Computer vision-based smart agricultural plant panorama measurement and calibration method and system

Abstract

This invention relates to the field of smart agriculture technology, and in particular to a method and system for measuring and calibrating the overall appearance of plants in smart agriculture based on computer vision. The method includes: constructing a regional simulation model of the planting area; measuring the longitudinal distance between any two adjacent target plants in the simulation model according to the start time of the target plant's development; generating a normal growth signal based on the longitudinal distance; a region division module setting independent growth regions based on the longitudinal distance value; an image analysis module further identifying the relationship between branches and leaves and growth points in a unit image, distinguishing between related and unrelated branches and leaves, accurately determining the interference state; and then marking and measuring related branches and leaves according to the interference state to obtain accurate plant growth information. This allows the measurement results to more realistically reflect the actual growth status of a single plant, thereby achieving precise measurement of the overall appearance of the plant.

Description

Technical Field

[0001] This invention relates to the field of smart agriculture technology, and in particular to a method and system for measuring and calibrating the overall appearance of plants in smart agriculture based on computer vision. Background Technology

[0002] In the field of smart agriculture, accurate measurement and calibration of plant growth status is key to achieving scientific planting management and improving crop yield and quality.

[0003] The prior art CN111578837A discloses a plant morphology visual tracking and measurement method for agricultural robot operations, including: acquiring main stem image information of the plant under multiple fields of view, and determining the three-dimensional morphology information of the plant main stem based on the multiple main stem image information; wherein, each main stem image information includes a tracking target reference point located at the lowest end of the center line of the plant main stem in the corresponding field of view and a tracking start reference point located in the middle of the plant main stem; acquiring main stem image information of the plant under multiple fields of view includes: using the tracking start reference point in the i-th field of view as the tracking target reference point in the (i+1)-th field of view.

[0004] However, plants are often planted in large quantities. When measuring and calibrating plants, the average value of the area is often used instead of the measurement results. It is difficult to systematically measure and calibrate the overall appearance of a single plant, and thus it is impossible to accurately judge the state of the plant at different growth stages. This results in large errors in the measurement results and cannot provide accurate and reliable data support for agricultural production. Summary of the Invention

[0005] The purpose of this invention is to solve the problems in the background art by proposing a smart agriculture plant overall measurement and calibration method and system based on computer vision.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A computer vision-based method for measuring and calibrating the overall plant topography in smart agriculture, which includes the following steps:

[0008] Step 1: Real-time acquisition of growth images of the planting area using image acquisition equipment;

[0009] Step 2: Based on the growth images, construct a regional simulation model of the planting area;

[0010] Step 3: Based on the start time of the target plant, set the start time point, obtain the regional simulation model, identify the target plant in the regional simulation model, and when the target plant is identified, measure the longitudinal distance between any two adjacent target plants in the regional simulation model, and generate a normal growth signal based on the longitudinal distance.

[0011] Step 4: When a normal growth signal is detected, an independent growth zone is set for the target plant based on the longitudinal distance value;

[0012] Step 5: Obtain unit images of independent growth areas, set the growth point of the target plant in the unit images, identify the plant branches and leaves in the unit images, determine whether there is a positional connection between the plant branches and leaves and the growth point, identify associated branches and leaves and unrelated branches and leaves, and set interference states for independent growth areas based on the associated and unrelated branches and leaves in the unit images.

[0013] Step 6: Based on the interference status of the independent growth area, mark the associated branches and leaves in the unit image, and measure the target plant in the independent growth area according to the marked position to obtain growth information.

[0014] As a further aspect of the present invention, the start of development time refers to the time when the target plant begins to grow. If the target plant is an annual plant, the initial planting time of the target plant is marked as the start time point. If the target plant is a perennial plant, the start time of the germination period is marked as the start time point.

[0015] As a further aspect of the present invention, the method for measuring the longitudinal distance value includes:

[0016] S1: Starting from the start time, acquire growth images of the target plants in the planting area, acquire growth images collected at the same time, and stitch the growth images according to the acquisition position of each growth image, and construct a regional simulation model based on the image scenes in the growth images.

