method of the first aspect or any corresponding embodiment thereof.

[0017] In a fifth aspect, the present invention provides a computer program product comprising computer instructions for causing a computer to execute the pollen viability detection method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 is a schematic flow chart of a pollen viability detection method according to an embodiment of the present invention; Figure 2 is a flow chart of another pollen viability detection method according to an embodiment of the present invention; Figure 3 2. It is a schematic diagram of the network structure of YOLOV10-CBAM according to an embodiment of the present invention; Figure 4 is a structural block diagram of a pollen vitality detection device according to an embodiment of the present invention; Figure 5 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0021] Related technologies use traditional pollen testing methods, such as triphenyltetrazolium chloride (TTC) staining, acetic acid carmine staining, I2-KI staining, and pollen tube germination, to process pollen. This processed pollen is then manually analyzed to determine pollen viability. While these traditional pollen viability testing methods can reflect pollen activity to a certain extent, they are generally difficult to interpret due to cumbersome procedures, time-consuming processes, and high subjectivity. These methods make it difficult to meet the urgent need for efficient and precise management in modern agriculture.

[0022] In view of this, the pollen viability detection method provided in the embodiments of the present application can be applied to a server to detect pollen viability. The method provided in the embodiments of the present application uses a pre-built pollen viability detection model to identify the viability type information of each individual pollen grain in the first target image of the pollen to be detected. The viability information of the pollen to be detected is determined based on the viability type information of different individual pollen grains, thereby achieving rapid and accurate detection of pollen viability. This solves the problem of low detection efficiency and accuracy of pollen viability detection using traditional pollen viability detection methods in the related art.

[0023] According to an embodiment of the present invention, an embodiment of a pollen viability detection method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0024] In this embodiment, a pollen vitality detection method is provided, which can be used in the above-mentioned server. Figure 1 FIG. 1 is a flow chart of a pollen viability detection method according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps: Step S101: Acquire a first target image of pollen to be detected.

[0025] For example, the pollen to be tested may be a pollen sample requiring pollen viability testing. In this embodiment, corn plants that are healthy, pest-free, and have begun to shed pollen at one-third of the tassel are selected as sampling plants. 0.01g of pollen is collected as a pollen sample for pollen viability testing. The first target image may be a microscopic image of the pollen obtained by processing the pollen sample to be tested using a preset pollen testing method. The preset pollen testing method may include, but is not limited to, triphenyltetrazolium chloride (TTC) staining. TTC staining is used for pollen viability testing and has been proven to be the simplest, fastest, and most accurate traditional pollen viability testing method. A small amount of dye solution is pipetted onto a glass slide using a plastic dropper, and a microscopic image of the stained pollen is collected using a microscope system. Microscopic images of the stained pollen are collected using a portable microscope system. Three images are taken in three different fields of view, corresponding to the pollen image data of the pollen sample.

[0026] Step S102: inputting a first target image of pollen to be detected into a pre-built pollen vitality detection model, so that the pollen vitality detection model outputs a second target image, wherein the second target image contains vitality type information of each single pollen grain in the first target image.

[0027] Exemplarily, the pollen viability detection model is used to detect the viability type information of each individual pollen grain in a first target image of the pollen to be detected, and output second target image data. The viability type information is used to characterize the viability status of the individual pollen grains. In this embodiment of the present application, the viability types of individual pollen grains may include, but are not limited to, four types: high-vigor pollen (HVP), low-vigor pollen (LVP), non-viable pollen (NVP), and cracked pollen (CP). Specifically, in a microscopic image of stained pollen, HVP appears as a red circle, LVP appears as a pink circle, NVP appears as a colorless circle, and CP appears as a green irregular shape.

[0028] Step S103: determining the vitality information of the pollen to be detected based on the type information of each single pollen grain.

[0029] For example, vitality information can be used to assess the vitality of the pollen population of the pollen to be tested. In this embodiment of the present application, the ratio of vigorous pollen to the total pollen population can be calculated based on the number of high-vitality pollen and low-vitality pollen and the total number of single pollen grains, and this ratio can be used to assess the vitality of the pollen population to be tested.

[0030] The pollen viability detection method provided in this embodiment uses a pre-constructed pollen viability detection model to identify the viability type information of each individual pollen grain in the first target image of the pollen to be detected, and determines the viability information of the pollen to be detected based on the viability type information of different individual pollen grains, thereby achieving rapid and accurate detection of pollen viability, and solving the problems of low detection efficiency and accuracy in pollen viability detection using traditional pollen viability detection methods in related technologies.

