Seed germination detection method and device, electronic device and storage medium

Abstract

The application provides a seed germination detection method and device, electronic equipment and storage medium, and belongs to the technical field of seed germination detection. The method comprises the following steps: performing seed individual image segmentation according to a plurality of global image frame data of a plurality of target seeds to obtain local sampling area data of each target seed; according to the local sampling area data, a plurality of local image frame data corresponding to each target seed is segmented from the plurality of global image frame data; for each target seed, pixel difference germination detection is performed according to the plurality of local image frame data to obtain pixel germination frame state data; for each target seed, semantic germination detection is performed according to the plurality of local image frame data and germination detection prompt text to obtain semantic germination frame state data; and the target germination detection result of the target seed is determined according to the pixel germination frame state data and the semantic germination frame state data. The application can improve the accuracy and efficiency of seed germination detection.

Description

Technical Field

[0001] This application relates to the field of seed germination detection technology, and in particular to a seed germination detection method and apparatus, electronic equipment and storage medium. Background Technology

[0002] In the field of seed germination detection technology, a combination of image acquisition and manual observation is used to detect the germination of specific seeds, such as mung beans, broad beans, and kidney beans, during the germination process. This determines the germination results of each seed, including its germination status, germination rate, germination potential, and germination time at different stages, providing data support for seed quality assessment. However, this detection method is highly dependent on human experience, resulting in subjective and inconsistent results, which limits both the accuracy and efficiency of seed germination detection.

[0003] Currently, some methods in the industry have improved upon this germination detection approach by introducing image processing methods or deep learning models to classify and identify seed images to determine the germination result. However, this germination detection method often relies on sample data of specific seeds for training. When applied to scenarios with different varieties or changing imaging environments, such as simultaneously analyzing multiple seeds of the same variety, it is difficult to continuously distinguish individual identities. The model's generalization ability is limited, and it is difficult to accurately determine the germination status of each seed, leading to a decrease in the accuracy of seed germination detection. Therefore, how to accurately determine the germination status of each seed to improve the accuracy and efficiency of seed germination detection is a technical problem that urgently needs to be solved in the industry. Summary of the Invention

[0004] The main objective of this application is to provide a seed germination detection method, apparatus, electronic device, and storage medium, which aims to accurately determine the germination status of each seed, thereby improving the accuracy and efficiency of seed germination detection.

[0005] To achieve the above objectives, a first aspect of this application provides a seed germination detection method, the method comprising: Acquire a collection of images showing the germination process of multiple target seeds and germination detection prompt text; wherein, the collection of images showing the germination process includes multiple global image frame data acquired in chronological order; Seed individual image segmentation is performed based on multiple global image frame data to obtain local sampling region data for each target seed; Based on the local sampling region data, multiple local image frame data corresponding to each target seed are segmented from the multiple global image frame data; For each target seed, pixel difference germination detection is performed based on multiple local image frame data to obtain pixel germination frame state data; For each target seed, semantic germination detection is performed based on multiple local image frame data and the germination detection prompt text to obtain semantic germination frame state data; The target germination detection result of the target seed is determined based on the pixel germination frame state data and the semantic germination frame state data.

[0006] In some embodiments, segmenting multiple local image frame data corresponding to each target seed from multiple global image frame data based on the local sampling region data includes: Understanding text in the germination scenario of multiple target seeds; Based on multiple global image frame data and the germination scene understanding text, seed scene parsing is performed to obtain seed layout parsing data; Based on the local sampling region data, multiple candidate local image frame data corresponding to each target seed are segmented from the multiple global image frame data; For each target seed, the image segmentation effect is verified based on the seed layout parsing data and multiple candidate local image frame data to obtain segmentation effect verification data; In response to the segmentation effect verification data satisfying the preset image segmentation effect conditions, multiple candidate local image frame data are determined as multiple local image frame data of the target seed.

[0007] In some embodiments, the step of segmenting multiple local image frame data corresponding to each target seed from multiple global image frame data based on the local sampling region data further includes: In response to the segmentation effect verification data not meeting the preset image segmentation effect conditions, segmentation difference analysis is performed based on the seed layout parsing data and multiple candidate local image frame data to obtain segmentation difference quantification data. The parameters of the local sampling region data are adjusted based on the segmentation difference quantification data to obtain the parameter-adjusted local sampling region data; Based on the local sampling region data adjusted according to the parameters, multiple candidate local image frame data corresponding to each target seed are re-segmented from multiple global image frame data until the segmentation effect verification data corresponding to the multiple candidate local image frame data obtained by re-segmentation meets the preset image segmentation effect conditions, and the multiple candidate local image frame data are determined as multiple local image frame data of the target seed.

[0008] In some embodiments, the step of segmenting individual seed images based on multiple global image frame data to obtain local sampling region data for each target seed includes: Color thresholding is performed on multiple global image frame data to obtain multiple masked global image frame data; Connectivity analysis is performed on multiple masked global image frame data to obtain local sampling region data for each target seed.

[0009] In some embodiments, the step of performing color thresholding based on multiple global image frame data to obtain multiple masked global image frame data includes: Color space conversion is performed on multiple global image frame data to obtain multiple color space image data; Threshold segmentation is performed on multiple color space image data to obtain multiple candidate mask global image frame data; Morphological optimization is performed on multiple candidate mask global image frame data to obtain multiple mask global image frame data.

[0010] In some embodiments, the step of performing pixel difference germination detection based on multiple local image frame data for each target seed to obtain pixel germination frame state data includes: For each target seed, the first frame data of the local image and the subsequent frame data of the local image are determined based on the multiple local image frame data; For each subsequent frame of the local image, pixel difference analysis is performed based on the first frame of the local image and the subsequent frame of the local image to obtain the pixel difference data corresponding to the subsequent frame of the local image. Pixel budding detection is performed based on the pixel difference data corresponding to subsequent frames of multiple local images to obtain pixel budding frame state data.

[0011] In some embodiments, for each target seed, semantic germination detection is performed based on multiple local image frame data and the germination detection prompt text to obtain semantic germination frame state data, including: For each subsequent frame of a local image for each target seed, semantic germination status detection is performed based on the first frame of the local image, the subsequent frame of the local image, and the germination detection prompt text to obtain semantic germination status data corresponding to the subsequent frame of the local image. Semantic germination detection is performed based on the semantic germination state data corresponding to the subsequent frame data of multiple local images to obtain semantic germination frame state data.

