Seedling condition information determination method and device, equipment and storage medium

Abstract

The invention discloses a seedling condition information determination method and device, equipment and a storage medium. The method comprises the steps of performing conventional segmentation on an acquired unmanned aerial vehicle image to obtain a conventional segmented image set, and performing dislocation segmentation on the acquired unmanned aerial vehicle image to obtain a dislocation segmented image set; determining a first image set according to the conventional segmentation image set, and determining a second image set according to the error segmentation image set; according to a pre-trained crop seedling identification model, identifying the first image set to obtain a first seedling identification result, and identifying the second image set to obtain a second seedling identification result; and determining seedling condition information according to the first seedling identification result and the second seedling identification result. According to the technical scheme of the embodiment, the crop seedling identification model is pre-trained, the two segmented images of the unmanned aerial vehicle image data are identified, the seedling condition information of the target crop field row crops is determined, and crop field seedling condition monitoring based on the unmanned aerial vehicle image is realized.

Description

Technical Field

[0001] This invention relates to the field of seedling monitoring technology, and in particular to a method, apparatus, equipment and storage medium for determining seedling information. Background Technology

[0002] Crop condition monitoring is an important component of smart agriculture implementation. Effective crop condition monitoring helps in precise field management, such as replanting missing seedlings, managing weak seedlings, and preventing diseases and pests, which is crucial to the formation of crop yield and quality.

[0003] Currently, my country mainly relies on manual inspections for crop growth monitoring, which is costly, inefficient, and difficult to adapt to the development of large-scale agriculture. Existing crop growth monitoring methods are based on satellite remote sensing imagery; however, satellite remote sensing imagery has low resolution, and crop growth monitoring based on satellite remote sensing imagery can only be carried out at the population level, and cannot achieve high-precision monitoring at the individual level.

[0004] In addition, current seedling condition monitoring methods can only monitor seedlings with abnormal growth in a small area, and the seedling condition monitoring function is relatively simple, making it impossible to carry out refined monitoring of missing seedlings.

[0005] In recent years, with the rapid development of drone technology, agricultural drones have gradually become an important tool for smart agriculture in crop monitoring. Drones can be equipped with various types of sensors to acquire crop growth and development status and field background information from multiple dimensions in a short period of time. Therefore, there is an urgent need for a seedling monitoring method based on drone imagery to monitor crop seedling conditions in the field and meet the needs of smart agriculture development. Summary of the Invention

[0006] This invention provides a method, apparatus, device, and storage medium for determining crop seedling conditions, thereby enabling crop seedling monitoring based on UAV imagery and improving the accuracy of seedling condition information.

[0007] In a first aspect, embodiments of the present invention provide a method for determining crop growth information, the method comprising:

[0008] The process involves acquiring UAV images of a target crop field, performing conventional segmentation on the UAV images to obtain a conventional segmented image set, and performing misaligned segmentation on the UAV images to obtain a misaligned segmented image set. The conventional segmentation is performed by dividing the image into a grid starting from the vertices of the UAV images, while the misaligned segmentation is performed by dividing the image into a grid starting from the non-vertices of the UAV images.

[0009] Based on the conventional segmented image set, a first image set is determined, and based on the erroneous segmented image set, a second image set is determined. The first image set is the data after superimposing the conventional segmented image set with the first crop row mask, and the second image set is the data after superimposing the misaligned segmented image set with the second crop row mask.

[0010] Based on the pre-trained crop seedling recognition model, the first image set is identified to obtain the first seedling recognition result, and based on the pre-trained crop seedling recognition model, the second image set is identified to obtain the second seedling recognition result.

[0011] Based on the first seedling identification result and the second seedling identification result, seedling condition information is determined.

[0012] Secondly, embodiments of the present invention also provide a crop condition information determination device, the device comprising:

[0013] The data acquisition module is used to acquire UAV images collected for the target crop field, perform conventional segmentation on the UAV images to obtain a conventional segmented image set, and perform misaligned segmentation on the UAV images to obtain a misaligned segmented image set; the conventional segmentation is performed by dividing the image grid starting from the vertices of the UAV images, and the misaligned segmentation is performed by dividing the image grid starting from the non-vertices in the UAV images.

[0014] The image set determination module is used to determine a first image set based on the conventional segmented image set, and to determine a second image set based on the erroneous segmented image set. The first image set is the data after superimposing the conventional segmented image set with a first crop row mask, and the second image set is the data after superimposing the misaligned segmented image set with a second crop row mask.

[0015] The seedling recognition module is used to identify the first image set according to the pre-trained crop seedling recognition model to obtain the first seedling recognition result, and to identify the second image set according to the pre-trained crop seedling recognition model to obtain the second seedling recognition result.

[0016] The seedling condition information determination module is used to determine seedling condition information based on the first seedling identification result and the second seedling identification result.

[0017] Thirdly, embodiments of the present invention also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the crop condition information determination method as described in any of the embodiments of the present invention.

[0018] Fourthly, embodiments of the present invention also provide a storage medium for storing computer-executable instructions, which, when executed by a computer processor, are used to perform the crop condition information determination method as described in any of the embodiments of the present invention.

[0019] The technical solution of this invention uses a pre-trained crop seedling recognition model to identify two segmented images of UAV image data, thereby determining the seedling condition information of the target crop in the field and realizing crop seedling condition monitoring based on UAV imagery.