[0017] S2: Based on the basic planting information of the target plants, extract the planting distance of the target plants. The planting distance includes the standard row spacing and the standard plant spacing. The standard plant spacing refers to the scientific planting space distance between plants, and the standard row spacing refers to the scientific planting distance between rows.

[0018] The region simulation model is acquired, and the target plants in the region simulation model are identified. When the target plants are detected in the region simulation model, a measurement signal is generated.

[0019] When a measurement signal is detected, a regional simulation model is acquired, and the positions of the target plants are marked with mass points in the regional simulation model. Using a measurement tool, the distance between any two adjacent mass points is measured, and the measured data is marked as the lateral distance value DHi. Here, the lateral distance value DHi represents the plant spacing between target plants, and i represents the number of different plant spacing intervals. At the same time, the longitudinal distance in the regional simulation model is measured, and the midpoint of the longitudinal distance is taken and marked as the longitudinal boundary position. Here, the longitudinal distance is the distance between rows in the planting area.

[0020] As a further aspect of the present invention, the method for generating normal growth signals includes:

[0021] When the lateral distance value DHi is obtained, it is first compared with the standard plant spacing of the target plant. If the lateral distance value DHi is within the standard plant spacing range, a normal growth signal is generated. Otherwise, if the lateral distance value DHi is not within the standard plant spacing range, an abnormal growth signal is generated and transmitted to the terminal display module.

[0022] As a further aspect of the present invention, multiple acquisition devices are installed in the planting area. Each acquisition device acquires a regional image of a local area in the planting area. The regional images acquired by multiple acquisition devices are stitched together to form a complete regional image of the planting area. Then, a regional simulation model is constructed based on the complete regional image. The regional simulation model is scaled relative to the actual planting area according to a preset ratio. At the same time, the regional simulation model is updated in real time based on the real-time acquired growth images.

[0023] As a further aspect of the present invention, if the target plant is initially planted in the planting area in the form of seeds, in the initial stage, since the seeds are buried by the soil, the specific location of the target plant in the planting area cannot be identified when the image is acquired. Therefore, it is necessary to monitor the planting area in real time until the seedling of the target plant is detected, and then generate a measurement signal.

[0024] As a further aspect of the present invention, the method for setting up independent growth regions includes:

[0025] When a normal growth signal is detected, a regional simulation model is obtained. In the regional simulation model, adjacent particles are connected by line segments to obtain distance segments. The midpoint of the distance segments is taken, the location of the midpoint is identified, and this location is marked as the lateral local boundary point.

[0026] By extending straight lines at the longitudinal boundary and the lateral partial boundary, the lines intersect. Based on the intersection points, the areas enclosed by the lines are obtained. At this time, each area contains one and only one target plant. This area is then marked as the independent growth area of ​​the target plant.

[0027] As a further aspect of the present invention, the method for setting the interference state of the independent growth region includes:

[0028] SS1: Acquire real-time growth images and update the region simulation model in real time based on the growth images;

[0029] Arbitrarily select an independent growth region as the unit analysis region. Starting from the signal generation time, collect images of the unit analysis module separately in the region simulation model and mark them as unit images.

[0030] Starting from the initial time, the acquired unit images are arranged sequentially according to time order to obtain an image sequence. In the image sequence, the acquisition interval between any two adjacent unit images is a fixed value.

[0031] SS2: First, select the first unit image in the image sequence, and identify the growth point of the target plant in the unit image based on this unit image. The growth point refers to the initial growth position of the target plant.

[0032] Following the image sequence, select the unit image at the second position, identify all the plant branches and leaves in the second unit image, and sequentially identify whether there is a positional connection between the plant branches and leaves and the growth point at each position. Here, positional connection refers to the connection between the plant branches and leaves and the growth point through the stem or branch.

