Seed sowing status monitoring method and system based on image recognition

Abstract

The present invention relates to the field of image recognition sowing and cultivation technology, and specifically to a seed sowing status monitoring method and system based on image recognition. First, soil from an area to be planted in batches is collected and filled into a transparent planting box. Then, plump and healthy target model seeds are selected and planted in the transparent planting box. Then, a field planting environment is simulated and the seeds are provided with the same planting parameters for cultivation. Then, a root system development map after seed sowing is collected at a set collection interval by an in-situ root system image collector. Then, image preprocessing is performed on the root system development map and characteristic values ​​are extracted. Finally, post-sowing seed characteristic data obtained by image recognition is compared with target model seed benchmark data, and a planting decision for the target model seed in the planting area is output. In this way, whether the seeds are suitable for the planting area is evaluated from the perspective of root system development, and the original evaluation basis, which was based on the planting results, is changed to an early prediction of potential risks in the rooting stage after seed planting.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition sowing and cultivation, and in particular to a seed sowing state monitoring method and system based on image recognition. Background Art

[0002] In agricultural production, seed breeding with high-quality traits is a key focus. Seed companies or research institutions will select and breed unique seed models based on technologies such as gene sequencing and stable trait inheritance. These seed models are either designed to enhance stress resistance for specific soils or specific planting areas, such as being more adaptable to saline-alkali land or resistant to lodging, or to enhance their taste or yield or increase the accumulation of flavor substances.

[0003] After these seeds are developed, seed companies or research institutions will provide a series of parameters and applicable planting scenarios for their planting. Farmers or agricultural extension stations will then select potentially applicable seeds for trial planting or small-batch planting based on the soil and environmental characteristics of the area. The traditional planting method is to divide the planting area into a grid for trial planting, and then conduct seed evaluation based on the trial planting situation. The overall trial planting cycle is long and is affected by the results.

[0004] On the one hand, with the development of image acquisition and image recognition technology, combined with image recognition technology, regular image collection is used to analyze the short-term root development of seeds after planting, and compared with the standard data provided by seed companies or research institutions, so as to more accurately judge the adaptability. This method can reduce costs, improve efficiency, real-time monitoring, and multi-dimensional evaluation.

[0005] On the other hand, after the seeds are planted, except for unavoidable sudden natural disasters, after they can successfully take root, human intervention can be direct, and the environmental factors can be changed by configuring nutrient solution to promote their growth. Therefore, the seed mortality rate during the planting period often occurs during the seed planting and rooting stage. Therefore, whether it can take root well and whether the rooting effect can meet expectations are important indicators for evaluating whether a type of seed can be planted in this planting scenario and the basis for making decisions. Summary of the Invention

[0006] The purpose of the present invention is to provide a seed sowing status monitoring method and system based on image recognition to address the above-mentioned problems.

[0007] The technical solution adopted by the present invention is as follows: a seed sowing status monitoring method based on image recognition comprises the following steps:

[0008] S1. Soil collection: Collect soil from the area to be planted in batches, group it, and fill it into transparent planting boxes;

[0009] S2. Seed sowing: Select plump and healthy target seeds and plant them in transparent planting boxes according to the planting interval;

[0010] S3. Seed cultivation, imitating the field planting environment, providing the same planting parameters for seeds to be cultivated;

[0011] S4. Image acquisition of seed sowing status: root development images after seed sowing are collected by an in-situ root image acquisition instrument at set collection intervals;

[0012] S5. Image recognition of seed sowing status, image preprocessing of root system development diagram and extraction of feature values ​​to obtain feature data after seed sowing;

[0013] S6. Feature data comparison: compare the feature data of the seeds after sowing obtained by image recognition with the benchmark data of the target model seeds, and output the planting decision of the target model seeds in the planting area.

[0014] Preferably, in S1, grid points are arranged in the area to be planted in batches. After the plowing operation is completed, the loose soil of 0-5 cm on the soil surface is removed from each grid point to obtain 5-20 cm tillage layer soil, 20-40 cm plow bottom layer soil and 40-60 cm parent material layer soil, and the ratio of the three layers of soil is 1:1:1. A total of 10 L of each of the three layers of soil are collected in the area to be planted in batches. The soil of the same layer at each grid point is mixed evenly and added to the transparent planting box in the order of parent material layer soil, plow bottom layer soil and tillage layer soil.