[0012] To achieve the above objectives, a second aspect of this application provides a seed germination detection device, the device comprising: The seed data acquisition unit is used to acquire a collection of images of the germination process of multiple target seeds and germination detection prompt text; wherein, the collection of images of the germination process includes multiple global image frame data acquired in chronological order; A local sampling and localization unit is used to segment individual seed images based on multiple global image frame data to obtain local sampling region data for each target seed; A local image segmentation unit is used to segment multiple local image frame data corresponding to each target seed from multiple global image frame data based on the local sampling region data; A pixel germination detection unit is used to perform pixel difference germination detection for each of the target seeds based on multiple local image frame data to obtain pixel germination frame status data. A semantic germination detection unit is used to perform semantic germination detection for each target seed based on multiple local image frame data and the germination detection prompt text, and obtain semantic germination frame state data. The germination result determination unit is used to determine the target germination detection result of the target seed based on the pixel germination frame state data and the semantic germination frame state data.

[0013] To achieve the above objectives, a third aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0014] To achieve the above objectives, a fourth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect.

[0015] This application performs seed individual image segmentation using multiple global image frame data acquired sequentially over time for multiple target seeds, obtaining local sampling region data for each target seed. Then, based on the local sampling region data, multiple local image frame data corresponding to each target seed are segmented from the multiple global image frame data. Next, for each target seed, pixel difference germination detection is performed based on the multiple local image frame data to obtain pixel germination frame state data. Semantic germination detection is also performed based on the multiple local image frame data and germination detection prompt text to obtain semantic germination frame state data. Finally, the target germination detection result of the target seed in the local sampling region data is determined based on the pixel germination frame state data and the semantic germination frame state data of the local sampling region data. In this way, it is possible to segment individual seed images to create multiple consecutive local image frames for each seed, then perform semantic germination detection through germination detection prompt text, and determine the germination status of each seed by combining the results of pixel difference germination detection. This reduces subjective errors caused by manual observation and reduces dependence on specific seed sample data. In other words, this application can accurately determine the germination status of each seed at different time periods, thereby improving the accuracy and efficiency of seed germination detection. Attached Figure Description

[0016] Figure 1 This is a flowchart of the seed germination detection method provided in the embodiments of this application; Figure 2A This is a schematic diagram of global image frame data of multiple target seeds provided in an embodiment of this application; Figure 2B This is another schematic diagram of global image frame data of multiple target seeds provided in the embodiments of this application; Figure 3 yes Figure 1 The flowchart of step S103 in the process; Figure 4 yes Figure 1 Another flowchart of step S103 in the process; Figure 5 yes Figure 1 The flowchart of step S102 in the document; Figure 6 yes Figure 5 The flowchart of step S501 in the process; Figure 7 yes Figure 1 The flowchart of step S104 in the process; Figure 8 yes Figure 1 The flowchart of step S105 in the process; Figure 9 This is a schematic diagram of the seed germination detection device provided in the embodiments of this application; Figure 10 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0018] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0020] Seeds are the foundation of agricultural production, and their quality directly affects crop yield and agricultural economic benefits. Seed germination ability is a crucial indicator of seed quality and production potential. Parameters such as germination rate, germination potential, and germination time directly reflect the physiological vitality and field performance potential of seeds, playing a vital role in crop breeding, seed industry supervision, and agricultural production practices. In seed quality testing and breeding research, accurate monitoring of individual seed germination is particularly important. By tracking the germination status of each seed at different times, we can gain a deeper understanding of the batch-to-batch quality uniformity, genetic stability, and environmental adaptability.

[0021] In the field of seed germination detection technology, a combination of image acquisition and manual observation is used to detect the germination of specific seeds, such as mung beans, broad beans, and kidney beans, during the germination process. This determines the germination results of each seed, including its germination status, germination rate, germination potential, and germination time at different times, providing data support for seed quality assessment. However, this detection method is highly dependent on human experience. It is difficult to handle the concurrent monitoring of multiple seeds simultaneously, and it is also difficult to continuously record the dynamic changes of each seed over several days of shooting. This results in highly subjective and inconsistent detection results, limiting the improvement of the accuracy and efficiency of seed germination detection.

[0022] In recent years, with the rapid development of computer vision and artificial intelligence technologies, automatic seed germination detection methods based on image processing technology have gradually become a research hotspot. These methods mainly employ threshold segmentation, edge detection, morphological operations, machine vision imaging, spectral analysis, various sensors, and infrared thermal imaging to detect seed germination. However, while this germination detection method achieves a certain degree of automation, its algorithm parameters often require manual setting and adjustment for specific seed varieties and imaging environments. When seed morphology, lighting conditions, or background changes, the stability and accuracy of detection are difficult to guarantee, and it is also difficult to effectively handle complex scenarios such as multiple seeds occluding each other and morphological changes during growth.

[0023] With the rise of deep learning technology, some methods have improved the aforementioned germination detection methods by introducing deep learning approaches, such as using Convolutional Neural Networks (CNNs) for seed germination detection. However, this germination detection method often relies on sample data of specific seeds for training. When applied to scenarios with different varieties or changing imaging environments, such as simultaneously analyzing multiple seeds of the same variety, it is difficult to continuously distinguish individual identities, leading to confusion in seed identities during growth and an inability to accurately record the specific germination time of each seed. Furthermore, existing deep learning models only perform well on specific crops and sample sets. When applied to different varieties or environmental conditions, the accuracy drops significantly, generally exhibiting limited model generalization ability, lack of single-seed temporal tracking capability, and poor interpretability, resulting in reduced accuracy in seed germination detection. Based on this, embodiments of this application provide a seed germination detection method and apparatus, electronic device, and storage medium, aiming to accurately determine the germination status of each seed at different time periods, thereby improving the accuracy and efficiency of seed germination detection.

[0024] The seed germination detection method provided in this application relates to the field of seed germination detection technology. The seed germination detection method provided in this application can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the seed germination detection method, but is not limited to the above forms.

[0025] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0026] Figure 1 This is an optional flowchart of the seed germination detection method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S101 to S106: Step S101: Obtain a collection of images showing the germination process of multiple target seeds and germination detection prompt text; Step S102: Perform seed individual image segmentation based on multiple global image frame data to obtain local sampling region data for each target seed; Step S103: Based on the local sampling region data, segment multiple local image frame data corresponding to each target seed from multiple global image frame data; Step S104: For each target seed, perform pixel difference germination detection based on multiple local image frame data to obtain pixel germination frame state data. Step S105: For each target seed, semantic germination detection is performed based on multiple local image frame data and germination detection prompt text to obtain semantic germination frame state data. Step S106: Determine the target germination detection result of the target seed based on the pixel germination frame state data and the semantic germination frame state data.

