Corn plant multi-parameter measurement system and measurement method

Abstract

The invention discloses a corn plant multi-parameter measurement system, which comprises a multi-mode sensing unit used for synchronously acquiring corn plant height, corn cob size and corn cob internal structure data; the ground intelligent mobile platform carries the multi-mode sensing unit and has autonomous navigation and obstacle avoidance capabilities; the edge calculation unit integrates a machine learning model to process multi-source data in real time; and the digital twinborn visualization module is used for generating a farmland three-dimensional growth model and an AR auxiliary diagnosis interface. The system is based on a ground intelligent mobile platform, adopts a multi-mode sensing fusion technology, can synchronously measure the plant height, the size of a corn cob, the internal structure of a stalk and grain components in a non-contact and high-precision manner, and can effectively solve the technical problems existing in manual measurement. The invention further discloses a measurement method, according to the method, the multi-modal sensing data fusion technology is adopted in the corn multi-parameter measurement process for the first time, and the accuracy and efficiency of corn growth data measurement can be remarkably improved.

Description

Technical Field

[0001] The present invention relates to a monitoring system based on a ground intelligent mobile platform and multi-modal sensor fusion, and in particular to a corn plant multi-parameter measurement system and a measurement method, belonging to the intersection field of precision agricultural equipment and intelligent sensor technology. Background Art

[0002] During the breeding process of new corn varieties, the growth status of corn is measured at various stages of corn growth, such as measuring corn plant height, corn cob size, etc.

[0003] There are several traditional methods for measuring corn plant height:

[0004] 1. Manual tape measure or tower ruler measurement

[0005] The traditional method mainly relies on manual measurement using a tape measure or a tower ruler. The operation steps include: inserting the tower ruler vertically into the soil, the measurement personnel holding the tape measure from the ground to the top of the corn tassel, recording the value and calculating the average value. Some improved solutions use a retractable rod with a ruler to calculate the plant height based on the principle of similar triangles, but it still requires manual alignment of the measuring rod and the top of the plant.

[0006] 2. Bamboo pole assisted measurement

[0007] Before heading, the surveyor inserts a bamboo pole into the soil, and another person holds a tape measure to measure the vertical distance between the bamboo pole and the highest point of the leaf. After heading, the length from the bamboo pole to the top of the male ear is directly measured, and efficiency is improved through collaboration among multiple people.

[0008] 3. Digital cameras and image processing

[0009] Some studies use digital cameras to take plant images and extract plant height data through image processing technology. For example, a mobile phone camera is used to take images of a ruler and a plant, and algorithm analysis is combined to achieve non-contact measurement.

[0010] The above measurement method has the following defects:

[0011] 1. Conflict between efficiency and accuracy

[0012] Manual measurement relies on manual operation, which is limited by the vision, experience and physical strength of the surveyor. A single measurement takes a long time, about 3-5 minutes per plant, and the repeatability error can reach ±2-5cm. In addition, in the late stage of corn growth, since the surveyors need to go deep into the corn field to measure, they are easily scratched by corn leaves, and some surveyors are allergic to corn pollen.

[0013] Although the digital camera solution can improve efficiency, it requires professional image processing software and is affected by factors such as lighting and leaf occlusion, so the accuracy of ear tip recognition is less than 70%.

[0014] 2. Poor environmental adaptability

[0015] Traditional tower rulers and tape measures are prone to failure in complex field environments such as rain, fog, and straw obstruction. The measuring rod cannot fit the plant perfectly, resulting in data deviation.

[0016] Although new equipment such as LiDAR has a higher accuracy, reaching ±0.5cm, it needs to be carried by drones, and the flight altitude is limited, usually >20m, and the cost is high, with the unit price of the equipment >50,000 yuan.

[0017] 3. Single function

[0018] Existing equipment can only measure plant height, but cannot simultaneously obtain key parameters such as stem diameter and three-dimensional size of the ear. It also lacks dynamic obstacle avoidance capabilities and is easily affected by obstacles in the field.

[0019] 4. Destructive testing issues

[0020] Some studies obtain internal structure data by cutting plants. Although this can accurately measure indicators such as stem density, it will disrupt the normal growth of the plants and cannot achieve long-term monitoring.

[0021] Therefore, there is an urgent need to design a new multi-parameter measurement method for corn plants to solve the above technical problems. Summary of the invention

[0022] In view of this, one of the objects of the present invention is to provide a multi-parameter measurement system for corn plants. The multi-parameter measurement system for corn plants is based on a ground intelligent mobile platform and adopts multimodal sensing fusion technology. It can synchronously measure plant height, corn cob size, stalk internal structure and grain composition in a non-contact and high-precision manner, which can effectively solve the technical problems existing in manual measurement.

[0023] The present invention solves the above technical problems by the following technical means:

[0024] The corn plant multi-parameter measurement system of the present invention is mainly composed of four modules:

[0025] 1. Multimodal sensing unit, used to simultaneously collect data on corn plant height, corn cob size, and corn cob internal structure;

[0026] 2. A ground intelligent mobile platform equipped with the multimodal sensing unit and having autonomous navigation and obstacle avoidance capabilities;

[0027] 3. Edge computing unit, integrating machine learning models to process multi-source data in real time;

[0028] 4. Digital twin visualization module, generating a three-dimensional growth model of farmland and an AR-assisted diagnosis interface.