[0033] If a positional connection exists, the plant branches and leaves at this position are marked as associated branches and leaves; otherwise, if no positional connection exists, the plant branches and leaves at this position are marked as unrelated branches and leaves. Traverse the plant branches and leaves in the unit image. When all plant branches and leaves in the unit image are associated branches and leaves, mark the disturbance state of the independent growth region corresponding to this unit image as convergent growth; otherwise, if there are unrelated branches and leaves in the plant branches and leaves in the unit image, mark the disturbance state of the independent growth region corresponding to this unit image as cross growth.

[0034] As a further aspect of the present invention, the method for associating branch and leaf markers in a unit image includes:

[0035] Identify the interference state of the independent growth area at this time. If the interference state is convergent growth, mark the position of the plant branches and leaves in this unit image directly, and measure according to the marked position to obtain the real-time growth information of the target plant in this independent growth area.

[0036] If the interference state is cross growth, obtain the associated branches and leaves in the unit image, process the associated branches and leaves according to the processing method in the above convergent growth, then obtain the irrelevant branches and leaves, identify the orientation of the irrelevant branches and leaves in the unit image, select the independent growth region adjacent to this orientation, obtain the unit image of this independent growth region, and mark it as the neighboring image. Mark the image where the irrelevant branches and leaves are located as the trunk image.

[0037] The main trunk image and adjacent side images are stitched together according to the image acquisition positions to obtain a stitched image. Then, in the stitched image, the positions of irrelevant branches and leaves in the main trunk image are obtained, and the related branches and leaves in the adjacent side images are identified. At the same time, the related branches and leaves at the positions connected to the main trunk image are selected, and image matching is performed between these related branches and leaves and the irrelevant branches and leaves. Using an intelligent algorithm, the matching result between the related branches and leaves and the irrelevant branches and leaves at the stitching position is calculated. If the matching result shows that the matching result is irrelevant, a corresponding abnormal signal is generated based on this irrelevant branch and leaf and transmitted to the device terminal of the relevant management personnel. Conversely, if the matching result shows that the matching result is relevant, the irrelevant branches and leaves in the main trunk image are marked as related branches and leaves in the adjacent side images. At the same time, when marking the target plant in the independent growth area of ​​the adjacent side images, the marking positions of all related branches and leaves of the target plant are obtained, and the target plant is measured according to the marking positions to obtain the real-time growth information of the target plant in this independent growth area.

[0038] A computer vision-based smart agriculture plant panorama measurement and calibration system includes:

[0039] The information collection module is used to collect basic planting information for the planting area;

[0040] The image acquisition module is used to acquire growth images of the planting area;

[0041] The model building module is used to construct a regional simulation model of the planting area based on the growth image;

[0042] The node setting module is used to set the start time point based on the start time of the target plant's development.

[0043] The growth detection module is used to acquire the regional simulation model after the start time point, identify the target plants, and then measure the longitudinal distance between any two adjacent target plants in the regional simulation model, and generate normal growth signals based on the longitudinal distance.

[0044] The region segmentation module is used to detect normal growth signals and set independent growth regions based on longitudinal distance values;

[0045] The image analysis module is used to acquire unit images of independent growth areas, set the growth point of the target plant, then identify the plant branches and leaves in the unit images, determine whether there is a positional connection between the plant branches and leaves and the growth point, and identify associated and unrelated branches and leaves. Based on the associated and unrelated branches and leaves in the unit images, the module sets the interference state for the independent growth areas.

[0046] The measurement and processing module is used to mark the associated branches and leaves in a unit image according to the interference status of independent growth areas, and to measure the target plant in each independent growth area according to the marked position to obtain growth information.

[0047] The terminal display module is used to display the growth information of the target plants, so that the managers of the planting area can understand the growth information of the target plants in real time.