[0020] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a method for determining crop growth information provided in Embodiment 1 of the present invention;

[0023] Figure 2 This is a flowchart of a method for determining crop growth information provided in Embodiment 2 of the present invention;

[0024] Figure 3 This is a schematic diagram of four image types obtained after preprocessing of UAV images provided in Embodiment 2 of the present invention;

[0025] Figure 4 This is a schematic diagram of a 4-channel image of a superimposed crop row mask provided in Embodiment 2 of the present invention;

[0026] Figure 5 This is a schematic diagram illustrating the prediction effect of the crop seedling identification model provided in Embodiment 2 of the present invention;

[0027] Figure 6 This is a schematic diagram illustrating conventional segmentation and misaligned segmentation of UAV images provided in Embodiment 2 of the present invention;

[0028] Figure 7 This is a schematic diagram of the crop row outline from a drone image provided in Embodiment 2 of the present invention;

[0029] Figure 8 This is a schematic diagram of the two crop seedling identification results provided in Embodiment 2 of the present invention in UAV imagery;

[0030] Figure 9 This is a schematic diagram of the merged seedling identification results provided in Embodiment 2 of the present invention;

[0031] Figure 10 This is a schematic diagram of missing seedling location provided in Embodiment 2 of the present invention;

[0032] Figure 11 This is a schematic diagram of the crop row outline and crop seedling identification results provided in Embodiment 2 of the present invention.

[0033] Figure 12 This is a schematic diagram of the structure of a crop condition information determination device provided in Embodiment 3 of the present invention;

[0034] Figure 13 This is a schematic diagram of the structure of an electronic device that implements the seedling information determination method of this invention. Detailed Implementation

[0035] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0036] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention 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 the invention 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 a 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.

[0037] Example 1

[0038] Figure 1 This is a flowchart illustrating a method for determining crop growth information according to Embodiment 1 of the present invention. This embodiment is applicable to situations requiring crop growth information determination. The method can be executed by a crop growth information determination device, which can be implemented in hardware and / or software. This device can be configured in any electronic device with network communication and computing capabilities. Figure 1As shown, the method includes:

[0039] S110. Acquire UAV images collected for the target crop field, perform conventional segmentation on the UAV images to obtain a conventional segmentation image set, and perform misaligned segmentation on the UAV images to obtain a misaligned segmentation image set; the conventional segmentation is performed by dividing the image grid starting from the vertices of the UAV images, and the misaligned segmentation is performed by dividing the image grid starting from the non-vertices in the UAV images.

[0040] In this embodiment, the drone imagery refers to image data of the target crop field collected by multiple types of sensors mounted on the drone. In practical applications, the types of sensors mounted on the drone include hyperspectral cameras, multispectral cameras, or visible light cameras, etc., resulting in drone imagery of the target crop field collected by these multiple types of sensors.

[0041] In this model, conventional segmentation divides the image into grids starting from the vertices of the UAV image, while misaligned segmentation divides the image into grids starting from the non-vertices of the UAV image. The conventional segmentation image set is an image dataset composed of small-sized UAV images after conventional segmentation, and the misaligned segmentation image set is an image dataset composed of small-sized UAV images after misaligned segmentation.

[0042] In this embodiment, the drone images include different types of drone image data, which can be used to identify crop rows in different types of drone images, thereby improving the image application scenarios for crop row information recognition.

[0043] As an optional but not limited implementation, after acquiring drone imagery of the target crop field, the following steps are included:

[0044] The UAV imagery is preprocessed, including but not limited to at least one of the following: image segmentation, hyperspectral dimensionality reduction, and band combination. The UAV imagery includes hyperspectral images, visible light images, and multispectral images.

[0045] In this embodiment, the UAV images of the crop field during the crop growth period are images of any stage of crop growth and development in the field. Typically, crops in the field are planted in a straight line with regular row spacing.

[0046] In this embodiment, the types of UAV images collected for the target crop field include hyperspectral images, visible light images, and multispectral images. Among them, hyperspectral images contain dozens to hundreds of bands, have low spatial resolution, and large data volume; visible light images contain only 3 bands, have small data volume, are easy to process and store, and can intuitively reflect the surface visual characteristics of crops such as color and shape; multispectral images contain about ten bands, each band reflecting the reflectance characteristics of crops at a specific wavelength.

[0047] It should be noted that hyperspectral images are characterized by large memory usage, high data dimensionality, and rich image information. To address these characteristics, dimensionality reduction and band combination processing are required to enable unified data analysis and identification of different types of UAV images after image processing.

[0048] If the acquired UAV imagery is a hyperspectral image, it needs to undergo dimensionality reduction and band combination processing. Specifically, the dimensionality reduction process involves uniformly collecting pixels from the hyperspectral image at a ratio of 0.5‰ of the total pixels, extracting their spectral information, and training a PCA model based on the extracted spectral information using Principal Component Analysis (PCA). The number of principal components, k, is set to 3, reducing the hyperspectral image dimension from n dimensions to 3 dimensions. The values ​​of each dimension are then scaled from 0 to 1 to 0 to 255, resulting in the PCA-reduced composite image. Further, the band combination processing involves selecting the red (680nm), green (550nm), blue (450nm), and near-infrared (NIR) bands of the hyperspectral image, and superimposing the values ​​of the red, green, and blue bands to obtain an RGB composite image. The NDVI (Normalized Difference Vegetation Index) and EVI (Enhanced Vegetation Index) are calculated using the following formulas, and corresponding images are generated.

[0049] NDVI = (NIR - Red) / (NIR + Red);

[0050] EVI=2.5*(NIR-Red) / (NIR+6*Red-7.5*Blue+1);

[0051] Furthermore, for visible light images and vegetation index images obtained from hyperspectral band combinations or UAV acquisition, the number of bands in the images is unified to 3 through channel cropping or channel duplication. If the visible light image includes a transparency channel, the transparency channel is cropped to unify the number of image channels to 3. Vegetation index images are often single-channel, with pixel values ​​ranging from 0 to 1; therefore, the number of image channels can be unified to 3 through channel duplication, and the image pixel values ​​can be scaled to 0–255. In this embodiment, after preprocessing and cropping various types of UAV images, four types of small-sized image datasets can be obtained.

[0052] In this embodiment, by specifying corresponding preprocessing procedures for different types of data and unifying the data format, the potential value of the data is fully explored while ensuring the compatibility of this crop growth information determination method with different types of UAV images.

