A method for detecting seedling deficiency

Abstract

The present invention discloses a method for detecting missing seedlings, comprising the following steps: obtaining an operating trajectory of a rice transplanter; constructing a rice transplanter transplanting model; calculating theoretical seedling geographic coordinates; collecting images of the rice seedlings after transplanting; identifying and locating the rice seedlings in the rice seedling images using a deep learning target detection model, and obtaining the seedling pixel coordinates; converting the seedling pixel coordinates into actual rice seedling geographic coordinates; combining the actual rice seedling geographic coordinates, using the theoretical rice seedling geographic coordinates as the origin, setting a specified distance as the region radius, and identifying missing seedling regions where no actual rice seedling geographic coordinate points exist within the region radius; and marking the theoretical rice seedling geographic coordinates within the missing seedling region as missing seedling coordinates. This method is applicable to irregular fields and to scenarios with non-linear operating directions, and has a wide range of applications. Furthermore, through image processing technology, combined with the theoretical rice seedling geographic coordinates calculated using a model constructed based on relevant rice transplanter parameters, a comprehensive judgment is made as to whether a rice seedling is missing, thereby improving the accuracy of missing seedling identification.

Description

Technical Field

[0001] The present invention relates to the field of agricultural information technology, and in particular to a seedling shortage detection method. Background Art

[0002] The advent of rice transplanters has significantly improved rice planting efficiency. However, during the transplanting process, transplanters are often prone to problems such as missing seedlings from the seedling tray or malfunctioning of the transplanting mechanism. This can lead to problems such as the seedling removal mechanism being unable to remove seedlings or seedlings being inserted too shallowly into the soil, ultimately resulting in missed or floating seedlings. Missing seedlings directly impact crop yields. Traditional methods for detecting missing seedlings rely primarily on manual inspections, which are inefficient, have high rates of missed detection, and are time-consuming and labor-intensive.

[0003] In order to solve the above problems, the existing technology uses automation to detect missing seedlings. For example, the invention patent application with application publication number CN113807128A discloses a missing seedling marking method, device, computer equipment and storage medium; this method generates a crop binary map based on a farmland image, obtains the main direction of the planting row based on the crop binary map, and extracts multiple crop connected domains included in the crop binary map; obtains at least one planting row area based on the crop binary map and the main direction of the planting row; and marks missing seedlings in the planting row area based on the distance between adjacent crop connected domains in each planting row area. However, the above method still has the following shortcomings:

[0004] 1. Since the above method needs to locate the seedling position based on the crop rows and their connected areas, its application scenario is only when the agricultural machinery is moving in a straight direction, and it is difficult to apply to non-linear working directions.

[0005] 2. The paddy field operation environment is more complex than the dry field operation environment, and side slipping is more likely to occur during the operation process; in the real environment, there are many irregular fields, so the above method has a narrow scope of application.

[0006] 3. Since image processing technology is greatly affected by ambient light, identifying missing seedlings only through visual image processing is inaccurate and unreliable. Summary of the Invention

[0007] The purpose of the present invention is to overcome the above-mentioned problems and provide a method for detecting missing seedlings. The method can be applied to irregular fields and scenes with non-linear working directions, and has a wide range of applications. Moreover, the method uses image processing technology, combined with the theoretical seedling geographical coordinates calculated based on the model constructed based on the relevant parameters of the rice transplanter, to make a comprehensive judgment on whether there are missing seedlings, thereby improving the accuracy of missing seedling identification.

[0008] The purpose of the present invention is achieved through the following technical solutions:

[0009] A method for detecting seedling shortage is provided, the method being used to detect seedling shortage after transplanting by a rice transplanter, the method comprising the following steps:

[0010] (1) Obtaining the operation trajectory of the rice transplanter;

[0011] (2) Constructing a rice transplanter model;

[0012] (3) Calculate the theoretical seedling geographical coordinates;

[0013] (4) Collect images of rice seedlings after transplanting using drones;

[0014] (5) Using the constructed deep learning target detection model, the rice seedlings in the rice seedling image are identified and located, and the pixel coordinates of the rice seedlings are obtained;

[0015] (6) converting the seedling pixel coordinates into the actual seedling geographic coordinates;

[0016] (7) combining the actual seedling geographical coordinates, taking the theoretical seedling geographical coordinates as the origin, setting the specified distance as the area radius; if the actual seedling geographical coordinate point does not exist within the area radius, then the area radius is judged to be a seedling-deficient area;

[0017] (8) Mark the theoretical seedling geographical coordinates in the seedling-missing area as the seedling-missing coordinates.