[0031] In this embodiment, a pollen vitality detection method is provided, which can be used in the above-mentioned server. Figure 2 FIG. 1 is a flow chart of a pollen viability detection method according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps: Step S201: Obtain a first target image of pollen to be detected. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0032] Step S202 : Inputting the first target image of the pollen to be detected into a pre-built pollen vitality detection model, so that the pollen vitality detection model outputs a second target image, wherein the second target image contains vitality type information of each single pollen grain in the first target image.

[0033] In some optional embodiments, the pollen viability detection model is constructed by the following steps: Step a1: Acquire a plurality of first images of pollen samples and a second image of each first image.

[0034] For example, the second image contains information about the viability type of each individual pollen grain in the first image. In this embodiment, the viability type information in the second image is represented by label data, and each first image is labeled using the image annotation tool (labelme). First, different viability types of pollen are labeled with different colors, and each pollen area is classified using different colored labels. Pink, orange, red, and yellow represent HVP, LVP, NVP, and CP, respectively. Finally, the label data in JSON format is converted to the YOLO dataset format.

[0035] Step a2: Associating the first image and the second image of each pollen sample to obtain an associated data set.

[0036] For example, in the embodiment of the present application, there is no limitation on the manner of associating the first image and the second image, and those skilled in the art may determine the manner of associating the first image and the second image according to their needs, as long as it is reasonable.

[0037] Step a3: Use the associated data set to train the preset target detection model until the model accuracy meets the preset conditions, thereby obtaining a pollen vitality detection model.

[0038] Exemplarily, the preset object detection model includes a backbone network, a neck network, and a detection head. The backbone network is used to extract features from the input first image to obtain feature maps at different levels. The neck network is used to fuse the feature maps at different levels to obtain a fused feature map. The detection head is used to output a detection result based on the fused feature map. The preset conditions can be determined empirically. In an embodiment of the present application, the preset object detection model can be the You Only Look Once v10 with Convolutional Block Attention Module (YOLOv10-CBAM) object detection model. YOLOv10 is a real-time object detection system that innovatively adopts a dual allocation strategy, abandoning the traditional non-maximum suppression (NMS). Through a one-to-many and one-to-one allocation approach, it significantly improves detection efficiency. YOLOv10 integrates Pyramid Split Attention (PSA) technology. Through refined multi-scale feature extraction and attention weight allocation, it significantly enriches the dimensional information of the feature map, enabling the model to more accurately parse complex image content. However, when the input data contains a lot of noise or outliers, PSA will overemphasize this non-critical information, affecting the model's performance. The pre-trained object detection model used in this application incorporates a Convolutional Block Attention Module (CBAM) into the YOLOv10 backbone network. The CBAM module, with its two submodules: channel attention and spatial attention, effectively enhances the model's ability to identify key features and reduces its sensitivity to noise and outliers. Together, these two submodules make YOLOv10-CBAM more comprehensive and detailed in its feature representation, significantly improving recognition efficiency for complex scenes and diverse objects. Furthermore, the YOLOv10 series offers multiple variants (such as YOLOv10-N, YOLOv10-S, and YOLOv10-X) to meet the needs of different application scenarios. Here, the pre-trained YOLOv10-S model is used, as its perfect balance of speed and accuracy makes it an ideal choice for low-latency applications. Through training on large-scale datasets, the model has accumulated rich feature representations and valuable knowledge. These assets can play a vital role in new tasks, accelerating training, improving model performance, reducing the risk of overfitting, and facilitating cross-task transfer. Regarding training parameters, YOLOV10-CBAM sets epochs to 500, batch size to 4, and image size (imgsz) to 1280 to ensure efficient and stable model training.