[0027] Steps S101 to S106 as illustrated in this embodiment involve segmenting individual seed images using multiple global image frame data collected sequentially over time from multiple target seeds to obtain local sampling region data for each target seed. Then, based on the local sampling region data, multiple local image frame data corresponding to each target seed are segmented from the multiple global image frame data. Next, for each target seed, pixel difference germination detection is performed based on the multiple local image frame data to obtain pixel germination frame state data. Semantic germination detection is also performed based on the multiple local image frame data and germination detection prompt text to obtain semantic germination frame state data. Finally, the target germination detection result of the target seed in the local sampling region data is determined based on the pixel germination frame state data and the semantic germination frame state data of the local sampling region data. In this way, it is possible to segment individual seed images to create multiple consecutive local image frames for each seed, then perform semantic germination detection through germination detection prompt text, and determine the germination status of each seed by combining the results of pixel difference germination detection. This reduces subjective errors caused by manual observation and reduces dependence on specific seed sample data. In other words, this application can accurately determine the germination status of each seed at different time periods, thereby improving the accuracy and efficiency of seed germination detection.

[0028] In step S101 of some embodiments, the target seed may refer to the individual seed to be tested for germination. For example, the target seed may be a corn seed, rapeseed seed, soybean seed, wheat seed, rice seed, or cotton seed, etc., and is not specifically limited. The germination process image set may refer to an image set including multiple global image frame data acquired in chronological order, used to record the complete germination process of multiple target seeds from the start to the end of germination. The global image frame data may refer to image data containing multiple target seeds acquired at a single time point. For example, please refer to... Figure 2A and Figure 2B , Figure 2A This is a schematic diagram of global image frame data for multiple target seeds provided in an embodiment of this application. Figure 2A The target seed in the experiment is a wheat seed, showing the initial state of multiple wheat seeds at the start of the germination experiment. Figure 2B This is another schematic diagram of global image frame data for multiple target seeds provided in the embodiments of this application. Figure 2BThe target seeds in the image are corn seeds, showing the initial state of multiple corn seeds at the start of the germination experiment. It should be noted that the germination process image set is obtained by capturing images of the target seeds during the germination process at preset time intervals, while keeping the seed position and acquisition conditions constant. The number of global image frames included in the germination process image set can be adjusted according to actual needs. For example, in this embodiment, a germination experiment can be conducted using the standard germination bed method. The germination box is placed in an intelligent temperature-controlled germination chamber, with a layer of sponge and a sheet of pressure paper laid in the germination bed. Twenty corn seeds are placed above the germination bed in preset rows and columns, and the target seeds during the germination process are photographed by a camera at preset time intervals, thereby acquiring multiple global image frames acquired in chronological order.

[0029] Germination detection prompt text refers to instruction text used to semantically determine the germination status of a target seed. For example, the germination detection prompt text could be, "Compare these two global image frame data to determine whether the target seed has germinated, such as whether the radicle has emerged or the seed coat has ruptured"; or, the germination detection prompt text could be, "Compare the global image frame data of the target seed in the first frame and the current frame to identify whether there are germination features." Understandably, the specific content and expression of the germination detection prompt text can be adjusted according to actual needs, but it must be able to determine the germination status of the target seed.

[0030] In steps S102 to S103 of some embodiments, seed individual image segmentation can refer to the process of identifying and separating the region where each target seed is located from multiple global image frame data. Local sampling region data can refer to coordinate data obtained through seed individual image segmentation, used to describe the position of each target seed in the global image frame data. For example, local sampling region data could be... Figure 2A The local sampling region data refers to the bounding rectangle coordinates of the first wheat seed in the top left corner of the first global image frame data; or, alternatively, the local sampling region data can be the collection of the bounding rectangle coordinates of the wheat seed in all global image frame data. It is understood that the specific form of the local sampling region data can be adjusted according to actual needs. Local image frame data can refer to the local image data corresponding to each target seed in each global image frame data, segmented from the global image frame data based on the local sampling region data. For example, the local image frame data can be based on... Figure 2AThe local image frame data is obtained by cropping the bounding rectangle boundary coordinates of the seed from the global image frame data of the first frame to all subsequent frames. Alternatively, the local image frame data can also be obtained by cropping the local image frame data of the seed in each corresponding frame from the global image frame data of the set of bounding rectangle boundary coordinates of the seed in all global image frame data. It can be understood that each target seed corresponds to a local image frame data in each global image frame data.

[0031] In step S104 of some embodiments, pixel difference germination detection can refer to the process of comparing the pixel differences of the target seed in the first frame with the pixel differences in the local image frame data of the target seed in subsequent frames, and determining whether the target seed has germinated based on the pixel differences in each frame. Pixel germination frame state data can refer to the germination state result data of the target seed in multiple local image frame data obtained by pixel difference germination detection. For example, for a certain target seed, the pixel germination frame state data can include a set of state data in which the target seed is determined to be "not germinated" in the second frame of local image frame data based on pixel differences, and "germinated" in the third frame of local image frame data based on pixel differences; or, the pixel germination frame state data can also include a sequence of germination state data of the target seed determined based on pixel differences in all local image frame data. It should be noted that, since seed germination is an irreversible process, once a target seed is determined to have germinated in a local image frame in a certain frame, the embodiments of this application do not need to perform pixel difference germination detection on the local image frame data of the target seed in subsequent frames, thereby reducing the overall computational resource consumption and improving the efficiency of germination detection.

[0032] In step S105 of some embodiments, semantic germination detection may refer to the process of performing semantic analysis on the local image frame data of the target seed in the first frame and the local image frame data in subsequent frames, combined with the germination detection prompt text, and determining whether the target seed has germinated based on the semantic analysis results. Semantic germination frame state data may refer to the germination state result data of the target seed in multiple local image frame data obtained through semantic germination detection. For example, for a certain target seed, the semantic germination frame state data may include a set of state data where the target seed is determined to be "not germinated" in the local image frame data of the 5th frame based on semantic analysis and "germinated" in the local image frame data of the 6th frame based on semantic analysis; or, the semantic germination frame state data may also include a sequence of germination state data of the target seed determined based on semantic analysis in all local image frame data. It should be noted that semantic germination detection, by introducing germination detection prompt text for semantic analysis, can identify subtle germination features such as radicle emergence and seed coat rupture, making up for the shortcomings of pixel difference germination detection in mid-to-late frames. Furthermore, semantic germination detection can stop performing semantic germination detection on subsequent local image frames of a target seed after it has been determined to have germinated in a local image frame of a certain frame, thereby reducing the overall computational resource consumption.