[0018] A preferred embodiment of the present invention, wherein, in step (1), an RTK antenna is installed on the top of the rice transplanter and an RTK receiver is installed inside the rice transplanter to record the operation trajectory of the rice transplanter during operation, wherein the operation trajectory includes the operation position and operation direction of the rice transplanter.

[0019] Preferably, in step (2), constructing the rice transplanter model requires obtaining the following parameters of the rice transplanter:

[0020] The spatial distance between the actual RTK position point and the transplanting points of each transplanting mechanism in the transplanter, the real-time travel speed of the transplanter, the real-time position relationship between the transplanting mechanism and the transplanter body, and the position signal of the transplanting mechanism.

[0021] In step (3), the theoretical seedling geographic coordinates can be calculated through the operation trajectory of the rice transplanter and the rice transplanter transplanting model.

[0022] Preferably, in step (4), the drone is an aerial survey drone with RTK positioning function, which can record the position information and attitude information of the camera at the moment of shooting.

[0023] Preferably, in step (5), the method for constructing the deep learning target detection model includes the following steps:

[0024] (5.1) Collect images of rice seedlings after transplanting;

[0025] (5.2) Image preprocessing, labeling seedling images;

[0026] (5.3) The rice seedling images are divided into training and validation sets, and a deep learning target detection model is obtained through training.

[0027] Preferably, in step (6), the specific steps of converting the seedling pixel coordinates into the actual seedling geographic coordinates include:

[0028] (6.1) Mapping the rice seedling image in step (4) using aerial survey modeling software to obtain an image with geographic coordinate information;

[0029] (6.2) Establish a correspondence between the seedling pixel coordinates in step (5) and the image with geographic coordinate information, and convert the seedling pixel coordinates into the actual seedling geographic coordinates.

[0030] Preferably, in step (7), if there is an actual seedling geographical coordinate point within the area radius, the area radius is judged to be a seedling area.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] The seedling missing detection method of the present invention is applied to seedling missing detection after transplanting. By collecting seedling images after transplanting and processing them, the influence of light on the recognition rate is reduced. The theoretical seedling geographical coordinates are obtained by calculating the operation trajectory of the transplanter and the transplanting model of the transplanter. Combined with the actual seedling geographical coordinates, the theoretical seedling geographical coordinates are used as the origin to comprehensively judge whether seedlings are missing, thereby greatly improving the accuracy of seedling missing recognition.

[0033] 2. The missing seedling detection method of the present invention compares the theoretical seedling geographical coordinates calculated by the rice transplanter transplanting model with the actual seedling geographical coordinates identified and calculated by the deep learning target detection model to form a comprehensive judgment on the missing seedling location, rather than relying solely on image processing to make a judgment; if it relies solely on image processing to make a judgment, the image processing is greatly affected by the ambient light; therefore, the missing seedling detection method of the present invention has a higher missing seedling detection rate and is more reliable.

[0034] 3. The seedling missing detection method of the present invention compares the theoretical seedling geographical coordinates calculated by the rice transplanter transplanting model with the actual seedling geographical coordinates identified and calculated by the deep learning target detection model to form a comprehensive judgment on the location of the missing seedlings. Therefore, it can identify the seedling missing situation in the normal operating area of ​​the rice transplanter. It is suitable for various irregular fields and also for scenes with non-linear operating directions. It has a wide range of applications. After transplanting, various fields can be detected for seedling missing. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The present invention is a flowchart of a method for detecting seedling deficiency.

[0036] Figure 2 Rice seedling images collected by drone.

[0037] Figure 3 This is a schematic diagram of the theoretical seedling geographical coordinates obtained by calculating the operation trajectory and the rice transplanter transplanting model in the present invention.

[0038] Figure 4 This is a schematic diagram of the actual geographical coordinates of the rice seedlings detected by the deep learning target detection model in the present invention.