[0039] The network structure of YOLOV10-CBAM is as follows Figure 3 As shown, YOLOv10-CBAM consists of three main components: the backbone network (Backbone), the neck network (Neck), and the detection head (Head). Backbone performs multiple convolutions (Conv) and custom modules (C2f and SCdown) on the input raw image to gradually extract basic features. C2f / SCdown in Backbone are YOLOv10's unique feature extraction modules. Similar to the residual structure of ResNet, they enhance feature extraction capabilities while reducing computational overhead through cross-layer connections and multi-branch convolutions. Backbone's fast version of Spatial Pyramid Pooling (SPPF) performs multi-scale pooling (e.g., MaxPool of different sizes) on feature maps, followed by concatenation (Concat) to improve the model's adaptability to objects of varying sizes. Backbone's CBAM first uses channel attention to focus on key semantic channels, then uses spatial attention to locate the target region, enhancing effective features and suppressing redundant information. Position-Sensitive Attention (PSA) in Backbone further optimizes feature spatial position awareness, assisting in the detection of small and densely packed objects. Neck fuses feature maps from different Backbone layers through upsampling (upsample), downsampling (conv, etc.), and concatenation (concat), allowing the model to simultaneously leverage low-level details (such as edges and textures) and high-level semantics (such as object type and overall outline). The head module includes multiple independent v10Detect, a YOLOv10-specific detection head module. Based on the fused multi-scale feature maps, v10Detect performs object classification (determining object identity) and bounding box regression (determining object location). The feature maps of different scales output by Neck are fed into independent v10Detect branches for detecting large, medium, and small objects (large objects rely on deep, low-resolution features, while small objects rely on shallow, high-resolution features). The final prediction is then output, including object type, confidence score, and bounding box coordinates, completing the mapping from features to detection results.

[0040] Step a4: obtaining a test set, which contains a plurality of image data to be tested.

[0041] For example, in the embodiments of the present application, the image data to be tested in the test set refers to the microscopic image of the stained pollen obtained by collecting the pollen samples using a portable microscope (3R-MSUSB390, Japan) system after the pollen samples are stained.

[0042] Step a5: input each image data to be tested into the pollen vitality detection model for testing, and obtain vitality type information of each single pollen grain in the corresponding image data to be tested.

[0043] For example, please refer to the above description of related contents for details, which will not be repeated here.

[0044] Step a6: evaluating the detection accuracy of the pollen vitality detection model based on the vitality type information of each single pollen grain in each image data to be tested.

[0045] Exemplarily, the detection accuracy of the pollen viability detection model is evaluated based on the viability type information of each single pollen grain detected by the model and the actual viability type of each single pollen grain.

[0046] In some optional implementations, the above step a6 includes: Step a61 : calculating the index values ​​corresponding to different preset evaluation indexes based on the vitality type information of each single pollen grain in each image data to be tested.

[0047] Exemplarily, different preset evaluation indicators may include but are not limited to the Accuracy indicator, the Precision indicator, the Recall indicator, the F1 score indicator, and the Detection Rate indicator. In the embodiment of the present application, in order to evaluate the performance of the model in the overall classification of all types of pollen, the Accuracy indicator is used. In order to evaluate the performance of the model in classifying a single type of pollen, the Precision and Recall indicators are used. In order to evaluate the comprehensive classification performance of the model for a single type of pollen, the F1 score indicator is used. These metrics are derived from the confusion matrix, which records the consistency between the actual type and the predicted type. In order to evaluate whether the pollen can be detected from the background, the detection rate is used to represent it. The calculation formulas for different preset evaluation indicators are shown below:

[0048]

[0049]

[0050]

[0051]

[0052] in, Indicates the accuracy rate, "TP" refers to the number of samples classified as positive when the label is positive. "TN" refers to the number of samples classified as negative when the label is negative. "FP" refers to the number of samples classified as positive when the label is negative; "FN" refers to the number of samples classified as negative when the label is positive. "TT" refers to the total number of labels. represents the accuracy, represents the recall rate, represents the F1 score, Indicates the detection rate.

[0053] Step a62 : evaluating the detection accuracy of the pollen vitality detection model based on the indicator values ​​corresponding to different preset evaluation indicators.

[0054] For example, the detection accuracy of the pollen viability detection model is comprehensively evaluated based on the index values ​​of different preset evaluation indicators. The embodiment of the present application does not limit the specific evaluation method, and those skilled in the art can determine it according to needs.

[0055] Step S203: determining the vitality information of the pollen to be detected based on the type information of each single pollen grain.

[0056] Specifically, the above step S203 includes: Step S2031 : determining the number of single pollen grains of each vitality type based on the vitality type information corresponding to different single pollen grains.

[0057] Illustratively, in this embodiment of the present application, the number of individual pollen of different activity types in the pollen to be detected is determined.

[0058] Step S2032: Substitute the number of single pollen grains corresponding to different vitality types into the pre-built pollen population vitality evaluation model for calculation to obtain vitality information of the pollen to be tested.