[0053] As an optional but not limited implementation, the UAV imagery is subjected to conventional segmentation to obtain a conventionally segmented image set, and the UAV imagery is subjected to misaligned segmentation to obtain a misaligned segmented image set, including:

[0054] Starting from the first starting point, the UAV imagery is divided into grids according to a preset grid size to obtain a conventional segmented image set;

[0055] Starting from the second starting point, the UAV image is divided into grids according to a preset grid size to obtain a misaligned segmented image set. The interval between the first starting point and the second starting point is half a preset grid size.

[0056] In this embodiment, since large images contain a lot of information, existing image processors have difficulty processing large images and cannot accurately identify image information.

[0057] Therefore, in this embodiment, the original UAV imagery is segmented into smaller images to facilitate subsequent data processing and obtain accurate image information.

[0058] Furthermore, in this embodiment, the acquired UAV images are divided into conventional segmentation and misaligned segmentation methods to obtain conventional segmented image sets and misaligned segmented image sets.

[0059] Specifically, conventional segmentation starts with image vertices and divides the UAV imagery into a grid according to a preset grid size, resulting in a conventional segmented image set consisting of multiple images of the same size. Misaligned segmentation starts with image vertices offset by half a preset grid size in the horizontal and vertical coordinate directions, and divides the UAV imagery into a grid according to the preset grid size, resulting in a misaligned segmented image set consisting of multiple smaller images.

[0060] In this embodiment, the UAV images are divided using two image segmentation methods to obtain a conventional segmentation image set and a misaligned segmentation image set. The conventional segmentation image set and the misaligned segmentation image set are used to cross-validate the results of subsequent crop seedling condition identification, thereby improving the accuracy of crop seedling condition identification.

[0061] S120. Based on the conventional segmented image set, determine a first image set, and based on the erroneous segmented image set, determine a second image set, wherein the first image set is the data after superimposing the conventional segmented image set with the first crop row mask, and the second image set is the data after superimposing the misaligned segmented image set with the second crop row mask.

[0062] In this embodiment, the first image set is the data superimposed with the conventional segmentation image set and the first crop row mask, and the second image set is the data superimposed with the misaligned segmentation image set and the second crop row mask. The first crop row mask is a crop row mask for recognizing crop rows in the conventional segmentation image set, and the second crop row mask is a crop row mask for recognizing crop rows in the misaligned segmentation image set.

[0063] Furthermore, the first cropping mask is superimposed on a small image of the regular segmentation image set to obtain a first image set, which consists of 4-channel images. The second cropping mask is superimposed on a small image of the incorrectly segmented image set to obtain a second image set, which also consists of 4-channel images.

[0064] As an optional but not limited implementation, determining a first image set based on the conventionally segmented image set, and determining a second image set based on the incorrectly segmented image set, includes:

[0065] Determine the first crop row mask for the conventional segmented image set and the second crop row mask for the misaligned segmented image set;

[0066] A first image set is determined based on the conventional segmented image set and the first crop row mask, and a second image set is determined based on the misaligned segmented image set and the second crop row mask.

[0067] In this embodiment, the first image set and the second image set can be determined using the methods described above. Specifically, a pre-trained crop row recognition model can be used to identify the conventionally segmented image set to determine the first crop row mask, and the pre-trained crop row recognition model can be used to identify the incorrectly segmented image set to determine the second crop row mask. The pre-trained crop row recognition model is used to identify and determine the crop row masks in the images.

[0068] Furthermore, the first cropping mask is superimposed on a small image of the regular segmentation image set to obtain a first image set, which consists of 4-channel images. The second cropping mask is superimposed on a small image of the incorrectly segmented image set to obtain a second image set, which also consists of 4-channel images.

[0069] S130. Based on the pre-trained crop seedling recognition model, the first image set is identified to obtain the first seedling recognition result, and based on the pre-trained crop seedling recognition model, the second image set is identified to obtain the second seedling recognition result.

[0070] In this embodiment, the pre-trained crop seedling recognition model is a pre-trained neural network model used to recognize row crop seedlings.

[0071] As an optional but not limited implementation, the process of determining the pre-trained crop seedling recognition model includes:

[0072] Determine the neural network model for crop seedling recognition and the model training dataset; the model training dataset includes crop row masks and corresponding crop seedling annotation data, and includes a training set, a test set, and a validation set; the crop seedling annotation data is the minimum bounding rectangle recognition box of the crop seedling;

[0073] The training set was used to train the neural network model for identifying crop seedlings;

[0074] The test set is identified based on the trained crop seedling recognition neural network model;

[0075] The loss value is determined based on the crop seedling identification results and crop seedling annotation data in the test set;

[0076] Based on the loss value, the loss function of the crop seedling recognition neural network model is optimized until the loss value is less than a preset threshold or the number of training iterations is reached.

[0077] Using the validation set, the hyperparameters of the optimized crop seedling recognition neural network model are adjusted to obtain a pre-trained crop seedling recognition model.

[0078] It should be noted that the input to the pre-trained crop seedling recognition model is the crop row mask and the crop seedling annotation data corresponding to the crop row mask, where the crop seedling annotation data is the minimum bounding rectangle recognition box of the crop seedling.

[0079] Furthermore, the pre-trained crop seedling recognition model is used to identify the first image set to obtain the first seedling recognition result, and the pre-trained model is used to identify the second image set to obtain the second seedling recognition result. The first seedling recognition result is the minimum bounding rectangle of the crop seedlings in the first image set, and the second seedling recognition result is the minimum bounding rectangle of the crop seedlings in the second image set.

[0080] S140. Determine seedling condition information based on the first seedling identification result and the second seedling identification result.

[0081] In this embodiment, the first seedling identification result is the minimum bounding rectangle identification box of each crop seedling in each small-sized image of the conventional segmented image set, and the first seedling identification result is the minimum bounding rectangle identification box of each crop seedling in each small-sized image of the misaligned segmented image set.

[0082] Furthermore, by cross-verifying the results of the first and second seedling identifications, the seedling identification results of the target crop field can be determined. By analyzing the seedling identification results, the location and number of missing seedlings in the target crop field can be determined.