[0039] Figure 5 It is a combined diagram of the theoretical rice seedling geographical coordinate points and the actual rice seedling geographical coordinates in the present invention.

[0040] Figure 6 This is a schematic diagram of the seedling-missing area in the present invention, where the background of the picture is an image of the seedlings taken by a drone.

[0041] Figure 7 This is a schematic diagram of the coordinate points of missing seedlings in the present invention, where the background of the picture is an image of seedlings taken by a drone. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described below in conjunction with embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0043] See also Figure 1-Figure 7 A method for detecting seedling shortage is provided, wherein the method is used to detect the shortage of seedlings after transplanting by a rice transplanter, and the method comprises the following steps:

[0044] (1) Obtaining the operation trajectory of the rice transplanter;

[0045] (2) Constructing a rice transplanter model;

[0046] (3) Calculate the theoretical seedling geographical coordinates;

[0047] (4) Collect images of rice seedlings after transplanting using drones;

[0048] (5) Using the constructed deep learning target detection model, the rice seedlings in the rice seedling image are identified and located, and the pixel coordinates of the rice seedlings are obtained;

[0049] (6) converting the seedling pixel coordinates into the actual seedling geographic coordinates;

[0050] (7) combining the actual seedling geographical coordinates, taking the theoretical seedling geographical coordinates as the origin, setting the specified distance as the area radius; if the actual seedling geographical coordinate point does not exist within the area radius, then the area radius is judged to be a seedling-deficient area;

[0051] (8) Mark the theoretical seedling geographical coordinates in the seedling-missing area as the seedling-missing coordinates.

[0052] In this embodiment, by comparing the theoretical seedling geographical coordinates calculated by the rice transplanter transplanting model with the actual seedling geographical coordinates identified and calculated by the deep learning target detection model, a comprehensive judgment of the missing seedling location is formed, thereby further improving the accuracy of missing seedling detection.

[0053] Specifically, in step (1), an RTK antenna is installed on the top of the rice transplanter and an RTK receiver is installed inside the rice transplanter to record the rice transplanter's operating trajectory during operation, wherein the operating trajectory includes the rice transplanter's operating position and operating direction. RTK positioning can accurately identify the rice transplanter's operating trajectory, making the theoretical rice seedling geographic coordinates more accurate.

[0054] Specifically, in step (2), constructing the rice transplanter model requires obtaining the following parameters of the rice transplanter:

[0055] The spatial distance between the RTK actual position point and the planting points of each transplanting mechanism in the rice transplanter, the real-time travel speed of the rice transplanter, the real-time position relationship between the transplanting mechanism and the transplanter body, and the position signal of the transplanting mechanism. In the above steps, the RTK actual position point refers to the position of the RTK antenna. The rice transplanter consists of the transplanter body and multiple transplanting mechanisms used for transplanting rice seedlings.

[0056] See also Figure 1-3 In step (3), the theoretical seedling geographic coordinates can be calculated using the rice transplanter's operating trajectory and the rice transplanter model. The rice transplanter model is actually a calculation formula based on the parameter values ​​of the above parameters. By inputting the parameter values ​​and performing program calculations, the theoretical seedling geographic coordinates can be obtained.

[0057] Specifically, in step (4), the drone is an aerial survey drone with an RTK positioning function, which can record the position information and attitude information of the camera at the moment of shooting. By obtaining the position information and attitude information, the pixel coordinates of the seedlings are converted into the actual geographical coordinates of the seedlings; specifically, the position information and attitude information of the camera at the moment of shooting are recorded by the aerial survey drone, so that the aerial survey modeling software can obtain images with geographical coordinate information during the mapping process, thereby obtaining the actual geographical coordinates of the seedlings. Therefore, the camera position information and attitude information at the moment of shooting are necessary conditions. Without this information, the map cannot be built.

[0058] Specifically, in step (5), the method for constructing the deep learning target detection model includes the following steps:

[0059] (5.1) Collect images of rice seedlings after transplanting;

[0060] (5.2) Image preprocessing, labeling seedling images;

[0061] (5.3) The rice seedling images are divided into training and validation sets, and a deep learning target detection model is obtained through training.