[0033] In step S106 of some embodiments, the target germination detection result may refer to the final germination state result data of the target seed determined comprehensively based on pixel germination frame state data and semantic germination frame state data. For example, if the pixel germination frame state data and the semantic germination frame state data have the same germination state determination for the same local image frame data, such as both being determined to be "germinated" in the local image frame data of frame 5, then this determination result can be determined as the germination state of that local image frame data in the target germination detection result. Alternatively, if the pixel germination frame state data and the semantic germination frame state data have different germination state determinations for the same local image frame data, then the germination state of the corresponding frame number in either the pixel germination frame state data or the semantic germination frame state data can be selected as the germination state of that local image frame data in the target germination detection result based on the frame number corresponding to that local image frame data. For example, if frames 1 to 3 are defined as early frames, and frames 4 and onwards as mid-to-late frames, if a target seed is determined to have "sprouted" in the pixel germination frame state data of the local image frame data in frame 2, but "not sprouted" in the semantic germination frame state data, then since frame 2 is an early frame, the "sprouted" in the pixel germination frame state data can be determined as the germination state of that local image frame data in the target germination detection result. If the target seed is determined to have "not sprouted" in the pixel germination frame state data of the local image frame data in frame 5, but "sprouted" in the semantic germination frame state data, then since frame 5 is a mid-to-late frame, the "sprouted" in the semantic germination frame state data can be determined as the germination state of that local image frame data in the target germination detection result. Alternatively, if the pixel germination frame state data and the semantic germination frame state data have inconsistent germination status determinations for the same local image frame data, a weighted fusion can be performed based on the respective confidence levels of the pixel germination frame state data and the semantic germination frame state data to obtain the germination status of that local image frame data in the target germination detection result. For example, if a target seed in the 4th frame of local image frame data is determined to be "germinated" by the pixel germination frame state data with a corresponding germination probability of 0.8, and the semantic germination frame state data is determined to be "not germinated" with a corresponding germination probability of 0.4, and the weights of both the pixel germination frame state data and the semantic germination frame state data are set to 0.5, then the weighted fusion germination probability is 0.6. If the preset threshold is 0.5, then the germination status of that local image frame data is determined to be "germinated". It is understood that the specific method for determining the target germination detection result in this embodiment can be adjusted according to actual needs.

[0034] Please see Figure 3 In some embodiments, step S103 may include, but is not limited to, steps S301 to S305: Step S301: Obtain the germination scene understanding text of multiple target seeds; Step S302: Based on multiple global image frame data and the germination scene understanding text, seed scene parsing is performed to obtain seed layout parsing data; Step S303: Based on the local sampling region data, segment multiple candidate local image frame data corresponding to each target seed from multiple global image frame data; Step S304: For each target seed, the image segmentation effect is verified based on the seed layout parsing data and multiple candidate local image frame data to obtain segmentation effect verification data. Step S305: In response to the segmentation effect verification data meeting the preset image segmentation effect conditions, multiple candidate local image frame data are determined as multiple local image frame data of the target seed.

[0035] In step S301 of some embodiments, the germination scene understanding text can refer to instruction text used for semantic understanding of the germination scene indicated by the global image frame data. For example, the germination scene understanding text could be "Please identify the germination bed region in the image and output the boundary coordinates of the region and the number of target seeds"; or, the germination scene understanding text could also be "Please find the location of the petri dish where the target seeds are placed in the image and estimate how many target seeds there are." It is understood that the specific content and expression of the germination scene understanding text can be adjusted according to actual needs, but it must be able to be used to identify the germination bed region and the distribution of target seeds.

[0036] In step S302 of some embodiments, seed scene parsing can refer to the process of calling a specific visual language model to perform semantic parsing on germination scenes indicated by multiple globally acquired image frames in chronological order based on the germination scene understanding text, in order to obtain germination scene layout information. Seed layout parsing data can refer to data obtained through seed scene parsing that describes the germination scene layout. For example, seed layout parsing data can be obtained by calling a Vision-Language Model (VLM) model to perform seed scene parsing, and the seed layout parsing data may include the number of target seeds; or, seed layout parsing data can also be obtained by calling a GPT-4V model to perform seed scene parsing, and the seed layout parsing data may include the distribution coordinate information of the target seeds within the germination bed. It is understood that the specific implementation of seed scene parsing can be adjusted according to actual needs, and the specific content and format of the seed layout parsing data can also be adjusted according to the instructions of the germination scene understanding text.

[0037] In step S303 of some embodiments, candidate local image frame data may refer to the initial local image frame data corresponding to each target seed in each global image frame data, which is obtained by segmenting from each global image frame data based on local sampling region data.

[0038] In step S304 of some embodiments, image segmentation effect verification can refer to the process of verifying multiple consecutive candidate local image frame data of the target seed based on seed layout parsing data, to determine whether the number or position of the seed corresponding to the candidate local image frame data conforms to the target seed number or target seed distribution location information indicated by the seed layout parsing data. Segmentation effect verification data can refer to data obtained through image segmentation effect verification, used to indicate the verification result of the candidate local image frame data. For example, segmentation effect verification data may include quantity verification result data, used to indicate whether the number of target seeds corresponding to the candidate local image frame data is consistent with the number of target seeds in the seed layout parsing data, wherein the number of target seeds corresponding to the candidate local image frame data is the same as the number of candidate local image frame data; or, segmentation effect verification data may also include position verification result data, used to indicate whether the position data of each target seed in the candidate local image frame data is located within the region corresponding to the target seed distribution location information indicated by the seed layout parsing data, wherein the position data of each target seed in the candidate local image frame data is the same as the position data indicated by the local sampling region data. It is understood that the specific content and form of the segmentation effect verification data can be adjusted according to actual verification needs.

[0039] In step S305 of some embodiments, the preset image segmentation effect condition may refer to the judgment condition used to determine whether the candidate local image frame data meets the requirements of the seed layout parsing data. If the segmentation effect verification data meets the preset image segmentation effect condition, it indicates that the number or position of seeds in the candidate local image frame data is consistent with the target seed number or distribution position information indicated by the seed layout parsing data. Multiple consecutive candidate local image frame data can be determined as multiple consecutive local image frame data of the target seeds. If the segmentation effect verification data does not meet the preset image segmentation effect condition, it indicates that the number or position of seeds in the candidate local image frame data is inconsistent with the target seed number or distribution position information indicated by the seed layout parsing data. In this case, the process of segmenting individual seed images can be optimized to obtain more accurate local sampling region data. Then, the segmentation of candidate local image frame data based on the local sampling region data and the image segmentation effect verification are re-executed until multiple local image frame data that meet the preset image segmentation effect condition are obtained.

[0040] It is understood that this application embodiment obtains germination scene understanding text for multiple target seeds, performs seed scene parsing based on multiple global image frame data and germination scene understanding text to obtain seed layout parsing data, and then segments multiple candidate local image frame data corresponding to each target seed from multiple global image frame data based on local sampling region data. Then, for each target seed, image segmentation effect verification is performed based on seed layout parsing data and multiple candidate local image frame data to obtain segmentation effect verification data. When the segmentation effect verification data meets preset image segmentation effect conditions, the multiple candidate local image frame data are determined as multiple local image frame data of the target seed. In this way, by introducing germination scene understanding text to obtain seed layout parsing data, reference seed layout information can be provided for candidate local image frame data obtained based on local sampling region data segmentation. Image segmentation effect verification ensures that the number and position of seeds in the final multiple local image frame data match the target seed number or target seed distribution information indicated by the seed layout parsing data, thereby improving the accuracy of individual seed image segmentation and providing a reliable data foundation for subsequent germination detection.