[0083] In practical applications, based on the seedling identification results, the length, width, and area of ​​the minimum bounding rectangle of each seedling can be calculated. Based on the minimum bounding rectangles, images of each seedling are cropped, and high-level features are extracted using a pre-trained deep convolutional neural network. The seedling morphological features and high-level features are then combined, and the seedlings are classified using a clustering algorithm.

[0084] Furthermore, based on the number of seedling minimum bounding rectangle recognition boxes and the clustering results, the total number of seedlings and the proportion of seedlings in each category are calculated. The total number of seedlings is the number of seedling minimum bounding rectangle recognition boxes, and the number of seedlings in each category is divided by the total number of seedlings to obtain the proportion of seedlings in each category.

[0085] Furthermore, the number of missing seedlings in the original UAV images is summed with the actual number of seedlings to calculate the theoretical number of seedlings. The number of missing seedlings in the original UAV images is divided by the theoretical number of seedlings to obtain the seedling shortage rate. For the different categories of seedlings obtained from clustering, their length, width, and area are statistically analyzed, and analysis of variance and multiple comparisons are performed. The average plant spacing, number of missing seedlings, seedling shortage rate, actual number of seedlings, theoretical number of seedlings, number and proportion of different categories of seedlings, analysis of variance and multiple comparison results are compiled and summarized to generate a seedling monitoring result report.

[0086] Furthermore, the obtained crop row outlines are plotted on the original UAV image; the identified missing seedling sites are plotted on the original image; and the different categories of seedlings obtained from clustering are plotted on the original image using different colored recognition boxes to generate a seedling monitoring visualization result.

[0087] The technical solution of this invention uses a pre-trained crop seedling recognition model to identify two segmented images of UAV images, thereby determining the seedling information of the target crop in the field and realizing crop seedling monitoring based on UAV images.

[0088] Example 2

[0089] Figure 2 This is a flowchart illustrating a method for determining crop growth information according to Embodiment 2 of the present invention. This embodiment further specifies the method based on the previous embodiment. This embodiment is applicable to situations requiring crop growth information determination. The method can be executed by a crop growth information determination device, which can be implemented in hardware and / or software. This device can be configured in any electronic device with network communication and computing capabilities. Figure 2 As shown, the method includes:

[0090] S210. Acquire UAV images collected for the target crop field, perform conventional segmentation on the UAV images to obtain a conventional segmentation image set, and perform misaligned segmentation on the UAV images to obtain a misaligned segmentation image set; the conventional segmentation is performed by dividing the image grid starting from the vertices of the UAV images, and the misaligned segmentation is performed by dividing the image grid starting from the non-vertices in the UAV images.

[0091] S220. Based on the conventional segmented image set, determine a first image set, and based on the erroneous segmented image set, determine a second image set, wherein the first image set is the data after superimposing the conventional segmented image set with the first crop row mask, and the second image set is the data after superimposing the misaligned segmented image set with the second crop row mask.

[0092] S230. Based on the pre-trained crop seedling recognition model, the first image set is identified to obtain the first seedling recognition result, and based on the pre-trained crop seedling recognition model, the second image set is identified to obtain the second seedling recognition result.

[0093] S240. The first seedling identification result is stitched together to obtain the first seedling identification frame information of the UAV image, and the second seedling identification result is stitched together to obtain the second seedling identification frame information of the UAV image.

[0094] In this embodiment, the first seedling identification box information is the seedling identification result obtained by splicing the first seedling identification results, and the second seedling identification box information is the seedling identification result obtained by splicing the second seedling identification results. Specifically, the first seedling identification box information and the second seedling identification box information are the minimum bounding rectangle identification boxes of each seedling.

[0095] It should be noted that by stitching the crop seedling identification results of each small-sized image to the original image size according to its segmentation method, the first seedling identification result and the second crop seedling identification result of the complete UAV image are obtained.

[0096] The first seedling identification result and the second crop seedling identification result are in the form of rectangular identification boxes. Each rectangular identification box is used to mark whether a seedling exists at the corresponding position. The crop seedling identification result can be determined by the first seedling identification result and the second crop seedling identification result of the complete UAV image.

[0097] S250. Based on the first seedling identification box information and the second seedling identification box information, determine the seedling identification results for each crop row in the target crop field.

[0098] In this embodiment, the first and second seedling identification box information of the stitched complete UAV image are analyzed, and the two crop seedling identification results are merged into the original UAV image. Based on the merged seedling identification results, the crop seedling information of the target crop field is determined.

[0099] As an optional but not limited implementation, the seedling identification results for each crop row in the target crop field are determined based on the first seedling identification box information and the second seedling identification box information, including:

[0100] The overlap is determined based on the information of the first seedling identification frame and the information of the second seedling identification frame;

[0101] Based on the overlap, the seedling identification box information for each crop row in the target crop field is determined.

[0102] In this embodiment, the overlap degree is used to represent the degree of overlap between seedling recognition boxes. The higher the overlap degree, the more accurate the two seedling recognition results are. Low overlap degree may be due to reasons such as image misalignment or differences in the position of the recognition box generation. It is necessary to further determine the final crop seedling recognition result based on the overlap degree.

[0103] As an optional but not limited implementation, the seedling identification box information for each crop row in the target crop field is determined based on the overlap, including:

[0104] The first seedling identification frame and the second seedling identification frame with an overlap greater than or equal to a first preset threshold are merged to obtain merged seedling identification frame information; the first seedling identification frame is the seedling identification frame in the first seedling identification frame information, and the second seedling identification frame is the seedling identification frame in the second seedling identification frame information.

[0105] Determine the distance between the center point of the first seedling identification box and the center point of the second seedling identification box in the merged seedling identification box information;

[0106] The first seedling identification frame and the second seedling identification frame with a distance less than the second preset threshold are merged to obtain the seedling identification frame information of each crop row in the target crop field.