[0048] Compared with existing technologies, the advantages of this invention are:

[0049] This invention constructs a regional simulation model of the planting area. Based on the starting development time of the target plant, it measures the longitudinal distance between any two adjacent target plants in the simulation model. A normal growth signal is generated based on this longitudinal distance. A region division module sets independent growth regions based on the longitudinal distance values. An image analysis module further identifies the relationship between branches and leaves and growth points in a unit image, distinguishing between related and irrelevant branches and leaves, and accurately determining interference states. This process achieves refined analysis of the plant growth environment, effectively eliminating the interference of environmental factors on the measurement results. This makes the measurement results more accurately reflect the actual growth status of individual plants, improving the accuracy and reliability of the measurements. Subsequently, related branches and leaves are marked and measured according to the interference state to obtain accurate plant growth information. This achieves precise measurement of the overall plant appearance, providing accurate data support for precise fertilization, irrigation, and pest and disease control in smart agriculture, helping to improve agricultural resource utilization efficiency and reduce production costs. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0051] Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0052] 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.

[0053] Reference Figure 1 and Figure 2 A computer vision-based method for measuring and calibrating the overall plant topography in smart agriculture, which specifically includes the following steps:

[0054] Step 1: Collect basic planting information for the planting area. This information includes the types of plants planted in the area and their corresponding growth information, such as planting distance, plant growth curve, ambient humidity, ambient temperature, light intensity, and duration.

[0055] Step 2: Install multiple image acquisition devices in the planting area. These devices are used to capture images of the plants in the planting area and mark the captured images as growth images. The locations captured by the multiple image acquisition devices will fully cover the planting area, thereby preventing blind spots in the monitoring of the planting area.

[0056] Step 3: Mark the time when the target plant begins to develop as the start time point. Here, the target plant refers to the plant currently planted in the planting area, and the start time of development refers to the time when the target plant begins to grow. For example, if the target plant is an annual plant (wheat, rice, corn, etc.), the start time of planting the target plant is marked as the start time point. If the target plant is a perennial plant (apple tree, peach tree, pear tree, etc.), the start time of budding is marked as the start time point.

[0057] Starting from the start time, growth images of the target plants within the planting area are acquired and labeled as target analysis images. Computer vision technology is then used to identify independent growth regions within these images. Specifically, the methods for identifying independent growth regions include:

[0058] S1: Acquire growth images collected at the same time, stitch the growth images together according to the acquisition position of each growth image, construct a regional simulation model based on the image scenes in the growth images, and then update the regional simulation model in real time based on the real-time acquired growth images.

[0059] Furthermore, multiple acquisition devices are installed in the planting area. Each acquisition device collects a regional image of a local area in the planting area. The regional images collected by multiple acquisition devices are stitched together to form a complete regional image of the planting area. Then, a regional simulation model is constructed based on the complete regional image. The regional simulation model is scaled relative to the actual planting area according to a preset ratio. The specific preset ratio is set by those skilled in the art based on big data experience.

[0060] S2: Based on the basic planting information of the target plants, extract the planting distance of the target plants. The planting distance includes the standard row spacing and the standard plant spacing. The standard plant spacing refers to the scientific planting space distance between plants, and the standard row spacing refers to the scientific planting distance between rows. For example, the standard row spacing of tomatoes is 50-60cm and the standard plant spacing is 30-40cm. The standard row spacing of rice is 20-30cm and the standard plant spacing is 15-20cm.

[0061] The region simulation model is acquired, and the target plants in the region simulation model are identified. When the target plants are detected in the region simulation model, a measurement signal is generated.

[0062] When the system detects a measurement signal, it acquires a regional simulation model and marks the position of the target plant with mass points in the regional simulation model. Using a measurement tool, it measures the distance between any two adjacent mass points and marks the measured data as the lateral distance value DHi, where the lateral distance value DHi represents the plant spacing between target plants and i represents the number of different plant spacing intervals. At the same time, it measures the longitudinal distance in the regional simulation model and takes the midpoint of the longitudinal distance, marking this position as the longitudinal boundary position. The longitudinal distance is the distance between rows in the planting area.

[0063] It should be further explained that if the target plant is initially planted in the planting area in the form of seeds, the specific location of the target plant in the planting area cannot be identified when the image is acquired because the seeds are buried in the soil in the initial stage. Therefore, it is necessary to monitor the planting area in real time until the seedling of the target plant is detected before generating a measurement signal.