[0041] Please see Figure 4 In some embodiments, step S103 may also include, but is not limited to, steps S401 to S403: Step S401: In response to the segmentation effect verification data not meeting the preset image segmentation effect conditions, segmentation difference analysis is performed based on seed layout parsing data and multiple candidate local image frame data to obtain segmentation difference quantification data. Step S402: Adjust the parameters of the local sampling region data based on the segmentation difference quantification data to obtain the parameter-adjusted local sampling region data; Step S403: Based on the local sampling region data adjusted by parameters, re-segment multiple candidate local image frame data corresponding to each target seed from multiple global image frame data until the segmentation effect verification data corresponding to the multiple candidate local image frame data obtained by re-segmentation meets the preset image segmentation effect conditions, and determine the multiple candidate local image frame data as multiple local image frame data of the target seed.

[0042] In step S401 of some embodiments, segmentation difference analysis can refer to the process of quantifying the degree of deviation between seed layout parsing data and target seed data from multiple consecutive candidate local image frames, focusing on the number or position of seeds. Segmentation difference quantification data can refer to a number obtained through segmentation difference analysis that describes the degree of difference between candidate local image frame data and seed layout parsing data. For example, segmentation difference quantification data may include the difference between the number of target seeds and the number of target seeds in the seed layout parsing data; or, segmentation difference quantification data may also include the offset distance data between the target seed position and the target seed distribution position in the seed layout parsing data.

[0043] In step S402 of some embodiments, parameter adjustment may refer to the process of optimizing local sampling region data based on segmentation difference quantization data to obtain more accurate position coordinate data. The parameter-adjusted local sampling region data may refer to the local sampling region data obtained after parameter optimization, used for re-segmenting candidate local image frame data. For example, if the segmentation difference quantization data indicates a deviation in the seed position corresponding to the candidate local image frame data, parameter adjustment may include adjusting the coordinate range in the local sampling region data to correct the seed position deviation, thereby obtaining parameter-adjusted local sampling region data; or, if the segmentation difference quantization data indicates that the size of the seed region corresponding to the candidate local image frame data does not match the actual seed size, parameter adjustment may further include adjusting the size of the bounding rectangle in the local sampling region data to more completely cover the target seed region, thereby obtaining parameter-adjusted local sampling region data. It is understood that the specific method of parameter adjustment can be adjusted according to the content of the segmentation difference quantization data.

[0044] In step S403 of some embodiments, after the parameter adjustment is completed, multiple candidate local image frame data corresponding to each target seed can be re-segmented from multiple global image frame data based on the local sampling region data of the parameter adjustment, and the image segmentation effect verification continues until the segmentation effect verification data corresponding to the multiple candidate local image frame data obtained by re-segmentation meets the preset image segmentation effect conditions. Finally, the multiple candidate local image frame data that meet the conditions are determined as the multiple local image frame data of the target seed.

[0045] It is understood that, in this embodiment of the application, when the segmentation effect verification data does not meet the preset image segmentation effect conditions, segmentation difference analysis is performed based on seed layout analysis data and multiple candidate local image frame data to obtain segmentation difference quantification data. Then, based on the segmentation difference quantification data, parameters of the local sampling region data are adjusted to obtain parameter-adjusted local sampling region data. Next, based on the parameter-adjusted local sampling region data, multiple candidate local image frame data corresponding to each target seed are re-segmented from multiple global image frame data until the segmentation effect verification data corresponding to the re-segmented multiple candidate local image frame data meets the preset image segmentation effect conditions. The multiple candidate local image frame data that meet the conditions are then determined as the multiple local image frame data of the target seed. In this way, when there is a deviation between the candidate local image frame data segmented based on the local sampling region data and the number or distribution position of the target seeds indicated by the seed layout analysis data, the degree of deviation can be quantified through segmentation difference analysis, and the local sampling region data can be specifically adjusted based on the quantified data. This allows the re-segmented candidate local image frame data to more accurately meet the requirements of the seed layout analysis data, thereby improving the accuracy of individual seed image segmentation and providing a reliable data foundation for subsequent germination detection.

[0046] In some embodiments, seed individual image segmentation can be implemented using image segmentation algorithms, such as OpenCV-based image segmentation algorithms or the Watershed Algorithm, to obtain local sampling region data for each target seed. Then, based on the obtained local sampling region data, multiple candidate local image frames corresponding to each target seed are segmented from multiple global image frames. If the segmentation performance verification data does not meet the preset image segmentation performance conditions, the corresponding parameters of the image segmentation algorithm can be adjusted based on the segmentation difference quantization data. For example, if the segmentation difference quantization data indicates a deviation in the seed position corresponding to the candidate local image frame data, parameters related to positioning accuracy in the image segmentation algorithm (such as color threshold range or connected component analysis parameters) can be adjusted to correct the coordinate range in the regenerated local sampling region data, thus obtaining parameter-adjusted local sampling region data. Alternatively, if the segmentation difference quantization data indicates that the seed region size corresponding to the candidate local image frame data does not match the actual seed size, parameters related to region range in the image segmentation algorithm (such as the size of the structuring element in morphological operations or the marker point parameters in the watershed algorithm) can be adjusted to adjust the size of the bounding rectangle in the regenerated local sampling region data, so as to more completely cover the target seed region, thus obtaining parameter-adjusted local sampling region data. Subsequently, image segmentation and subsequent verification steps are performed again based on the parameter-adjusted local sampling region data until multiple local image frame data that meet the preset conditions are obtained. It is understood that the specific method of parameter adjustment can be adjusted according to the content of the segmentation difference quantization data.

[0047] It should be noted that in practical applications, the above-mentioned segmentation difference analysis and parameter adjustment process can be implemented by calling a Large Language Model (LLM). The LLM can analyze the specific problems with the current segmentation parameters based on the comparison results between the seed layout parsing data and the candidate local image frame data, and generate targeted parameter optimization suggestions. For example, when the number of seeds in the candidate local image frame data is inconsistent with the number of target seeds in the seed layout parsing data, the LLM can analyze that it may be due to an improper color threshold range setting, and thus suggest adjusting the relevant parameters; or, when there is a deviation between the seed position in the candidate local image frame data and the target seed distribution position in the seed layout parsing data, the LLM can analyze that it may be due to a deviation in the coordinate extraction parameters or morphological operation parameters in the connected component analysis, and thus suggest adjusting the relevant parameters. By introducing the LLM, intelligent diagnosis and optimization of segmentation parameters can be achieved, improving the efficiency and accuracy of parameter adjustment.