[0107] In this embodiment, the first preset threshold can be preset, the first seedling identification box is the seedling identification box in the first seedling identification box information, and the second seedling identification box is the seedling identification box in the second seedling identification box information.

[0108] Specifically, based on the position of the crop seedling identification boxes in the original UAV imagery, the results of the two crop seedling identification attempts are aggregated into the original UAV imagery. A first merging process is performed based on the overlap of the identification boxes, merging identification results with an overlap greater than a certain threshold into a single identification box. Based on this first merging, a second merging process is performed using a clustering algorithm to merge identification boxes with a distance less than a certain threshold, obtaining the crop seedling identification results for the original image. The distance between identification boxes can be determined based on the distance between the center points of each identification box.

[0109] S260. Based on the seedling identification results and crop row outlines, determine the seedling condition information of the target crop field.

[0110] In this embodiment, the crop row outline is obtained by recognizing the crop row mask image. Based on the crop row outline and crop seedling recognition results of the original image, the data is summarized into the same data file through layer overlay. Crop row outlines that do not contain any crop seedling recognition boxes, as well as crop seedling recognition boxes that are not in the crop row outline, are removed.

[0111] As an optional but not limited implementation, determining a first image set based on the conventionally segmented image set, and determining a second image set based on the incorrectly segmented image set, includes:

[0112] The crop row mask of the target crop field is determined based on the union of the first crop row mask and the second crop row mask.

[0113] The crop row mask is identified to determine at least one crop row outline.

[0114] In this embodiment, the crop row outline is the outline of the region where the pixel is 1 or 0 in the crop row mask information.

[0115] In this embodiment, the crop row masks are merged to determine the final crop row mask result, and the crop row outline is determined based on the crop row mask. The crop seedling identification results are then verified to verify their accuracy based on the crop row outline.

[0116] As an optional but not limited implementation, the seedling condition information of the target crop field is determined based on the seedling identification results and crop row outlines, including:

[0117] For each crop row, determine the coordinates of the first midpoint and the second midpoint of the short side of the minimum bounding rectangle of the crop row outline, as well as the coordinates of at least one center point corresponding to each seedling identification box in the seedling identification box information of the crop row.

[0118] Based on the coordinates of the first midpoint, the coordinates of each center point, and the coordinates of the second midpoint, at least two line segments are determined;

[0119] Based on the average length of the line segments, the seedling condition information of the target crop field is determined, including the location and number of missing seedlings.

[0120] In this embodiment, the minimum bounding rectangle of the crop row outline is obtained by traversing the crop row outline in the original UAV image. The coordinates of the midpoint of the short side of the minimum bounding rectangle are calculated, and the coordinates of the center point of the seedling identification box within the crop row outline are also calculated. Starting from one of the midpoints of the short side of the minimum bounding rectangle of the crop row, the nearest seedling center point within the crop row is connected sequentially until the other midpoint of the short side of the minimum bounding rectangle of the crop row is connected as the endpoint. The lengths of all connecting line segments are counted, sorted, and the truncated mean is calculated. The truncated mean is used as the length of the average plant spacing.

[0121] Furthermore, the length of all line segments is traversed to determine if it exceeds the average plant spacing or a specified multiple thereof. If it does, the line segment is considered to have missing seedlings. The number of missing seedlings is calculated by dividing the line segment length by the average plant spacing, and the location of the missing seedlings is the coordinates of the points on the corresponding line segment where the missing seedlings are evenly distributed. The number of missing seedlings in the original UAV image is the sum of the number of missing seedlings for all missing line segments.

[0122] The technical solution of this invention uses a pre-trained crop seedling recognition model to identify two segmented images of UAV images, thereby determining the seedling information of the target crop in the field and realizing crop seedling monitoring based on UAV images.

[0123] Specific application examples

[0124] The following is a specific application example of this embodiment, where flue-cured tobacco is selected as the crop. 7–15 days after transplanting, multi-rotor or fixed-wing drones are used to collect images above the tobacco field, employing manual or pre-defined flight paths. The drone's flight altitude is set according to the resolution of its onboard sensors to ensure that the tobacco seedlings appear complete and detailed in the acquired drone images. The types of sensors onboard the drone include visible light, multispectral, and hyperspectral sensors. The collected images are then stitched together using a professional image stitching program to obtain complete drone images of the tobacco seedling stage.

[0125] Furthermore, the acquired UAV imagery includes visible light, multispectral, and hyperspectral data. Taking advantage of the large memory size, high dimensionality, and rich information of hyperspectral data, dimensionality reduction and band combination processing were performed on it.

[0126] In hyperspectral images, pixels are uniformly collected at a ratio of 0.5‰ of the total pixels, and spectral information is extracted. Principal Component Analysis (PCA) is used to train a PCA model based on the extracted spectral information, with the number of principal components (k) set to 3. For each small-sized image, the PCA model is applied to perform feature dimensionality reduction, reducing the image dimension from n dimensions to 3 dimensions. The values ​​of each dimension are scaled from 0 to 1 to 0 to 255, resulting in the PCA-reduced composite image (PCA-CI).

[0127] The red (680nm), green (550nm), blue (450nm), and near-infrared (NIR) bands of hyperspectral data were selected. The values ​​of the red, green, and blue bands were superimposed to obtain an RGB composite image. The NDVI and EVI vegetation indices were calculated using the following formulas, and corresponding images were generated.

[0128] NDVI = (NIR - Red) / (NIR + Red);

[0129] EVI=2.5*(NIR-Red) / (NIR+6*Red-7.5*Blue+1);

[0130] For visible light images and vegetation index images obtained from hyperspectral band combinations or UAV acquisition, the number of bands in the image is unified to 3 through channel cropping or channel duplication. If the visible light image includes a transparency channel, the transparency channel is cropped to unify the number of image channels to 3. Vegetation index images are often single-channel with pixel values ​​ranging from 0 to 1; therefore, the number of image channels is unified to 3 through channel duplication, and the image pixel values ​​are scaled to 0–255. Four types of small-sized images are obtained after preprocessing and cropping of various types of UAV imagery. See [link to documentation]. Figure 3 The diagram shows four image types obtained after preprocessing of UAV images.