[0015] Preferably, in S1, the transparent planting box is made of PMMA material, a plurality of planting cavities are separated in the transparent planting box, and a water hole is provided at the bottom of the planting cavity, an observation room is provided on one side of the transparent planting box, and an in-situ root image acquisition device is provided in the observation room, and the in-situ root image acquisition device can capture images of the roots of seeds planted in the planting cavity.

[0016] Preferably, in S1, a side of the transparent planting box close to the observation chamber is provided with an inner convex cavity, and the inner convex cavity protrudes toward a side of the planting cavity, and the image acquisition terminal of the in-situ root system image acquisition instrument can extend into the inner convex cavity.

[0017] Preferably, in S3, the planting parameters include planting time, planting light intensity, planting water amount and planting fertilizer amount, and the same planting light intensity and planting water amount are used in daily cultivation.

[0018] Preferably, in S4, timing is started from the time after the seeds are sown, and images of the root system after the seeds are sown are collected at a frequency of 6 hours, including real-life images of the root system and fluorescence images of the root system bands. In a night environment, image collection is performed by supplementing light with an external light source.

[0019] Preferably, in said S5, the characteristic data after seed sowing include main root morphological parameters, xylem health and root water use efficiency, including the following sub-steps:

[0020] The root system development diagrams obtained by S4 after seed sowing were grouped into five-day cycles, namely the first-week root growth group, the second-week root growth group, and the Nth-week root growth group.

[0021] Select the root system images from the root growth group image data, remove the root xylem vessel noise to retain the edge details, extract the main root outline skeleton, calculate the horizontal angle between the first-level bifurcation point and the vertical direction to complete the main root morphological parameter evaluation;

[0022] Select the root fluorescence image from the root growth group image data, extract the xylem fluorescence intensity, calculate the vessel density and average diameter, and complete the xylem health assessment;

[0023] The root system real-life images from the root growth group image data were selected to create the root system LiDAR point cloud data. At the same time, soil moisture data from the same time period was collected. The root system water utilization efficiency was evaluated based on the root system LiDAR volume change rate and soil moisture content data.

[0024] Preferably, in said S6, the target model seed benchmark data includes taproot morphological parameter benchmark data, xylem health benchmark data and root water utilization efficiency benchmark data provided by the seed company;

[0025] Compare the evaluation data obtained by S5 with the target model seed benchmark data, assign weights to the main root morphological parameters, xylem health and root water use efficiency, and output the decision S based on the weights;

[0026] When the decision S is ≥ 0.75, the soil in the area to be planted in batches is suitable for planting target model seeds;

[0027] When 0.5≤decision S<0.75, adjust the planting parameters when planting target model seeds in the area to be planted in batches;

[0028] When the decision S<0.5, the area to be planted in batches is not suitable for planting the target model seeds.

[0029] The present application provides a seed sowing status monitoring system based on image recognition, including a processor, a memory, and a program or instruction stored in the memory for the processor to run, wherein the program or instruction is used to implement the above-mentioned image recognition steps.

[0030] The beneficial effects of the present invention are:

[0031] 1. A technical solution based on image acquisition and image recognition after seed planting is provided. The adaptability of seeds to the planting area is evaluated from the perspective of root development. Compared with existing technologies, the timeliness of the evaluation is improved. The original evaluation basis, which was changed from planting results to early prediction of potential risks during the rooting stage after seed planting, is changed to image acquisition and image recognition after seed planting.

[0032] 2. Combining traditional agricultural planting with digital assessment not only retains the verification method of small-scale trial planting and careful exploration, but also overcomes the problems of traditional agricultural planting such as high cost, long cycle, and incomplete data. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a schematic diagram of the transparent planting box structure;

[0034] Figure 2 The flowchart of the seed sowing status monitoring method based on image recognition is as follows;

[0035] Figure 3 It is the benchmark data diagram of main root morphological parameters;

[0036] Figure 4 Measured data of main root morphological parameters.

[0037] The accompanying drawings are marked as follows:

[0038] 1 is a transparent planting box, 2 is an observation room, 3 is a planting cavity, 4 is an in-situ root system image acquisition device, 5 is a data line, 6 is an inner convex cavity, and 7 is a planting point. DETAILED DESCRIPTION

[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0040] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are also within the scope of protection of the present invention.