[0048] Please see Figure 5In some embodiments, step S102 may include, but is not limited to, steps S501 to S502: Step S501: Perform color thresholding based on multiple global image frame data to obtain multiple masked global image frame data; Step S502: Perform connected component analysis based on multiple masked global image frame data to obtain local sampling region data for each target seed.

[0049] In step S501 of some embodiments, color thresholding segmentation can refer to the process of setting a threshold range based on multiple globally acquired image frame data in chronological order, segmenting pixel regions that meet the threshold conditions from the globally acquired image frame data, and converting them into binary mask images. The masked global image frame data can refer to the binary mask image data obtained through color thresholding segmentation, used to identify the region where the target seed is located. For example, the global image frame data for the mask can be obtained by setting threshold ranges for hue, saturation, and brightness based on the target seed's HSV color space, converting the global image frame data to the HSV color space, and then performing threshold segmentation on the converted global image frame data. Pixel regions that simultaneously meet the hue, saturation, and brightness threshold ranges are assigned a value of 255, while the remaining regions are assigned a value of 0, thus obtaining a binary mask image. Alternatively, the global image frame data for the mask can also be obtained by setting threshold ranges for the red, green, and blue channels based on the target seed's Red-Green-Blue (RGB) color space, performing threshold segmentation on the global image frame data, assigning a value of 255 to pixel regions that simultaneously meet the red, green, and blue channel threshold ranges, and assigning the remaining regions a value of 0, thus obtaining a binary mask image. It is understandable that the specific implementation of color threshold segmentation can be adjusted according to actual needs.

[0050] In step S502 of some embodiments, connected component analysis can refer to performing connected component analysis based on multiple masked global image frame data to identify and label connected regions composed of adjacent pixels in each masked global image frame data, and extracting the bounding rectangle boundary coordinate data of each connected region. Local sampling region data can refer to the set of coordinate data obtained through connected component analysis that describes the region where each target seed is located in each image frame. For example, local sampling region data may include the bounding rectangle boundary coordinate data of the target seed in the first global image frame data; or, local sampling region data may also include the set of bounding rectangle boundary coordinate data of the target seed in all global image frame data. It is understood that the specific content and form of the local sampling region data can be adjusted according to actual needs.

[0051] It is understood that the embodiments of this application obtain multiple masked global image frame data by performing color thresholding based on multiple global image frame data, and then perform connected component analysis based on the multiple masked global image frame data to obtain local sampling region data for each target seed. In this way, the inherent color characteristics of the seed can be used to weaken background noise interference in the global image frame data, and more accurate position coordinate information of each target seed can be extracted from the obtained masked global image frame data, providing accurate location information for subsequent segmentation of local image frame data from the global image frame data and germination detection.

[0052] It should be noted that after obtaining the local sampling region data through connected component analysis, the embodiments of this application can further optimize the local sampling region data. For example, based on preset minimum and maximum area thresholds, the connected component areas of each target seed in the local sampling region data are filtered to exclude noise regions with areas smaller than the minimum area threshold and contiguous regions with areas larger than the maximum area threshold, resulting in optimized local sampling region data; or, based on preset boundary expansion coefficients, the bounding rectangle boundary coordinates of each target seed in the local sampling region data are expanded to completely cover the target seed region, and the expanded boundary coordinates are cropped so that they do not exceed the image range, resulting in optimized local sampling region data. It is understood that by optimizing the local sampling region data, the accuracy of the local sampling region data can be improved, providing more accurate positional data support for subsequent segmentation of local image frame data from global image frame data.

[0053] Please see Figure 6 In some embodiments, step S501 may also include, but is not limited to, steps S601 to S603: Step S601: Perform color space conversion based on multiple global image frame data to obtain multiple color space image data; Step S602: Threshold segmentation is performed based on multiple color space image data to obtain multiple candidate mask global image frame data; Step S603: Morphological optimization is performed based on multiple candidate mask global image frame data to obtain multiple mask global image frame data.

[0054] In step S601 of some embodiments, color space conversion can refer to the process of converting global image frame data from an original color space to a target color space based on multiple global image frame data. Color space image data can refer to image data represented in the target color space after color space conversion. For example, color space image data can be hue, saturation, and lightness component data obtained after HSV color space conversion based on multiple global image frame data; or, color space image data can also be grayscale value data of each pixel obtained after grayscale space conversion based on multiple global image frame data. It is understood that the specific method of color space conversion can be selected according to actual needs.

[0055] In step S602 of some embodiments, threshold segmentation can refer to the process of setting a threshold range based on multiple color space image data, segmenting pixel regions that meet the threshold conditions from the color space image data, and converting them into a binary image. Candidate mask global image frame data can refer to the initial binary segmented image data obtained through threshold segmentation, used to identify the region where the target seed is located. For example, candidate mask global image frame data can be obtained by setting threshold ranges for hue, saturation, and brightness based on HSV color space image data, performing threshold segmentation on the HSV color space image data, assigning a value of 255 to pixel regions that simultaneously meet the hue, saturation, and brightness threshold ranges, and assigning a value of 0 to the remaining regions; or, candidate mask global image frame data can also be obtained by setting a grayscale threshold range based on grayscale color space image data, performing threshold segmentation on the grayscale color space image data, assigning a value of 255 to pixel regions that meet the grayscale threshold range, and assigning a value of 0 to the remaining regions. It is understood that the specific parameters and methods of threshold segmentation can be adjusted according to actual needs.

[0056] In step S603 of some embodiments, morphological optimization can refer to the process of performing morphological operations on multiple candidate mask global image frame data to remove noise points and fill holes inside the target seed, thereby improving the quality of the segmented image. The mask global image frame data can refer to binary mask image data obtained through morphological optimization, after noise removal and hole filling. For example, morphological optimization can involve first performing an opening operation on the candidate mask global image frame data (i.e., erosion followed by dilation) to remove small noise points in the candidate mask global image frame data, and then performing a closing operation on the opened image data (i.e., dilation followed by erosion) to fill the small holes inside the target seed, ultimately obtaining the mask global image frame data; or, morphological optimization can also involve first performing a closing operation on the candidate mask global image frame data (i.e., dilation followed by erosion) to fill the small holes inside the target seed, and then performing an opening operation on the closed image data (i.e., erosion followed by dilation) to remove small noise points in the candidate mask global image frame data, ultimately obtaining the mask global image frame data. Understandably, the specific order of operations, the size of structural elements, and the number of iterations in morphological optimization can be adjusted based on the actual segmentation results.

[0057] It is understood that the embodiments of this application obtain multiple color space image data by performing color space conversion based on multiple global image frame data, then perform threshold segmentation based on the multiple color space image data to obtain multiple candidate mask global image frame data, and finally perform morphological optimization based on the multiple candidate mask global image frame data to obtain multiple mask global image frame data. In this way, it is possible to highlight the color features of the target seed by utilizing color space conversion, initially separate the seed region from the background by threshold segmentation, and then remove noise points and fill the internal holes of the seed by morphological optimization, thereby obtaining a more accurate binary mask image, providing accurate data support for subsequent connected component analysis and acquisition of local sampling region data.