[0131] Furthermore, all types of small-sized images are aggregated into a single dataset for crop row and seedling annotation. Crop row annotation uses image segmentation software to annotate crop rows on each small-sized image and generate corresponding mask images. Similarly, crop seedling annotation software is used to annotate crop seedlings on each small-sized image and generate corresponding annotation files. The annotated dataset contains four types of data, and stratified random sampling is used to evenly divide each type of dataset into training, validation, and test sets in an 8:1:1 ratio.

[0132] Furthermore, the trained crop row recognition model is used to predict crop row patterns in the images of the training dataset, obtaining corresponding crop row masks. These predicted crop row masks are then overlaid onto the corresponding images in the training dataset to generate a 4-channel dataset. See also... Figure 4 The image shown is a schematic diagram of a 4-channel image with a crop row mask superimposed.

[0133] Furthermore, based on the synthesized 4-channel dataset, a crop seedling recognition model was trained to obtain the crop seedling recognition results for each small-sized image. See also Figure 5 The diagram shows the prediction performance of the crop seedling identification model.

[0134] See Figure 6 The diagram illustrates conventional and misaligned segmentation of drone images. In this embodiment, the data preparation process involves performing conventional and misaligned segmentation on different types of drone images. Conventional segmentation starts from the top-left corner of the image and segments the drone image according to a segmentation step size, resulting in a conventionally segmented image set. Misaligned segmentation starts from the top-left corner of the image, offsetting it by 1 / 2 of the segmentation step size along both the x and y axes, to obtain the image segmentation starting point. The drone image is then segmented using a preset segmentation step size, resulting in a misaligned segmented image set.

[0135] Further, the crop row segmentation and recognition results are summarized and denoised: the crop row recognition results obtained from the two segmentation steps are stitched onto the original UAV image. The stitched crop row layers are then overlaid, and the union of the two recognition results is used as the crop row recognition result for the original image. The crop row contours are determined based on these results, and denoising is applied to them. By appropriately selecting the width of the crop row contours, small-scale noise in the crop row mask image is effectively removed, improving the saliency of the target region and providing clearer and more accurate input data for subsequent image analysis or processing. See also... Figure 7 The image shown is a schematic diagram of the crop row outline from a drone image.

[0136] Furthermore, the crop seedling identification results obtained from the two segmentations are plotted on the original UAV imagery, see [link to relevant documentation]. Figure 8 The image shows a schematic diagram of the results of two crop seedling identifications in UAV imagery.

[0137] In this embodiment, overlapping crop seedling identification boxes can be merged based on the Intersection over Union (IoU) method. For two crop seedling identification boxes, their IoU is calculated. If the IoU exceeds a preset threshold (default 0.5), they are considered to belong to the same crop seedling, and the merging method is to take the bounding rectangle of the two crop seedling identification boxes. For adjacent but not completely overlapping boxes, the identification boxes are further merged using the DBSCAN clustering algorithm. The clustering parameter eps (i.e., neighborhood radius) is adaptively calculated based on the average distance between the center points of adjacent boxes, and the distance threshold can be effectively adjusted according to the spatial distribution of the detection boxes. DBSCAN clustering groups crop seedling identification boxes that are close to each other into one class, and then calculates the minimum bounding rectangle of the crop seedling identification boxes in each class to generate merged crop seedling identification boxes, thereby reducing false detection boxes caused by small-scale movement or segmentation problems.

[0138] In practical applications, based on the merged crop seedling identification bounding box, a semi-transparent rectangle can be drawn on the image, using different colors and transparency to highlight the detection results, and the merged seedling identification label file can be exported. See also Figure 9 This is a schematic diagram of the seedling identification results after merging.

[0139] Furthermore, the crop seedling identification results can be analyzed to determine seedling condition information or generate seedling monitoring reports. The seedling identification tag file is parsed, the coordinates of each crop seedling identification box are extracted, and the coordinates of the center point of each crop seedling identification box are calculated. The center points of all crop seedling identification boxes are matched with the target contours in the crop row contours. For each crop row contour, its minimum bounding rectangle is calculated, and the midpoint of the short side is determined. Then, based on these midpoints, the center points are connected using nearest neighbor connections to form a series of line segments, and the length of each line segment is calculated. The average plant spacing is calculated based on the length of the connecting line segments. By selecting line segments suitable for drawing missing seedlings (length greater than 1.5 times and less than 5 times the average plant spacing), the missing seedlings are calculated and marked on the image, as shown in Figure 15. Finally, the missing seedling data is saved in CSV format, and a report containing analysis information is generated, including the calculated average plant spacing, the number of missing seedlings, etc. See also Figure 10 The diagram shows the location of missing seedlings, and Table 1 records the crop seedling information.

[0140] Table 1 Crop Seedling Status Information

[0141] Seedling monitoring project Seedling monitoring information Average plant spacing (absolute pixel value) 20.77 Number of missing seedlings 97 strains Missing seedling location 1 Coordinates (1337, 1496) Missing seedling location 2 Coordinates (1352, 1497) 3 missing seedling locations Coordinates (1061, 1439) …… ……

[0142] Furthermore, based on the location information in the label file of the crop seedling identification results, images of target regions can be extracted from images of different channels (e.g., target bounding boxes from RGB and near-infrared images). Each extracted target region is stacked into multi-channel samples to ensure that each target crop identification box has information from multiple channels. To ensure that all samples are of consistent size, the maximum width and height of all samples are calculated, and each sample is padded to achieve a consistent size. For each sample, its width, height, area, and confidence score are calculated, and this information is stored along with the image samples.