[0059] For example, the vitality information may be a pollen vitality index. In the embodiment of the present application, the pollen population vitality assessment model may be a pollen vitality index (MPVI) calculation formula. The MPVI calculation formula may include the following three types:

[0060]

[0061]

[0062] in, Indicates the first MPVI calculation formula, Indicates the second MPVI calculation formula, Indicates the third MPVI calculation formula; The value range is 1-2, with a step size of 0.1; b The value range is 0-1, the step size is 0.1, and a+b=2; The value range is 0-1, with a step size of 0.1; The value range is 1-6, with a step size of 0.5; HVP, LVP, NVP and CP refer to the number of pollen of the corresponding type.

[0063] Specifically, in the embodiments of the present application, the first MPVI calculation formula is used to calculate the pollen viability index.

[0064] To further validate the effectiveness of MPVI, a pollen tube germination assay was conducted. Liquid pollen culture was prepared by weighing 0.6g CaCl2, 0.2g H3BO3, and 445g sucrose and dissolving them in water until the volume reached 2000ml. 0.01g of treated pollen was placed in a Petri dish, and 10ml of the prepared liquid culture was added to the dish and shaken evenly. The dish was then placed in an incubator at 25°C for 2 hours. The pollen was then observed under a microscope, and the number of germinated pollen was recorded (pollen tube length greater than the diameter of the pollen grain was considered germinated). The pollen tube germination rate (PTGR) formula is as follows:

[0065] Where M is the number of germinated pollen, and N is the total pollen count. Finally, the final MPVI index is determined based on the results of Spearman correlation analysis of multiple MPVIs and PTGRs. An MPVI greater than or equal to 0.5 is considered a high-viability pollen batch (HVPB), while an MPVI less than 0.5 is considered a low-viability pollen batch (LVPB).

[0066] The testing process of the pollen viability detection model is described below through a specific embodiment.

[0067] Example: Table 1 shows the test results of different models on the Test1 dataset. The results show that, for the overall model performance evaluation, YOLOv10-CBAM and YOLOv10 achieve very high detection rates of 99.4% and 99.3%, respectively. This is very important, as it ensures that all pollen is detected. Furthermore, YOLOv10-CBAM achieves the highest accuracy of 96.9%, followed by YOLOv10 at 95.9%. YOLOv8 performs the worst, with an accuracy of 89.4%. YOLOv8 suffers from overlapping detections. For the individual category evaluation, YOLOv10-CBAM achieves the best recognition performance for HVP, LVP, and CP, with F1 scores of 97.1%, 94.5%, and 99.2%, respectively. YOLOv10 achieves the best recognition performance for NVP, with an F1 score of 98.8%. Notably, CP achieves the best recognition performance, with a precision of 100% and a recall of 98.4%. Compared to YOLOv8, YOLOv10 and YOLOv10-CBAM display more hotspots. This suggests that YOLOv10 and YOLOv10-CBAM can learn subtle features (particularly pollen at the edges of the image) that YOLOv8 cannot detect, thus avoiding missed detections. Furthermore, YOLOv10-CBAM exhibits darker colors and higher confidence levels. This indicates that YOLOv10-CBAM has stronger feature responsiveness and more accurate pollen detection than YOLOv8 and YOLOv10. Overall, the YOLOv10 and YOLOv10-CBAM models demonstrate outstanding performance on the Test1 dataset. Therefore, in subsequent modeling, we selected YOLOv10 and YOLOv10-CBAM for testing on the Test2 dataset.

[0068] Table 1

[0069] Table 2 further shows the test results for pollen images of varying densities. In the Test2-LD pollen dataset (with fewer than 50 pollen plants), the model's detection rate and overall accuracy improved, with the YOLOV10-CBAM model achieving the best performance, with a detection rate of 100% and an accuracy of 98.5%, respectively. In the Test2-HD dataset (with more than 50 pollen plants), the detection rate and overall accuracy decreased significantly, by 3-5%. YOLOV10 achieved the highest detection rate of 96.9%, while the YOLOV10-CBAM model achieved an accuracy of 91.5%. Detection errors primarily stem from adjacent classes, particularly HVP and LVP. Furthermore, high pollen density images affected the model's detection rate, thereby reducing its accuracy. In the low-density pollen dataset, YOLOV10-CBAM achieved the best classification performance for LVP, NVP, and CP. Notably, CP performed the best, achieving 100% precision and recall. In the high-density pollen dataset, YOLOV10-CBAM has the best classification performance for HVP, LVP, and CP, while YOLOV10 has the best classification performance for NVP. It is worth noting that the classification effect of HVP has dropped significantly, with the F1-score decreasing by 15%.