[0058] Please see Figure 7 In some embodiments, step S104 may also include, but is not limited to, steps S701 to S703: Step S701: For each target seed, determine the first frame data of the local image and the subsequent frame data of the multiple local images based on the local image frame data. Step S702: For each local image subsequent frame data, perform pixel difference analysis based on the local image first frame data and local image subsequent frame data to obtain the pixel difference data corresponding to the local image subsequent frame data; Step S703: Pixel budding detection is performed based on the pixel difference data corresponding to subsequent frame data of multiple local images to obtain pixel budding frame state data.

[0059] In step S701 of some embodiments, the local image first frame data may refer to the local image data corresponding to each target seed in the first frame. The local image subsequent frame data may refer to the local image data corresponding to each target seed in all subsequent frames after the first frame.

[0060] In step S702 of some embodiments, pixel difference analysis can refer to the process of calculating and normalizing the absolute difference of grayscale values ​​pixel by pixel after performing grayscale conversion based on the first frame data and subsequent frame data of the local image. Pixel difference data can refer to the quantitative data obtained through pixel difference analysis that describes the degree of pixel difference between the subsequent frame data and the first frame data of the local image. For example, for a target seed, after converting the first frame data and the subsequent frame data of the local image in the third frame into grayscale images respectively, the absolute difference of grayscale values ​​at each corresponding pixel position is calculated. The sum of the absolute differences of all pixels is then divided by the total number of pixels to obtain the normalized average absolute difference value, which is used as the pixel difference data of the target seed in the third frame.

[0061] In step S703 of some embodiments, pixel germination detection can refer to the process of determining whether a target seed has germinated in each subsequent frame of a local image based on the pixel difference data corresponding to each target seed in the subsequent frame data of each local image. Pixel germination frame state data can refer to the result data obtained through pixel germination detection, used to indicate the germination state of each target seed in each subsequent frame of a local image. For example, for a target seed, based on the pixel difference data corresponding to all its subsequent frame data of local images, by comparing a preset difference threshold, the germination state of the target seed is determined frame by frame, resulting in a sequence composed of the germination states of each frame, which serves as the pixel germination frame state data of the target seed; or, for another target seed, based on its pixel difference data, if the preset threshold is exceeded for the first time in a certain frame, then that frame and subsequent frames are determined to be in a germination state, obtaining the corresponding pixel germination frame state data. It is understood that the specific determination method of pixel germination detection can be adjusted according to actual needs.

[0062] It is understood that, in this embodiment of the application, the first frame data and subsequent frames of a local image are determined for each target seed based on multiple local image frame data. Then, for each subsequent frame of a local image, pixel difference analysis is performed based on the first and subsequent frames to obtain pixel difference data corresponding to the subsequent frames. Finally, pixel germination detection is performed based on the pixel difference data corresponding to the subsequent frames to obtain pixel germination frame status data. In this way, by using the first frame image of each target seed as a benchmark and quantifying the degree of pixel difference between each subsequent frame and the first frame image, the germination status of the target seed can be determined frame by frame, thereby accurately determining the germination time point of each target seed and improving the accuracy of seed germination detection.

[0063] Please see Figure 8 In some embodiments, step S105 may include, but is not limited to, steps S801 to S802: Step S801: For each local image subsequent frame data of each target seed, perform semantic germination state detection based on the local image first frame data, local image subsequent frame data and germination detection prompt text to obtain semantic germination state data corresponding to the local image subsequent frame data. Step S802: Semantic germination detection is performed based on the semantic germination state data corresponding to subsequent frame data of multiple local images to obtain semantic germination frame state data.

[0064] In step S801 of some embodiments, semantic germination state detection can refer to the process of calling a specific visual language model to perform semantic analysis based on the first frame data of the local image, subsequent frame data of the local image, and germination detection prompt text for each local image subsequent frame data, in order to determine the germination state corresponding to that local image subsequent frame data. Here, the specific visual language model can refer to an artificial intelligence model capable of simultaneously processing image data and text data and performing cross-modal understanding. For example, the specific visual language model can be the Qwen3-VL-Plus model or the GPT-4V model, without specific limitations. Semantic germination state data can refer to the data obtained through semantic germination state detection, used to indicate the germination state corresponding to each target seed in each local image subsequent frame data. For example, for a target seed, the first frame data of its local image, the subsequent frame data of the local image in the 5th frame, and the germination detection prompt text are input into a specific visual language model. The specific visual language model outputs that the germination status of the target seed in the 5th frame is "germinated", thus obtaining the semantic germination status data of the target seed in the 5th frame. Alternatively, for another target seed, the first frame data of its local image, the subsequent frame data of the local image in the 8th frame, and the germination detection prompt text are input into a specific visual language model. The specific visual language model outputs that the germination status of the target seed in the 8th frame is "not germinated", thus obtaining the semantic germination status data of the target seed in the 8th frame.

[0065] In step S802 of some embodiments, semantic germination detection can refer to the process of determining the final germination state of a target seed in each subsequent frame of local images based on the semantic germination state data corresponding to all subsequent frames of local images for each target seed. Semantic germination frame state data can refer to the result data obtained through semantic germination detection, used to indicate the final germination state of each target seed in each subsequent frame of local images. For example, for a target seed, its semantic germination state data from frame 2 to all subsequent frames are combined in frame order to obtain the germination state sequence of the target seed in each frame, which serves as the semantic germination frame state data of the target seed; or, for another target seed, based on its semantic germination state data in each frame, the frame number where "germinated" first appears and the state of its subsequent frames are recorded to obtain the corresponding semantic germination frame state data. It is understood that the specific method of semantic germination detection can be adjusted according to actual needs.

[0066] It is understood that, in this embodiment of the application, semantic germination state detection is performed on each subsequent frame of a local image for each target seed, based on the first frame of the local image, subsequent frame data, and germination detection prompt text. This yields semantic germination state data corresponding to the subsequent frame data of the local image. Then, semantic germination detection is performed on the semantic germination state data corresponding to multiple subsequent frame data of the local images to obtain semantic germination frame state data. In this way, by introducing germination detection prompt text for semantic germination state detection, subtle germination features such as radicle emergence and seed coat rupture can be identified, thereby more accurately determining the germination state of each target seed in each image frame and improving the accuracy of seed germination detection.