[0143] Furthermore, a pre-trained ResNet50 model is loaded, and its first convolutional layer is modified to accommodate multi-channel input (e.g., the input image is a multi-channel RGB+NIR image instead of a single-channel RGB image). In the modified ResNet model, the parameters of all convolutional layers are set to non-trainable, so that no parameter updates are performed during the feature extraction stage. Feature extraction is then performed on the padded samples using the modified ResNet50 model. The input samples are converted into PyTorch tensors and fed into the ResNet50 model to obtain the feature vector for each sample.

[0144] The feature vectors extracted by ResNet are fused with the geometric information (such as width, height, and area) of each sample to generate a comprehensive feature vector. This feature fusion combines the visual features of the image with the geometric information of the bounding boxes for more effective clustering. The KMeans algorithm is used to cluster the fused features, dividing the seedlings into several categories (the default is 3 categories). The clustering results are saved, and the category label of each seedling is combined with the bounding box information (such as location, area, and confidence score) to generate a CSV file recording this information. Table 2 shows the seedling features and classification results.

[0145] Table 2. Seedling characteristics and classification results

[0146]

[0147] Furthermore, variance analysis and multiple comparisons were performed on the length, width, and area characteristics of different seedling categories. The total number of theoretical seedlings, the total number of effective seedlings, the missing seedling rate, and the proportion of different seedling categories were statistically analyzed and calculated. The results were then summarized and a seedling monitoring report was generated.

[0148] Draw rectangular boxes for clustered regions on the original image, assign different colors (such as blue, green, and yellow) to different clusters (strong, normal, and weak), and draw crop row outlines and missing seedling sites on the original image to generate visualized seedling monitoring results and export them.

[0149] See Figure 11 The diagram shows the crop row outline and crop seedling identification results, where rectangular identification boxes of different colors represent different seedling categories.

[0150] Example 3

[0151] Figure 12 This is a schematic diagram of a crop seedling condition information determination device provided in Embodiment 3 of the present invention. This embodiment is applicable to situations requiring crop seedling condition information determination. The device can be implemented in hardware and / or software, and can be configured in any electronic device with network communication and computing capabilities. Figure 12 As shown, the device includes:

[0152] The data acquisition module 310 is used to acquire UAV images collected for a target crop field, perform conventional segmentation on the UAV images to obtain a conventional segmented image set, and perform misaligned segmentation on the UAV images to obtain a misaligned segmented image set; the conventional segmentation is performed by dividing the image grid starting from the vertices of the UAV images, and the misaligned segmentation is performed by dividing the image grid starting from the non-vertices in the UAV images.

[0153] The image set determination module 320 is used to determine a first image set based on the conventional segmented image set, and to determine a second image set based on the erroneous segmented image set. The first image set is the data after superimposing the conventional segmented image set with a first crop row mask, and the second image set is the data after superimposing the misaligned segmented image set with a second crop row mask.

[0154] The seedling recognition module 330 is used to recognize a first image set according to a pre-trained crop seedling recognition model to obtain a first seedling recognition result, and to recognize a second image set according to a pre-trained crop seedling recognition model to obtain a second seedling recognition result.

[0155] The seedling condition information determination module 340 is used to determine seedling condition information based on the first seedling identification result and the second seedling identification result.

[0156] Optionally, the UAV imagery is subjected to conventional segmentation to obtain a conventionally segmented image set, and the UAV imagery is subjected to misaligned segmentation to obtain a misaligned segmented image set, including:

[0157] Starting from the first starting point, the UAV imagery is divided into grids according to a preset grid size to obtain a conventional segmented image set;

[0158] Starting from the second starting point, the UAV image is divided into grids according to a preset grid size to obtain a misaligned segmented image set. The interval between the first starting point and the second starting point is half a preset grid size.

[0159] Optionally, seedling condition information is determined based on the first seedling identification result and the second seedling identification result, including:

[0160] The first seedling identification result is stitched together to obtain the first seedling identification box information of the UAV image, and the second seedling identification result is stitched together to obtain the second seedling identification box information of the UAV image.

[0161] Based on the first seedling identification box information and the second seedling identification box information, determine the seedling identification results for each crop row in the target crop field;

[0162] Based on the seedling identification results and crop row outlines, the seedling condition information of the target crop field is determined; wherein, the crop row outlines are obtained by identifying the crop row mask image.

[0163] Optionally, determining a first image set based on the conventionally segmented image set, and determining a second image set based on the incorrectly segmented image set, includes:

[0164] The crop row mask of the target crop field is determined based on the union of the first crop row mask and the second crop row mask.

[0165] The crop row mask is identified to determine at least one crop row outline.

[0166] Optionally, based on the first seedling identification box information and the second seedling identification box information, the seedling identification results for each crop row in the target crop field are determined, including:

[0167] The overlap is determined based on the information of the first seedling identification frame and the information of the second seedling identification frame;

[0168] Based on the overlap, the seedling identification box information for each crop row in the target crop field is determined.

[0169] Optionally, based on the overlap, seedling identification box information for each crop row in the target crop field is determined, including:

[0170] The first seedling identification frame and the second seedling identification frame with an overlap greater than or equal to a first preset threshold are merged to obtain merged seedling identification frame information; the first seedling identification frame is the seedling identification frame in the first seedling identification frame information, and the second seedling identification frame is the seedling identification frame in the second seedling identification frame information.

[0171] Determine the distance between the center point of the first seedling identification box and the center point of the second seedling identification box in the merged seedling identification box information;

[0172] The first seedling identification frame and the second seedling identification frame with a distance less than the second preset threshold are merged to obtain the seedling identification frame information of each crop row in the target crop field.

[0173] Optionally, based on the seedling identification results and crop row outlines, the seedling condition information of the target crop field is determined, including:

[0174] For each crop row, determine the coordinates of the first midpoint and the second midpoint of the short side of the minimum bounding rectangle of the crop row outline, as well as the coordinates of at least one center point corresponding to each seedling identification box in the seedling identification box information of the crop row.

[0175] Based on the coordinates of the first midpoint, the coordinates of each center point, and the coordinates of the second midpoint, at least two line segments are determined;

[0176] Based on the average length of the line segments, the seedling condition information of the target crop field is determined, including the location and number of missing seedlings.