[0064] S3: When the lateral distance value DHi is obtained, the lateral distance value DHi is first compared with the standard plant spacing of the target plant. If the lateral distance value DHi is within the standard plant spacing range, a normal growth signal is generated. Conversely, if the lateral distance value DHi is not within the standard plant spacing range, an abnormal growth signal is generated and transmitted to the terminal display module. The terminal response module generates corresponding audio-visual reminder information based on the abnormal growth signal and provides real-time reminders to relevant management personnel.

[0065] When a normal growth signal is detected, a regional simulation model is obtained. In the regional simulation model, adjacent particles are connected by line segments to obtain distance segments. The midpoint of the distance segments is taken, the location of the midpoint is identified, and this location is marked as the lateral local boundary point.

[0066] S4: Extend the straight lines at the longitudinal boundary and the lateral partial boundary to make the straight lines intersect. Based on the intersection points, obtain the regions enclosed by the straight lines. At this time, there is one and only one target plant in each region. Then mark this region as the independent growth area of ​​the target plant.

[0067] Step 4: Once all independent growth zones in the planting area have been determined, the growth process of individual target plants within each independent growth zone is tracked in real time, and interference states in each independent growth zone are identified. Interference states include convergent growth and cross-growth. Specific methods for identifying interference states include:

[0068] SS1: Acquire real-time growth images and update the region simulation model in real time based on the growth images;

[0069] Choose any independent growth region as the unit analysis region, and take this unit analysis region as an example. Starting from the signal generation time, collect images of the unit analysis module separately in the region simulation model and mark them as unit images.

[0070] Starting from the initial time, the collected unit images are arranged sequentially according to time order to obtain an image sequence. In the image sequence, the acquisition interval between any two adjacent unit images is a fixed value, and the specific value of this fixed value is set by those skilled in the art based on big data experience.

[0071] SS2: First, select the first unit image in the image sequence, and identify the growth point of the target plant in the unit image based on this unit image. The growth point refers to the initial growth position of the target plant.

[0072] Following the image sequence, select the unit image at the second position, identify all the plant branches and leaves in the second unit image, and then sequentially identify whether there is a positional connection between the plant branches and leaves and the growth point at each position. Here, positional connection refers to the connection between the plant branches and leaves and the growth point through the stem or branch. For example, the leaves on the branch grow on the branch, and the branch is connected to the trunk (the growth point of the main trunk). This connection relationship of leaf → branch → trunk is positional connection.

[0073] If a positional connection exists, the plant branches and leaves at this position are marked as associated branches and leaves; otherwise, if no positional connection exists, the plant branches and leaves at this position are marked as unrelated branches and leaves. Traverse the plant branches and leaves in the unit image. When all plant branches and leaves in the unit image are associated branches and leaves, mark the interference state of the independent growth region corresponding to this unit image as convergent growth; otherwise, if there are unrelated branches and leaves in the plant branches and leaves in the unit image, mark the interference state of the independent growth region corresponding to this unit image as cross growth.

[0074] Step 5: Obtain the disturbance status of the independent growth area corresponding to each target plant, and measure the overall growth status of the target plant based on the disturbance status. The specific measurement methods include:

[0075] Acquire a real-time unit image of the independent growth region and identify the interference state of the independent growth region at this time. If the interference state is convergent growth, directly mark the position of the plant branches and leaves in this unit image and measure according to the marked position to obtain the real-time growth information of the target plant in this independent growth region.

[0076] If the interference state is cross-growth, first obtain the associated branches and leaves in the unit image, and process the associated branches and leaves according to the processing method in the above-mentioned convergent growth. Then obtain the irrelevant branches and leaves, and identify the orientation of the irrelevant branches and leaves in the unit image. At the same time, based on this orientation, select the independent growth region adjacent to this orientation, and obtain the unit image of this independent growth region. Mark this unit image as the neighboring image, and mark the image where the irrelevant branches and leaves are located as the trunk image.