[0067] Please see Figure 9 This application also provides a seed germination detection device, which can implement the above-mentioned seed germination detection method. The device includes: The seed data acquisition unit 901 is used to acquire a collection of images of the germination process of multiple target seeds and germination detection prompt text; wherein, the collection of images of the germination process includes multiple global image frame data acquired in chronological order; The local sampling and localization unit 902 is used to segment individual seed images based on multiple global image frame data to obtain local sampling region data for each target seed. The local image segmentation unit 903 is used to segment individual seed images based on multiple global image frame data to obtain local sampling region data for each target seed; The pixel germination detection unit 904 is used to perform pixel difference germination detection for each target seed based on multiple local image frame data to obtain pixel germination frame state data. The semantic germination detection unit 905 is used to perform semantic germination detection for each target seed based on multiple local image frame data and germination detection prompt text, and obtain semantic germination frame state data. The germination result determination unit 906 is used to determine the target germination detection result of the target seed based on the pixel germination frame state data and the semantic germination frame state data.

[0068] The specific implementation of this seed germination detection device is basically the same as the specific implementation of the seed germination detection method described above, and will not be repeated here.

[0069] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described seed germination detection method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0070] Please see Figure 10 , Figure 10The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 1001 can be implemented using a general-purpose central processing unit (CPU), microprocessor, application specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 1002 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 1002 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1002 and is called and executed by the processor 1001 to execute the seed germination detection method of the embodiments of this application. Input / output interface 1003 is used to implement information input and output; The communication interface 1004 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 1005 transmits information between various components of the device (e.g., processor 1001, memory 1002, input / output interface 1003, and communication interface 1004); The processor 1001, memory 1002, input / output interface 1003 and communication interface 1004 are connected to each other within the device via bus 1005.

[0071] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described seed germination detection method.

[0072] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0073] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0074] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0075] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0076] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0077] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0078] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0079] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0080] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0081] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0082] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0083] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for detecting seed germination, characterized in that, The method includes: Acquire a collection of images showing the germination process of multiple target seeds and germination detection prompt text; wherein, the collection of images showing the germination process includes multiple global image frame data acquired in chronological order; Seed individual image segmentation is performed based on multiple global image frame data to obtain local sampling region data for each target seed; Based on the local sampling region data, multiple local image frame data corresponding to each target seed are segmented from the multiple global image frame data; For each target seed, pixel difference germination detection is performed based on multiple local image frame data to obtain pixel germination frame state data; For each target seed, semantic germination detection is performed based on multiple local image frame data and the germination detection prompt text to obtain semantic germination frame state data; The target germination detection result of the target seed is determined based on the pixel germination frame state data and the semantic germination frame state data.

2. The method according to claim 1, characterized in that, The step of segmenting multiple local image frame data corresponding to each target seed from multiple global image frame data based on the local sampling region data includes: Understanding text in the germination scenario of multiple target seeds; Based on multiple global image frame data and the germination scene understanding text, seed scene parsing is performed to obtain seed layout parsing data; Based on the local sampling region data, multiple candidate local image frame data corresponding to each target seed are segmented from the multiple global image frame data; For each target seed, the image segmentation effect is verified based on the seed layout parsing data and multiple candidate local image frame data to obtain segmentation effect verification data; In response to the segmentation effect verification data satisfying the preset image segmentation effect conditions, multiple candidate local image frame data are determined as multiple local image frame data of the target seed.

3. The method according to claim 2, characterized in that, The step of segmenting multiple local image frame data corresponding to each target seed from multiple global image frame data based on the local sampling region data further includes: In response to the segmentation effect verification data not meeting the preset image segmentation effect conditions, segmentation difference analysis is performed based on the seed layout parsing data and multiple candidate local image frame data to obtain segmentation difference quantification data. The parameters of the local sampling region data are adjusted based on the segmentation difference quantification data to obtain the parameter-adjusted local sampling region data; Based on the local sampling region data adjusted according to the parameters, multiple candidate local image frame data corresponding to each target seed are re-segmented from multiple global image frame data until the segmentation effect verification data corresponding to the multiple candidate local image frame data obtained by re-segmentation meets the preset image segmentation effect conditions, and the multiple candidate local image frame data are determined as multiple local image frame data of the target seed.

4. The method according to claim 1, characterized in that, The step of segmenting individual seed images based on multiple global image frame data to obtain local sampling region data for each target seed includes: Color thresholding is performed on multiple global image frame data to obtain multiple masked global image frame data; Connectivity analysis is performed on multiple masked global image frame data to obtain local sampling region data for each target seed.

5. The method according to claim 4, characterized in that, The step of performing color threshold segmentation based on multiple global image frame data to obtain multiple masked global image frame data includes: Color space conversion is performed on multiple global image frame data to obtain multiple color space image data; Threshold segmentation is performed on multiple color space image data to obtain multiple candidate mask global image frame data; Morphological optimization is performed on multiple candidate mask global image frame data to obtain multiple mask global image frame data.

6. The method according to claim 1, characterized in that, For each target seed, pixel difference germination detection is performed based on multiple local image frame data to obtain pixel germination frame state data, including: For each target seed, the first frame data of the local image and the subsequent frame data of the local image are determined based on the multiple local image frame data; For each subsequent frame of the local image, pixel difference analysis is performed based on the first frame of the local image and the subsequent frame of the local image to obtain the pixel difference data corresponding to the subsequent frame of the local image. Pixel budding detection is performed based on the pixel difference data corresponding to subsequent frames of multiple local images to obtain pixel budding frame state data.

7. The method according to claim 6, characterized in that, For each target seed, semantic germination detection is performed based on multiple local image frame data and the germination detection prompt text to obtain semantic germination frame state data, including: For each subsequent frame of a local image for each target seed, semantic germination status detection is performed based on the first frame of the local image, the subsequent frame of the local image, and the germination detection prompt text to obtain semantic germination status data corresponding to the subsequent frame of the local image. Semantic germination detection is performed based on the semantic germination state data corresponding to the subsequent frame data of multiple local images to obtain semantic germination frame state data.

8. A seed germination detection device, characterized in that, The device includes: The seed data acquisition unit is used to acquire a collection of images of the germination process of multiple target seeds and germination detection prompt text; wherein, the collection of images of the germination process includes multiple global image frame data acquired in chronological order; A local sampling and localization unit is used to segment individual seed images based on multiple global image frame data to obtain local sampling region data for each target seed; A local image segmentation unit is used to segment multiple local image frame data corresponding to each target seed from multiple global image frame data based on the local sampling region data; A pixel germination detection unit is used to perform pixel difference germination detection for each of the target seeds based on multiple local image frame data to obtain pixel germination frame status data. A semantic germination detection unit is used to perform semantic germination detection for each target seed based on multiple local image frame data and the germination detection prompt text, and obtain semantic germination frame state data. The germination result determination unit is used to determine the target germination detection result of the target seed based on the pixel germination frame state data and the semantic germination frame state data.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the seed germination detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the seed germination detection method according to any one of claims 1 to 7.

Images