[0177] The seedling information determination device provided in the embodiments of the present invention can execute the seedling information determination method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0178] Example 4

[0179] Figure 13 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0180] like Figure 13As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0181] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0182] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the crop condition information determination method.

[0183] In some embodiments, the crop condition information determination method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the crop condition information determination method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the crop condition information determination method by any other suitable means (e.g., by means of firmware).

[0184] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication unit 19, or installed from storage unit 18, or installed from ROM 12. When the computer program is executed by processor 11, it performs the functions defined in the methods of the embodiments of the present invention.

[0185] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0186] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0187] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0188] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0189] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0190] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through a communication network. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0191] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0192] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for determining crop growth information, characterized in that, include: The process involves acquiring UAV images of a target crop field, performing conventional segmentation on the UAV images to obtain a conventional segmented image set, and performing misaligned segmentation on the UAV images to obtain a misaligned segmented image set. The conventional segmentation is performed by dividing the image into a grid starting from the vertices of the UAV images, while the misaligned segmentation is performed by dividing the image into a grid starting from the non-vertices of the UAV images. Based on the conventional segmented image set, a first image set is determined, and based on the erroneous segmented image set, a second image set is determined. The first image set is the data after superimposing the conventional segmented image set with the first crop row mask, and the second image set is the data after superimposing the misaligned segmented image set with the second crop row mask. Based on the pre-trained crop seedling recognition model, the first image set is identified to obtain the first seedling recognition result, and based on the pre-trained crop seedling recognition model, the second image set is identified to obtain the second seedling recognition result. Based on the first seedling identification result and the second seedling identification result, seedling condition information is determined.

2. The method according to claim 1, characterized in that, Perform conventional segmentation on the UAV imagery to obtain a conventionally segmented image set, and perform misaligned segmentation on the UAV imagery to obtain a misaligned segmented image set, including: Starting from the first starting point, the UAV imagery is divided into grids according to a preset grid size to obtain a conventional segmented image set; Starting from the second starting point, the UAV image is divided into grids according to a preset grid size to obtain a misaligned segmented image set. The interval between the first starting point and the second starting point is half a preset grid size.

3. The method according to claim 1, characterized in that, Based on the first seedling identification result and the second seedling identification result, seedling condition information is determined, including: The first seedling identification result is stitched together to obtain the first seedling identification box information of the UAV image, and the second seedling identification result is stitched together to obtain the second seedling identification box information of the UAV image. Based on the first seedling identification box information and the second seedling identification box information, determine the seedling identification results for each crop row in the target crop field; Based on the seedling identification results and crop row outlines, the seedling condition information of the target crop field is determined; wherein, the crop row outlines are obtained by identifying the crop row mask image.

4. The method according to claim 1, characterized in that, Determining a first image set based on the conventionally segmented image set, and determining a second image set based on the incorrectly segmented image set, includes: The crop row mask of the target crop field is determined based on the union of the first crop row mask and the second crop row mask. The crop row mask is identified to determine at least one crop row outline.

5. The method according to claim 3, characterized in that, Based on the first seedling identification box information and the second seedling identification box information, the seedling identification results for each crop row in the target crop field are determined, including: The overlap is determined based on the information of the first seedling identification frame and the information of the second seedling identification frame; Based on the overlap, the seedling identification box information for each crop row in the target crop field is determined.

6. The method according to claim 5, characterized in that, Based on the overlap, seedling identification box information for each crop row in the target crop field is determined, including: The first seedling identification frame and the second seedling identification frame with an overlap greater than or equal to a first preset threshold are merged to obtain merged seedling identification frame information; the first seedling identification frame is the seedling identification frame in the first seedling identification frame information, and the second seedling identification frame is the seedling identification frame in the second seedling identification frame information. Determine the distance between the center point of the first seedling identification box and the center point of the second seedling identification box in the merged seedling identification box information; The first seedling identification frame and the second seedling identification frame with a distance less than the second preset threshold are merged to obtain the seedling identification frame information of each crop row in the target crop field.

7. The method according to claim 3, characterized in that, Based on the seedling identification results and crop row outlines, determine the seedling condition information of the target crop in the field, including: For each crop row, determine the coordinates of the first midpoint and the second midpoint of the short side of the minimum bounding rectangle of the crop row outline, as well as the coordinates of at least one center point corresponding to each seedling identification box in the seedling identification box information of the crop row. Based on the coordinates of the first midpoint, the coordinates of each center point, and the coordinates of the second midpoint, at least two line segments are determined; Based on the average length of the line segments, the seedling condition information of the target crop field is determined, including the location and number of missing seedlings.

8. A device for determining crop growth information, characterized in that, include: The data acquisition module is used to acquire UAV images collected for the target crop field, perform conventional segmentation on the UAV images to obtain a conventional segmented image set, and perform misaligned segmentation on the UAV images to obtain a misaligned segmented image set; the conventional segmentation is performed by dividing the image grid starting from the vertices of the UAV images, and the misaligned segmentation is performed by dividing the image grid starting from the non-vertices in the UAV images. The image set determination module is used to determine a first image set based on the conventional segmented image set, and to determine a second image set based on the erroneous segmented image set. The first image set is the data after superimposing the conventional segmented image set with a first crop row mask, and the second image set is the data after superimposing the misaligned segmented image set with a second crop row mask. The seedling recognition module is used to identify the first image set according to the pre-trained crop seedling recognition model to obtain the first seedling recognition result, and to identify the second image set according to the pre-trained crop seedling recognition model to obtain the second seedling recognition result. The seedling condition information determination module is used to determine seedling condition information based on the first seedling identification result and the second seedling identification result.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the seedling information determination method as described in any one of claims 1-7.

10. A storage medium for storing computer-executable instructions, characterized in that, The computer-executable instructions, when executed by a computer processor, are used to perform the crop condition information determination method as described in any one of claims 1-7.

Images