[0077] The main trunk image and adjacent side images are stitched together according to the image acquisition positions to obtain a stitched image. Then, in the stitched image, the positions of irrelevant branches and leaves in the main trunk image are obtained, and the related branches and leaves in the adjacent side images are identified. At the same time, the related branches and leaves at the positions connected to the main trunk image are selected, and image matching is performed between these related branches and leaves and the irrelevant branches and leaves. Using an intelligent algorithm, the matching result between the related branches and leaves and the irrelevant branches and leaves at the stitching position is calculated. If the matching result shows that the matching result is irrelevant, a corresponding abnormal signal is generated based on this irrelevant branch and leaf and transmitted to the device terminal of the relevant management personnel. The relevant management personnel further manually process this irrelevant branch and leaf. Conversely, if the matching result shows that the matching result is relevant, the irrelevant branches and leaves in the main trunk image are marked as related branches and leaves in the adjacent side images. At the same time, when marking the target plant in the independent growth area of ​​the adjacent side images, the marking positions of all related branches and leaves of the target plant are obtained, and the target plant is measured according to the marking positions to obtain the real-time growth information of the target plant in this independent growth area.

[0078] A computer vision-based smart agriculture plant panorama measurement and calibration system includes:

[0079] The information collection module is used to collect basic planting information for the planting area;

[0080] The image acquisition module is used to acquire growth images of the planting area;

[0081] The model building module is used to construct a regional simulation model of the planting area based on the growth image;

[0082] The node setting module is used to set the start time point based on the start time of the target plant's development.

[0083] The growth detection module is used to acquire the regional simulation model after the start time point and identify the target plants in the regional simulation model. When a target plant is identified, the longitudinal distance between any two adjacent target plants in the regional simulation model is measured, and a normal growth signal is generated based on the longitudinal distance.

[0084] The region division module is used to detect normal growth signals and set independent growth regions for target plants in the region simulation model based on longitudinal distance values.

[0085] The image analysis module is used to acquire unit images of independent growth areas, set the growth point of the target plant in the unit image, identify the plant branches and leaves in the unit image, determine whether there is a positional connection between the plant branches and leaves and the growth point, and identify associated and unrelated branches and leaves. Based on the associated and unrelated branches and leaves in the unit image, an interference state is set for the independent growth area.

[0086] The measurement and processing module is used to mark the associated branches and leaves in a unit image according to the interference status of independent growth areas, and to measure the target plant in each independent growth area according to the marked position to obtain growth information.

[0087] The terminal display module is used to display the growth information of the target plants, so that the managers of the planting area can understand the growth information of the target plants in real time.

[0088] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for measuring and calibrating the overall plant topography in smart agriculture based on computer vision, characterized in that, The method specifically includes the following steps: Step 1: Real-time acquisition of growth images of the planting area using image acquisition equipment; Step 2: Based on the growth images, construct a regional simulation model of the planting area; Step 3: Based on the start time of the target plant, set the start time point, obtain the regional simulation model, identify the target plant in the regional simulation model, and when the target plant is identified, measure the longitudinal distance between any two adjacent target plants in the regional simulation model, and generate a normal growth signal based on the longitudinal distance. Step 4: When a normal growth signal is detected, an independent growth zone is set for the target plant based on the longitudinal distance value; Step 5: Obtain unit images of independent growth areas, set the growth point of the target plant in the unit images, identify the plant branches and leaves in the unit images, determine whether there is a positional connection between the plant branches and leaves and the growth point, identify associated branches and leaves and unrelated branches and leaves, and set interference states for independent growth areas based on the associated and unrelated branches and leaves in the unit images. The methods for setting the interference state include: SS1: Acquire real-time growth images and update the region simulation model in real time based on the growth images; Arbitrarily select an independent growth region as the unit analysis region. Starting from the signal generation time, collect images of the unit analysis module separately in the region simulation model and mark them as unit images. Starting from the initial time, the acquired unit images are arranged sequentially according to time order to obtain an image sequence. In the image sequence, the acquisition interval between any two adjacent unit images is a fixed value. SS2: First, select the first unit image in the image sequence, and identify the growth point of the target plant in the unit image based on this unit image. The growth point refers to the initial growth position of the target plant. Following the image sequence, select the unit image at the second position, identify all the plant branches and leaves in the second unit image, and sequentially identify whether there is a positional connection between the plant branches and leaves and the growth point at each position. Here, positional connection refers to the connection between the plant branches and leaves and the growth point through the stem or branch. If a positional connection exists, the plant branches and leaves at this position are marked as associated branches and leaves; otherwise, if no positional connection exists, the plant branches and leaves at this position are marked as unrelated branches and leaves. Traverse the plant branches and leaves in the unit image. When all plant branches and leaves in the unit image are associated branches and leaves, mark the interference state of the independent growth region corresponding to this unit image as convergent growth; otherwise, if there are unrelated branches and leaves in the plant branches and leaves in the unit image, mark the interference state of the independent growth region corresponding to this unit image as cross growth. Step 6: Based on the interference status of the independent growth area, mark the associated branches and leaves in the unit image, and measure the target plant in the independent growth area according to the marked position to obtain growth information.