[0070] Table 2

[0071] Table 3 shows the test results for pollen images at different illumination intensities. In the Test2-LL pollen dataset, the YOLOV10-CBAM model's detection rate improved slightly, while that of the YOLOV10 model decreased by 0.5%. The YOLOV10 model's accuracy decreased by 4%. The YOLOV10-CBAM model maintained relatively stable performance, achieving an accuracy of 94.2%. In the Test2-HL pollen dataset, the detection rate and overall accuracy decreased by 2-5%. YOLOV10-CBAM achieved the highest detection rate and accuracy, at 98.3% and 93.9%, respectively. Detection errors primarily stemmed from adjacent classes, particularly between HVP and LVP, and between LVP and NVP. Furthermore, high-brightness pollen images affected the model's detection rate, thereby reducing its accuracy. In the low-brightness pollen dataset, YOLOV10-CBAM achieved the best classification performance for HVP, LVP, and NVP. Notably, both models achieved 100% precision and recall. In the high-brightness pollen dataset, YOLOV10-CBAM has the best classification performance for HVP, LVP, and CP, while YOLOV10 has the best classification performance for NVP.

[0072] Table 3

[0073] Table 4 shows the test results for pollen images of different varieties. The detection rate and accuracy of YOLOV10-CBAM and YOLOV10 remained stable, with the YOLOV10-CBAM model achieving superior detection and accuracy. The detection rate was 98.4% and the accuracy was 95.2%. Detection errors primarily stemmed from adjacent classes, particularly between HVP and LVP, and between LVP and NVP. Across the pollen datasets of different varieties, YOLOV10-CBAM achieved the best classification performance for HVP, LVP, NVP, and CP. Notably, the classification performance for NVP and CP was superior, with F1 scores of 97.2% and 96.8%, respectively.

[0074] Table 4

[0075] Table 5 shows the test results for pollen images from different regions and image quality. In the Test2-ZN pollen dataset, the detection rate and precision of the YOLOV10-CBAM model remained stable at 99.0% and 96.9%, respectively. However, the detection rate and precision of the YOLOV10 model decreased slightly, by approximately 1%. In the Test2-LQ pollen dataset, the detection rate and overall accuracy of both models decreased significantly, by 4-7%. YOLOV10-CBAM achieved the highest detection rate and precision, at 94.3% and 89.4%, respectively. The confusion matrix shows that detection errors primarily arise from adjacent classes, particularly between HVP and LVP, and between LVP and NVP. In the Test2-ZN pollen dataset, YOLOV10-CBAM achieved the best classification performance for LVP, NVP, and CP. Notably, the precision and recall for CP both reached 100%. YOLOV10 also achieved the best classification performance for HVP. In the Test2-LQ pollen dataset, YOLOV10-CBAM has the best classification performance for HVP, LVP, NVP and CP.

[0076] Table 5

[0077] In the present embodiment, in order to find a reliable MPVI as an indicator for evaluating pollen viability, a Spearman correlation analysis was performed between multiple MPVIs and pollen tube germination rate. The analysis results showed that the correlation of MPVI-1 was higher than that of MPVI-2 and MPVI-3. a Set to 1.5, bWhen the correlation coefficient was set at 0.5, the correlation between MPVI-1 and pollen tube germination rate reached a maximum value of 0.820, which was within the confidence interval of 0.769 to 0.860, strongly proving that there was a high positive correlation between the two. It is worth mentioning that this correlation was statistically highly significant (P<0.01). Further observation of the performance of MPVI-2 and MPVI-3 revealed that when the weight coefficient of LVP was is set to 0.3, and the weight coefficient of HVP When set to 3.0, the correlation between the two also reached its maximum value. This finding is similar to the results of MPVI-1: when the weight of HVP is higher than the weight of LVP, the correlation is generally better than when the weights of the two are equal. When the weight of HVP is three times that of LVP, the correlation reaches its optimal state. However, when the weight of HVP exceeds three times the weight of LVP, the correlation begins to show a downward trend. This strong evidence not only confirms the close correlation between MPVI and pollen tube germination rate, but also proves from a practical perspective the scientific and practical nature of MPVI as an effective tool for assessing pollen vitality.