[0041] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other.

[0042] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not require further definition or explanation in subsequent drawings.

[0043] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, or are the orientations or positional relationships in which the inventive product is typically placed when in use, or are the orientations or positional relationships commonly understood by those skilled in the art. These terms are intended only to facilitate the description of the present invention and to simplify the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0044] In the description of the present invention, it should also be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0045] Example 1

[0046] like Figure 2 As shown, the seed sowing status monitoring method based on image recognition includes the following steps:

[0047] S1. Soil collection: collect soil from the area to be planted in batches, group it, and fill it into a transparent planting box 1;

[0048] S2. Seed sowing: Select plump and healthy target seeds and plant them in transparent planting box 1 according to the planting interval;

[0049] S3. Seed cultivation, imitating the field planting environment, providing the same planting parameters for seeds to be cultivated;

[0050] S4 seed sowing state image acquisition, through the in situ root image acquisition instrument 4 to set the acquisition interval to collect seed root development map after sowing;

[0051] S5. Image recognition of seed sowing status, image preprocessing of root system development diagram and extraction of feature values ​​to obtain feature data after seed sowing;

[0052] S6. Feature data comparison: compare the feature data of the seeds after sowing obtained by image recognition with the benchmark data of the target model seeds, and output the planting decision of the target model seeds in the planting area.

[0053] This design aims to provide a technical solution based on post-planting image acquisition and image recognition, assessing the suitability of seeds for a particular planting area from the perspective of root development. This improves the timeliness of assessment compared to existing technologies, shifting from the original evaluation basis based on planting results to early prediction of potential risks during the rooting phase after seed planting. Combining traditional agricultural planting with digital assessment preserves the careful verification method of small-scale trial planting while overcoming the high costs, long cycles, and incomplete data inherent in traditional agricultural planting.

[0054] Example 2

[0055] Taking the complete implementation of this solution as an example, the seed sowing status monitoring method based on image recognition includes the following steps:

[0056] S1. Soil collection: Soil from the area to be planted in batches is collected, grouped, and filled into a transparent planting box 1. A grid is formed in the area to be planted in batches. After the plowing operation is completed, the topsoil of 0-5 cm on the soil surface is removed from each grid point. A 5-20 cm tillage layer of soil, a 20-40 cm plow bottom layer of soil, and a 40-60 cm parent material layer of soil are obtained. The ratio of the three layers of soil is 1:1:1. A total of 10 L of soil from each of the three layers of soil are collected from the area to be planted in batches. The soil from the same layer at each grid point is mixed evenly and added to the transparent planting box 1 in the order of parent material layer soil, plow bottom layer soil, and tillage layer soil.

[0057] S2. Seed sowing: Select plump and healthy target seeds and plant them in the transparent planting box 1 according to the planting interval. First, fill the parent layer soil into the planting cavity 3, then fill the plow bottom soil into the planting cavity 3, and finally gradually add the tillage layer soil and bury the target seeds therein. The target seeds are close to the inner convex cavity 6 when sowing;

[0058] S3. Seed cultivation, imitating the field planting environment, provides the same planting parameters for the seeds to be cultivated, where the planting parameters include planting time, planting light intensity, planting water and planting fertilizer amount, using the same planting light intensity and planting water in daily cultivation;

[0059] S4. Image acquisition of seed sowing status: Root system development images after seed sowing are collected using an in-situ root system image acquisition device 4 at set acquisition intervals. Images of the root system after seed sowing are collected at a 6-hour interval starting from the time the seeds are sown. These images include both real-life root system images and root system fluorescence images. At night, image acquisition is performed using an external light source.

[0060] S5. Image recognition of seed sowing status: performing image preprocessing on the root development map and extracting eigenvalues ​​to obtain post-sowing seed feature data, including taproot morphological parameters, xylem health, and root water use efficiency. This includes the following sub-steps:

[0061] The root system development diagrams obtained by S4 after seed sowing were grouped into five-day cycles, namely the first-week root growth group, the second-week root growth group, and the Nth-week root growth group.