2. The method for measuring and calibrating the overall plant topography in smart agriculture based on computer vision according to claim 1, characterized in that, The start time of development refers to the time when the target plant begins to grow. If the target plant is an annual, the initial planting time of the target plant is marked as the start time point. If the target plant is a perennial, the start time of the budding period is marked as the start time point.

3. The method for measuring and calibrating the overall plant topography in smart agriculture based on computer vision according to claim 1, characterized in that, Methods for measuring longitudinal distance include: S1: Starting from the start time, acquire growth images of the target plants in the planting area, acquire growth images collected at the same time, and stitch the growth images according to the acquisition position of each growth image, and construct a regional simulation model based on the image scenes in the growth images. S2: Based on the basic planting information of the target plants, extract the planting distance of the target plants. The planting distance includes the standard row spacing and the standard plant spacing. The standard plant spacing refers to the scientific planting space distance between plants, and the standard row spacing refers to the scientific planting distance between rows. The region simulation model is acquired, and the target plants in the region simulation model are identified. When the target plants are detected in the region simulation model, a measurement signal is generated. When a measurement signal is detected, a regional simulation model is acquired, and the positions of the target plants are marked with mass points in the regional simulation model. Using a measurement tool, the distance between any two adjacent mass points is measured, and the measured data is marked as the lateral distance value DHi. Here, the lateral distance value DHi represents the plant spacing between target plants, and i represents the number of different plant spacing intervals. At the same time, the longitudinal distance in the regional simulation model is measured, and the midpoint of the longitudinal distance is taken and marked as the longitudinal boundary position. Here, the longitudinal distance is the distance between rows in the planting area.

4. The method for measuring and calibrating the overall plant topography in smart agriculture based on computer vision according to claim 3, characterized in that, Methods for generating normal growth signals include: When the lateral distance value DHi is obtained, it is first compared with the standard plant spacing of the target plant. If the lateral distance value DHi is within the standard plant spacing range, a normal growth signal is generated. Otherwise, if the lateral distance value DHi is not within the standard plant spacing range, a growth abnormality signal is generated and transmitted to the terminal display module.

5. The method for measuring and calibrating the overall plant topography in smart agriculture based on computer vision according to claim 3, characterized in that, Multiple acquisition devices are installed in the planting area. Each acquisition device captures a local area image of the planting area. The images captured by multiple acquisition devices are stitched together to form a complete image of the planting area. Then, a regional simulation model is built based on the complete regional image. The regional simulation model is scaled relative to the actual planting area according to a preset ratio. At the same time, the regional simulation model is updated in real time based on the real-time acquired growth images.

6. The method for measuring and calibrating the overall plant topography in smart agriculture based on computer vision according to claim 4, characterized in that, If the target plant is initially planted in the planting area in the form of seeds, the specific location of the target plant in the planting area cannot be identified during image acquisition because the seeds are buried in the soil in the initial stage. Therefore, it is necessary to monitor the planting area in real time until the seedling of the target plant is detected, and then generate a measurement signal.