[0078] By comparing and analyzing the MPVI results of different varieties of pollen viability, it was determined that the three varieties, V1, V4, and V5, showed high pollen viability under various temperature conditions. Moreover, as the temperature increased, their viability decreased relatively slowly, demonstrating high-temperature resistance. This discovery is of great significance for subsequent corn breeding work. At the specific temperature point of 31°C, the pollen viability of all tested varieties reached its peak, demonstrating the promoting effect of this temperature on improving pollen viability. Further observation of the temporal dynamics of pollen viability revealed a general pattern: pollen viability increased in the initial stage, then gradually decreased, until it almost completely lost its viability after 8 hours. This discovery not only reveals the typical pattern of pollen viability changes over time, but also provides an important reference for subsequent pollen preservation and application strategies.

[0079] To evaluate the model's performance in detecting pollen viability in different plant species, pollen samples were collected from 10 different plant species: wheat (Poaceae), green onion (Amaryllidaceae), pea (Leguminosae), alfalfa (Leguminosae), potato (Solanaceae), tobacco (Solanaceae), cucumber (Cucurbitaceae), honeysuckle (Caprifoliaceae), salvia miltiorrhiza (Lamiaceae), and rose (Rosaceae), representing a total of eight different families and genera. In terms of pollen morphology, wheat pollen is similar to corn pollen; green onion and pea pollen are oval; honeysuckle pollen has a distinctive triangular round shape; and potato and tobacco pollen are relatively small.

[0080] The pollen of these 10 plant species was stained using the TTC staining method. Subsequently, three images of each plant were annotated using LabelMe software. The best-performing YOLOv10-CBAM model was selected and fine-tuned using transfer learning techniques. It performed optimally in pollen viability detection for wheat, honeysuckle, cucumber, and rose, achieving detection and accuracy rates exceeding 95%. Honeysuckle achieved the best results in pollen viability detection, with an accuracy rate of 98.3% and a detection rate of 99.4%. However, pollen viability detection results for potato, tobacco, and alfalfa were relatively poor, with accuracy rates below 95% and detection rates below 90%. Further analysis revealed that the pollen grains of these three plants are relatively small and tend to stick together, posing significant challenges to the model's detection and thus impacting detection rates. (From the perspective of error distribution, the main detection errors were misclassified between NVP and CP.) However, it's worth noting that this misjudgment doesn't affect the accurate assessment of overall pollen viability, as both NVP and CP are considered inactive pollen. Overall, the accuracy of pollen viability testing for these 10 plant species remained above 90%. This result demonstrates that, using transfer learning technology, the model originally developed for corn pollen viability testing can be successfully transferred to other plant pollen viability testing in a short period of time.

[0081] This embodiment also provides a pollen vitality detection device for implementing the aforementioned embodiments and preferred embodiments. Details already described will not be repeated. As used below, the term "module" may refer to a combination of software and / or hardware that implements a predetermined function. While the devices described in the following embodiments are preferably implemented using software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0082] This embodiment provides a pollen vitality detection device, such as Figure 4 Shown, including: An acquisition module 401 is used to acquire a first target image of pollen to be detected; The detection module 402 is configured to input a first target image of pollen to be detected into a pre-built pollen vitality detection model, so that the pollen vitality detection model outputs a second target image, wherein the second target image includes vitality type information of each single pollen grain in the first target image; The determination module 403 is configured to determine the vitality information of the pollen to be detected based on the type information of each single pollen grain.

[0083] In some optional embodiments, the pollen viability detection model is constructed by the following steps: Acquire a plurality of first images of pollen samples and a second image of each first image, wherein the second image includes vitality type information corresponding to each single pollen grain in the first image; Associating the first image and the second image of each pollen sample to obtain an associated data set; The preset target detection model is trained using the associated data set until the model accuracy meets the preset conditions, and a pollen vitality detection model is obtained.

[0084] In some optional implementations, the determining module 403 includes: A determination submodule, configured to determine the number of single pollen grains of each vitality type based on the vitality type information corresponding to different single pollen grains; The calculation submodule is used to substitute the number of single pollen grains corresponding to different vitality types into the pre-built pollen population vitality evaluation model for calculation to obtain the vitality information of the pollen to be tested.