[0062] Select the root system images from the root growth group image data, remove the root xylem vessel noise to retain the edge details, extract the main root outline skeleton, calculate the horizontal angle between the first-level bifurcation point and the vertical direction to complete the main root morphological parameter evaluation;

[0063] Select the root fluorescence image from the root growth group image data, extract the xylem fluorescence intensity, calculate the vessel density and average diameter, and complete the xylem health assessment;

[0064] Select the root system real-life image from the root growth group image data to create the root system LiDAR point cloud data. Simultaneously, collect soil moisture data for the same time period. Based on the root system LiDAR volume change rate and soil moisture data, complete the root system water use efficiency assessment.

[0065] S6. Feature data comparison: comparing the post-sowing seed feature data obtained through image recognition with the target seed model's baseline data, and outputting a planting decision for the target seed model in the planting area. The target seed model's baseline data includes the taproot morphology parameter baseline data, xylem health baseline data, and root water use efficiency baseline data provided by the seed company.

[0066] Compare the evaluation data obtained by S5 with the target model seed benchmark data, assign weights to the main root morphological parameters, xylem health and root water use efficiency, and output the decision S based on the weights;

[0067] When the decision S is ≥ 0.75, the soil in the area to be planted in batches is suitable for planting target model seeds;

[0068] When 0.5≤decision S<0.75, adjust the planting parameters when planting target model seeds in the area to be planted in batches;

[0069] When the decision S<0.5, the area to be planted in batches is not suitable for planting the target model seeds.

[0070] This design aims to provide a technical solution based on post-planting image acquisition and image recognition, assessing the suitability of seeds for a particular planting area from the perspective of root development. This improves the timeliness of assessment compared to existing technologies, shifting from the original evaluation basis based on planting results to early prediction of potential risks during the rooting phase after seed planting. Combining traditional agricultural planting with digital assessment preserves the careful verification method of small-scale trial planting while overcoming the high costs, long cycles, and incomplete data inherent in traditional agricultural planting.

[0071] For the structure of the transparent planting box 1, please refer to Figure 1 , wherein the transparent planting box 1 is made of PMMA material, and a plurality of planting cavities 3 are separated in the transparent planting box 1, and a water hole is provided at the bottom of the planting cavity 3. An observation chamber 2 is provided on one side of the transparent planting box 1, and an in-situ root image collector 4 is provided in the observation chamber 2, and the in-situ root image collector 4 can collect images of the roots of seeds planted in the planting cavity 3. An inner convex cavity 6 is provided on one side of the transparent planting box 1 close to the observation chamber 2, and the inner convex cavity 6 protrudes toward one side of the planting cavity 3. The image acquisition terminal of the in-situ root image collector 4 can extend into the inner convex cavity 6. When planting, the target model seeds are planted at the planting point 7, so that the distance from the inner convex cavity 6 is 2-5 cm. As the seed roots develop, they will gradually approach the inner concave cavity, so that the image acquisition terminal of the in-situ root image collector 4 can more easily obtain the seed root map.

[0072] In the specific execution of step S3, multiple coordinations are required. Taking the target model seeds as an example, the planting time is early March. The lighting conditions of the outdoor planting area, that is, the area to be planted in batches, are collected, and then the indoor light compensation is adjusted according to the real-time collected results. At the same time, the precipitation of the outdoor planting area, that is, the area to be planted in batches, is collected, and then the soil moisture is adjusted according to the real-time collected results. As for the amount of fertilizer applied for planting, on the one hand, it can be based on the recommended usage of the target model seeds provided by the seed company or research institution. On the other hand, the on-site staff will make adjustments based on the recommended usage based on the subsequent planting conditions, and record the daily addition of planting.

[0073] Regarding step S5, this embodiment provides an achievable form. In the specific implementation, the anisotropic diffusion algorithm is first used to eliminate the xylem vessel noise and retain the edge details. Then, the soil and stone impurities in the image are removed by Fourier transform, and the root system map is retained. Then, the U-Net network is used to perform image segmentation on the root system map.

[0074] Specifically, the root system map obtained above is manually annotated, and then the annotated root system map is converted into a data set, which includes a training set, a validation set, and a test set. Then, an improved U-Net network is constructed, and a cascaded dilated convolution module is provided in the improved U-Net network.