7. The method for measuring and calibrating the overall plant topography in smart agriculture based on computer vision according to claim 1, characterized in that, Methods for setting up independent growth zones include: When a normal growth signal is detected, a regional simulation model is obtained. In the regional simulation model, adjacent particles are connected by line segments to obtain distance segments. The midpoint of the distance segments is taken, the location of the midpoint is identified, and this location is marked as the lateral local boundary point. By extending straight lines at the longitudinal boundary and the lateral partial boundary, the lines intersect. Based on the intersection points, the areas enclosed by the lines are obtained. At this time, each area contains one and only one target plant. This area is then marked as the independent growth area of ​​the target plant.

8. The method for measuring and calibrating the overall plant topography in smart agriculture based on computer vision according to claim 1, characterized in that, Methods for associating branch and leaf labels in a unit image include: Identify the interference state of the independent growth area at this time. If the interference state is convergent growth, mark the position of the plant branches and leaves in this unit image directly, and measure according to the marked position to obtain the real-time growth information of the target plant in this independent growth area. If the interference state is cross growth, obtain the associated branches and leaves in the unit image, process the associated branches and leaves according to the processing method in the above convergent growth, then obtain the irrelevant branches and leaves, identify the orientation of the irrelevant branches and leaves in the unit image, select the independent growth region adjacent to this orientation, obtain the unit image of this independent growth region, and mark it as the neighboring image. Mark the image where the irrelevant branches and leaves are located as the trunk image. The main trunk image and adjacent side images are stitched together according to the image acquisition positions to obtain a stitched image. Then, in the stitched image, the positions of irrelevant branches and leaves in the main trunk image are obtained, and the related branches and leaves in the adjacent side images are identified. At the same time, the related branches and leaves at the positions connected to the main trunk image are selected, and image matching is performed between these related branches and leaves and the irrelevant branches and leaves. Using an intelligent algorithm, the matching result between the related branches and leaves and the irrelevant branches and leaves at the stitching position is calculated. If the matching result shows that the matching result is irrelevant, a corresponding abnormal signal is generated based on this irrelevant branch and leaf and transmitted to the device terminal of the relevant management personnel. Conversely, if the matching result shows that the matching result is relevant, the irrelevant branches and leaves in the main trunk image are marked as related branches and leaves in the adjacent side images. At the same time, when marking the target plant in the independent growth area of ​​the adjacent side images, the marking positions of all related branches and leaves of the target plant are obtained, and the target plant is measured according to the marking positions to obtain the real-time growth information of the target plant in this independent growth area.

9. A computer vision-based intelligent agriculture plant overall measurement and calibration system, wherein the system uses the computer vision-based intelligent agriculture plant overall measurement and calibration method according to any one of claims 1-8, characterized in that, include: The information collection module is used to collect basic planting information for the planting area; The image acquisition module is used to acquire growth images of the planting area; The model building module is used to construct a regional simulation model of the planting area based on the growth image; The node setting module is used to set the start time point based on the start time of the target plant's development. The growth detection module is used to acquire the regional simulation model after the start time point, identify the target plants, and then measure the longitudinal distance between any two adjacent target plants in the regional simulation model, and generate normal growth signals based on the longitudinal distance. The region segmentation module is used to detect normal growth signals and set independent growth regions based on longitudinal distance values; The image analysis module is used to acquire unit images of independent growth areas, set the growth point of the target plant, then identify the plant branches and leaves in the unit images, determine whether there is a positional connection between the plant branches and leaves and the growth point, and identify associated and unrelated branches and leaves. Based on the associated and unrelated branches and leaves in the unit images, the module sets the interference state for the independent growth areas. The measurement and processing module is used to mark the associated branches and leaves in a unit image according to the interference status of independent growth areas, and to measure the target plant in each independent growth area according to the marked position to obtain growth information. The terminal display module is used to display the growth information of the target plants, so that the managers of the planting area can understand the growth information of the target plants in real time.

Images