[0085] In some optional embodiments, the step of constructing the pollen viability detection model further includes: Obtain a test set, which contains multiple image data to be tested; Input each image data to be tested into the pollen vitality detection model for testing, and obtain the vitality type information of each single pollen grain in the corresponding image data to be tested; The detection accuracy of the pollen vitality detection model is evaluated based on the vitality type information of each single pollen grain in each test image data.

[0086] In some optional embodiments, the detection accuracy of the pollen vitality detection model is evaluated based on the vitality type information of each single pollen grain in each image data to be tested, including: Based on the vitality type information of each single pollen grain in each image data to be tested, the index values ​​corresponding to different preset evaluation indicators are calculated; The detection accuracy of the pollen viability detection model is evaluated based on the indicator values ​​corresponding to different preset evaluation indicators.

[0087] In some optional embodiments, the preset target detection model includes a backbone network, a neck network and a detection head; the backbone network is used to extract features from the input first image to obtain feature maps of different levels, the neck network is used to perform feature fusion on feature maps of different levels to obtain a fused feature map, and the detection head is used to output detection results based on the fused feature map.

[0088] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0089] The pollen vitality detection device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0090] The embodiment of the present invention also provides a computer device having the above Figure 4 The pollen vitality detection device shown.

[0091] See also Figure 5 , Figure 5 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 5 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of a GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 10 is taken as an example.

[0092] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0093] The memory 20 stores instructions that can be executed by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0094] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0095] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0096] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0097] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0098] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0099] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A pollen viability detection method, characterized in that: The method comprises: Acquire a first target image of pollen to be detected; Inputting the first target image of the pollen to be detected into a pre-built pollen vitality detection model so that the pollen vitality detection model outputs a second target image, wherein the second target image contains vitality type information of each single pollen grain in the first target image; The vitality information of the pollen to be detected is determined based on the type information of each single pollen grain.

2. The method according to claim 1, characterized in that The pollen viability detection model is constructed by the following steps: Acquire a plurality of first images of pollen samples and a second image of each first image, wherein the second image includes vitality type information corresponding to each single pollen grain in the first image; Associating the first image and the second image of each pollen sample to obtain an associated data set; The preset target detection model is trained using the associated data set until the model accuracy meets the preset conditions, thereby obtaining the pollen viability detection model.

3. The method according to claim 1 or 2, characterized in that The step of determining the vitality information of the pollen to be detected based on the type information of each single pollen grain includes: Determining the number of single pollen grains of each vitality type based on the vitality type information corresponding to different single pollen grains; The number of single pollen grains corresponding to different vitality types is substituted into a pre-constructed pollen population vitality evaluation model for calculation to obtain vitality information of the pollen to be tested.

4. The method according to claim 2, characterized in that The steps of constructing the pollen viability detection model further include: Acquire a test set, wherein the test set includes a plurality of image data to be tested; Input each image data to be tested into the pollen vitality detection model for testing, and obtain the vitality type information of each single pollen grain in the corresponding image data to be tested; The detection accuracy of the pollen vitality detection model is evaluated based on the vitality type information of each single pollen grain in each image data to be tested.

5. The method according to claim 4, characterized in that The step of evaluating the detection accuracy of the pollen vitality detection model based on the vitality type information of each single pollen grain in each image data to be tested includes: Based on the vitality type information of each single pollen grain in each image data to be tested, the index values ​​corresponding to different preset evaluation indicators are calculated; The detection accuracy of the pollen viability detection model is evaluated based on the indicator values ​​corresponding to different preset evaluation indicators.

6. The method according to claim 2, characterized in that The preset target detection model includes a backbone network, a neck network and a detection head; The backbone network is used to extract features from the input first image to obtain feature maps of different levels. The neck network is used to fuse features of the feature maps of different levels to obtain a fused feature map. The detection head is used to output a detection result based on the fused feature map.

7. A pollen vitality detection device, characterized in that: The device comprises: An acquisition module, configured to acquire a first target image of the pollen to be detected; a detection module, configured to input the first target image of the pollen to be detected into a pre-built pollen vitality detection model, so that the pollen vitality detection model outputs a second target image, wherein the second target image contains vitality type information of each single pollen grain in the first target image; The determination module is used to determine the vitality information of the pollen to be detected based on the type information of each single pollen grain.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the pollen viability detection method according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the pollen viability detection method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the pollen viability detection method according to any one of claims 1 to 6.

Images