[0075] The improved U-Net network is trained and verified through the training set and the validation set to obtain the final U-Net network; the test set is preprocessed, and the preprocessed test is input into the final U-Net network, and the corresponding root segmentation image is output through the final U-Net network. It should be pointed out that the image recognition and segmentation of the U-Net network here can be understood as an existing technology. When implementing it specifically, those skilled in the art can refer to the corresponding existing technology and make adjustments according to actual needs, such as Figure 3 and Figure 4 As shown, from a morphological point of view, the benchmark data of the main root morphological parameters of the target model seeds can be compared with the measured data of the main root morphological parameters. It can be seen that the deviation between the lateral roots and the main root is within 15°, and there will be no scattered roots after planting.

[0076] Then, after completing the image segmentation and obtaining the xylem vessels, the main root outline skeleton is extracted, the horizontal angle between the first-level bifurcation point and the vertical direction is calculated, and the main root morphological parameter evaluation is completed.

[0077] In terms of water use efficiency, it is necessary to identify the root volume and collect additional soil moisture content in the image data collected previously. By identifying the changes in root volume in the image and matching them with the changes in soil moisture content in the corresponding time period, the two can be compared to obtain the transpiration rate, thereby evaluating the water use efficiency. This is used to determine whether the target model seeds can take root successfully and whether the roots can subsequently grow healthily.

[0078] In the assessment of xylem health, root fluorescence images are needed. The principle is that xylem vessels have characteristic autofluorescence in the 450nm band. The specific implementation method is to first use wavelet transform to remove random noise in the hyperspectral image, then use a standard white board for radiation calibration to eliminate the influence of light source fluctuations, and then extract the 450nm single-band image. Then, the vessel area is separated by the Otsu threshold method, and the number of vessel branches per unit length is counted.

[0079] In step S6, the weighted output decision S can be expressed as:

[0080] S=W1·S1+W2·S2+W3·S3;

[0081] Among them, S1 is the main root morphological parameter evaluation score, which is obtained based on the difference between the professional and technical personnel from the morphological perspective and the standard data graph;

[0082] W1 is the main root morphology parameter evaluation weight, which is based on the main root morphology and planting requirements when professional technicians plant from target model seeds;

[0083] S2 is the xylem health assessment score, which is obtained by scoring the xylem health data obtained by professional technicians from image recognition;

[0084] W2 is the xylem health assessment weight, which is based on the xylem health and planting requirements when professional technicians plant from target model seeds;

[0085] S3 is the root water use efficiency evaluation score, which is obtained based on the scoring of the root water use efficiency obtained by professional technicians;

[0086] W3 is the root water use efficiency assessment weight, which is based on the root water use efficiency and planting requirements when professional technicians plant from target model seeds;

[0087] And W1 +W2 +W3=1.

[0088] When the decision S is ≥ 0.75, the soil in the area to be planted in batches is suitable for planting target model seeds;

[0089] When 0.5≤decision S<0.75, adjust the planting parameters when planting target model seeds in the area to be planted in batches;

[0090] When the decision S<0.5, the area to be planted in batches is not suitable for planting the target model seeds.

[0091] It should also be pointed out that in order to facilitate the description of the entire plan, the term target model seeds is adopted. However, it can be understood that since seeds can be divided into seed planting and seed bud planting at a certain stage of cultivation when planted, the target model seeds here are not limited to seeds, but also include target model seed buds, and the implementation methods and principles are consistent.

[0092] Example 3

[0093] In this embodiment, a seed sowing status monitoring method based on image recognition provides a seed sowing status monitoring system based on image recognition, which includes a processor, a memory, and a program or instruction stored in the memory for the processor to run, and the program or instruction is used to implement the aforementioned image recognition steps.

[0094] It should be noted that, in this embodiment, the program or instruction referred to may be implemented in the form of a software functional unit and may be stored in a computer-readable storage medium when sold or used as an independent product.