[0062] In the above steps, the seedling images are collected in advance, annotated, divided into training sets and validation sets, and then the deep learning target detection model is trained. The obtained deep learning target detection model can be used to detect other acquired seedling images to obtain the seedling pixel coordinates.

[0063] See also Figure 2-Figure 4 In step (6), the specific steps of converting the seedling pixel coordinates into the actual seedling geographic coordinates include:

[0064] (6.1) Mapping the rice seedling image in step (4) using commercially available aerial survey modeling software to obtain an image with geographic coordinate information;

[0065] (6.2) Establish a correspondence between the seedling pixel coordinates in step (5) and the image with geographic coordinate information, and convert the seedling pixel coordinates into the actual seedling geographic coordinates.

[0066] In step (7), the area radius is a circular area, and the origin is the center of the circular area.

[0067] Furthermore, in step (7), if there is an actual seedling geographical coordinate point within the area radius, the area radius is determined to be a seedling area.

[0068] Figure 3 In the figure, the 50% gray point is the theoretical seedling geographical coordinate point, and the background is the seedling image taken by the drone; Figure 4 In the figure, the white dots are the actual geographical coordinates of the rice seedlings, and the background is the rice seedling image taken by the drone; Figure 5 In the figure, the white dots are the actual seedling geographical coordinate points, the 50% gray dots are the theoretical seedling geographical coordinate points, the black frame is the seedling-missing area, the black dots are the seedling-missing coordinate points, and the background is the seedling image taken by the drone.

[0069] The above is a preferred embodiment of the present invention, but the embodiment of the present invention is not limited to the above content. Any other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A method for detecting seedling shortage, characterized in that: The method is used to detect the lack of seedlings after transplanting by a rice transplanter, and the method comprises the following steps: (1) Obtain the operation trajectory of the rice transplanter; (2) Constructing a rice transplanting model; (3) Calculate the theoretical seedling geographical coordinates; (4) Collect images of rice seedlings after transplanting using drones; (5) Identify and locate the rice seedlings in the rice seedling image through the constructed deep learning target detection model, and obtain the pixel coordinates of the rice seedlings; (6) Convert the seedling pixel coordinates into the actual seedling geographic coordinates; (7) Combined with the actual seedling geographical coordinates, with the theoretical seedling geographical coordinates as the origin, set the specified distance as the area radius. If there is no actual seedling geographical coordinate point within the area radius, the area radius is judged to be a seedling-missing area; (8) Mark the theoretical seedling geographical coordinates in the seedling-missing area as the seedling-missing coordinates; In step (1), an RTK antenna is installed on the top of the rice transplanter and an RTK receiver is installed inside the rice transplanter to record the operation trajectory of the rice transplanter during operation, wherein the operation trajectory includes the operation position and operation direction of the rice transplanter; In step (6), the specific steps of converting the seedling pixel coordinates into the actual seedling geographic coordinates include: (6.1) Use aerial survey modeling software to map the rice seedling image in step (4) to obtain an image with geographic coordinate information; (6.2) Establish a correspondence between the seedling pixel coordinates in step (5) and the image with geographic coordinate information, and convert the seedling pixel coordinates into the actual seedling geographic coordinates.

2. A method for detecting seedling deficiency according to claim 1, characterized in that: In step (2), to construct the rice transplanter model, the following parameters of the rice transplanter need to be obtained: The spatial distance between the actual RTK position point and the transplanting points of each transplanting mechanism in the transplanter, the real-time travel speed of the transplanter, the real-time position relationship between the transplanting mechanism and the transplanter body, and the position signal of the transplanting mechanism.

3. A method for detecting seedling deficiency according to claim 1, characterized in that: In step (4), the drone is an aerial survey drone with RTK positioning function, which can record the position information and attitude information of the camera at the moment of shooting.

4. A method for detecting seedling deficiency according to claim 1, characterized in that: In step (5), the method for constructing the deep learning target detection model includes the following steps: (5.1) Collect images of rice seedlings after transplanting; (5.2) Image preprocessing and annotation of rice seedling images; (5.3) Divide the rice seedling images into a training set and a validation set, and train a deep learning target detection model.

5. A method for detecting seedling deficiency according to claim 1, characterized in that: In step (7), if there is an actual seedling geographical coordinate point within the area radius, the area radius is judged to be a seedling area.

Images