[0095] Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0096] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A seed sowing status monitoring method based on image recognition, characterized in that: The following steps are involved: S1. Soil collection: collect soil from the area to be planted in batches, group it, and fill it into a transparent planting box (1); S2. Seed sowing: Select plump and healthy target type seeds and plant them in a transparent planting box (1) according to the planting interval; S3. Seed cultivation, imitating the field planting environment, providing the same planting parameters for seeds to be cultivated; S4. Image acquisition of seed sowing status: using an in-situ root system image acquisition device (4) to acquire root system development images after seed sowing at a set acquisition interval. The timing starts from the time of seed sowing, and the root system images after seed sowing are acquired at a frequency of 6 hours, including root system real-life images and root system band fluorescence images. In a night environment, image acquisition is performed using an external light source for supplementary lighting. S5. Image recognition of seed sowing status: performing image preprocessing on the root development map and extracting eigenvalues ​​to obtain post-sowing seed feature data, including taproot morphological parameters, xylem health, and root water use efficiency. This includes the following sub-steps: The root system development diagrams obtained by S4 after seed sowing were grouped into five-day cycles, namely the first-week root growth group, the second-week root growth group, and the Nth-week root growth group. Select the root system images from the root growth group image data, remove the root xylem vessel noise to retain the edge details, extract the main root outline skeleton, calculate the angle between the first-level bifurcation point and the vertical direction to complete the main root morphological parameter evaluation; Select the root fluorescence image from the root growth group image data, extract the xylem fluorescence intensity, calculate the vessel density and average diameter, and complete the xylem health assessment; Select the root system real-life image from the root growth group image data to create the root system LiDAR point cloud data. Simultaneously, collect soil moisture data for the same time period. Based on the root system LiDAR volume change rate and soil moisture data, complete the root system water use efficiency assessment. S6. Feature data comparison: comparing the post-sowing seed feature data obtained through image recognition with the target seed model's baseline data, and outputting a planting decision for the target seed model in the planting area. The target seed model's baseline data includes the taproot morphology parameter baseline data, xylem health baseline data, and root water use efficiency baseline data provided by the seed company. Compare the evaluation data obtained by S5 with the target model seed benchmark data, assign weights to the main root morphological parameters, xylem health and root water use efficiency, and output the decision S based on the weights; When the decision S is ≥ 0.75, the soil in the area to be planted in batches is suitable for planting target model seeds; When 0.5≤decision S<0.75, adjust the planting parameters when planting target model seeds in the area to be planted in batches; When the decision S<0.5, the area to be planted in batches is not suitable for planting the target model seeds.

2. The seed sowing status monitoring method based on image recognition according to claim 1, characterized in that: In said S1, a grid is arranged in the area to be planted in batches. After the plowing operation is completed, the loose soil of 0-5 cm on the soil surface is removed from each grid point to obtain 5-20 cm tillage layer soil, 20-40 cm plow bottom layer soil and 40-60 cm parent material layer soil, and the ratio of the three layers of soil is 1:1:

1. A total of 10 L of each of the three layers of soil are collected in the area to be planted in batches. The soil of the same layer is mixed evenly at each grid point and added to the transparent planting box (1) in the order of parent material layer soil, plow bottom layer soil and tillage layer soil.

3. The seed sowing status monitoring method based on image recognition according to claim 1, characterized in that: In the S1, the transparent planting box (1) is made of PMMA material, a plurality of planting cavities (3) are separated in the transparent planting box (1), and a water-permeable hole is provided at the bottom of the planting cavity (3), an observation chamber (2) is provided on one side of the transparent planting box (1), and an in-situ root system image acquisition device (4) is provided in the observation chamber (2), and the in-situ root system image acquisition device (4) can acquire images of the root systems of the seeds planted in the planting cavities (3).

4. The seed sowing status monitoring method based on image recognition according to claim 3 is characterized in that: In S1, the transparent planting box (1) is provided with an inner convex cavity (6) on a side close to the observation chamber (2), and the inner convex cavity (6) protrudes toward a side of the planting cavity (3), and the image acquisition terminal of the in-situ root system image acquisition instrument (4) can extend into the inner convex cavity (6).

5. The seed sowing status monitoring method based on image recognition according to claim 4 is characterized in that: In the S2, the parent material layer soil is first filled into the planting cavity (3), and then the plow bottom layer soil is filled into the planting cavity (3). Finally, the tillage layer soil is gradually added and the target type seed is buried therein. The target type seed is close to the inner convex cavity (6) when sowing.

6. The seed sowing status monitoring method based on image recognition according to claim 1, characterized in that: In said S3, the planting parameters include planting time, planting light intensity, planting water amount and planting fertilizer amount, and the same planting light intensity and planting water amount are used in daily cultivation.

7. A seed sowing status monitoring system based on image recognition, characterized in that: The system comprises a processor, a memory and a program or instruction stored in the memory for execution by the processor, wherein the program or instruction is used to implement the steps of image recognition described in any one of claims 1 to 6.

Images