[0029] Furthermore, the multimodal sensing unit includes a multispectral camera array, a millimeter wave radar and a micro CT probe. The multispectral camera array is used to capture crop growth status images in real time, the millimeter wave radar is used to non-contactly measure plant height and stem diameter, and the micro CT probe is used to detect corn cobs to obtain three-dimensional size data. In order to be suitable for operation in corn planting areas, we have specially improved the existing micro CT probe. Specifically, the micro CT probe uses a liquid metal alloy probe with a diameter of ≤1mm, and the surface is covered with a biocompatible silicone layer. The silicone layer is designed in a sheet shape with a Shore hardness of 20HA to adapt to the curved surface of the corn cob; the working principle of the micro CT probe: the probe has a built-in X-ray source (power ≤5mW) and a detector array, and scans the cob with a rotation step of 0.1° to generate a 0.1mm resolution three-dimensional point cloud; the control method of the micro CT probe: the probe is adaptively fitted through a piezoelectric ceramic driver, and the contact pressure is controlled within the range of 0.1-0.5N to avoid damaging the corn tissue.

[0030] Furthermore, the ground intelligent mobile platform includes an autonomous driving vehicle, which includes a frame, wheels arranged on both sides of the frame, a power supply arranged in the middle of the frame and a control system connected to the power supply circuit, and a liftable robotic arm arranged on the top of the frame. The autonomous driving vehicle is equipped with a high-precision GNSS positioning system, the multi-spectral camera array and the micro CT probe are arranged at the end of the robotic arm, the millimeter-wave radar is arranged on the robotic arm, and the multi-spectral camera array, millimeter-wave radar and micro CT probe are respectively connected to the control system circuit.

[0031] Furthermore, the frame includes a chassis and wheel frames arranged on both sides of the chassis, the wheel frames include a swing plate, the wheels use hub motors, the hub motors include a stator and a rotor, the stator is fixedly connected to the lower part of the swing plate, the top of the swing plate is hinged to the chassis through a hinge shaft, the frame also includes an electric telescopic cylinder, one end of the electric telescopic cylinder is hinged to the chassis, and the other end of the electric telescopic cylinder is hinged to the middle part of the swing plate, and the electric telescopic cylinder and the hub motor are respectively connected to the control system circuit.

[0032] Furthermore, the edge computing unit includes:

[0033] Lightweight YOLOv9s-Plant neural network model for real-time identification of plant height, ear size, and pest and disease risks;

[0034] The LSTM-Transformer hybrid network predicts the plant height growth curve and the optimal pollination period based on historical data, and the training data of the LSTM-Transformer hybrid network includes at least 5000 groups of field measured plant height samples.

[0035] The second object of the present invention is to provide a method for measuring the growth parameters of corn plants based on a corn plant multi-parameter measurement system. This measurement method adopts multimodal sensor data fusion technology for the first time in the corn multi-parameter measurement process, which can significantly improve the accuracy and measurement efficiency of corn growth data measurement.

[0036] The present invention solves the above technical problems by the following technical means:

[0037] The measuring method of the present invention comprises the following steps:

[0038] Step 1: Based on the ridge depth and width, the control system and the electric telescopic cylinder are used to adjust the inclination angle of the swing plate, and then the sampling path of the ground intelligent mobile platform is planned;

[0039] Step 2: The ground intelligent mobile platform walks in the corn planting area according to the sampling path, and uses millimeter wave radar to obtain stalk profile data and calculate plant height and diameter;

[0040] Step 3: Scan the corn ear using a micro CT probe installed on a ground intelligent mobile platform to generate a three-dimensional model and extract the parameters of ear length and ear diameter;

[0041] Step 4: The edge computing unit integrates multi-source data and outputs visual reports and AR diagnostic suggestions.

[0042] Furthermore, the ground intelligent mobile platform described in step 2 obtains road information through a multi-spectral camera array and millimeter-wave radar to avoid obstacles. When the ground intelligent mobile platform moves to the designated sampling location, it stops moving and then uses the millimeter-wave radar to obtain stem contour data and calculate plant height and diameter.

[0043] Furthermore, the edge computing unit in step 2 adopts a lightweight YOLOv9s-Plant neural network model. The YOLOv9s-Plant model uses knowledge distillation technology to compress the number of parameters to 15% of the original model, and adopts the following data fusion rules:

[0044] H final =ω 1 H radar +ω 2 H CT

[0045] where ω 1 =0.6,ω 2 =0.4,ω 1 and ω 2 Corresponding to the data weights of radar and CT respectively.

[0046] The present invention has the following beneficial effects:

[0047] 1. High-precision non-contact measurement

[0048] Through the coordinated work of the millimeter-wave radar and the micro CT probe, the corn plant multi-parameter measurement system of the present invention can achieve a corn plant height measurement error of ≤0.5cm and an ear length measurement error of ≤0.2mm, which is 3-5 times more accurate than traditional manual measurement.

[0049] 2. Multi-dimensional data fusion

[0050] By integrating multimodal data such as millimeter-wave radar, CT scanning, and Raman spectroscopy, parameters such as stem diameter, three-dimensional size of the ear, and grain composition can be acquired simultaneously, breaking through the limitations of traditional single parameter measurement.

[0051] 3. Real-time dynamic processing

[0052] The edge computing unit is equipped with a lightweight YOLOv9s-Plant model with a single-frame processing time of <8ms, supporting real-time output of AR diagnostic suggestions while the ground intelligent mobile platform is driving.

[0053] 4. Precision agriculture support

[0054] Generate a three-dimensional growth model of farmland and an AR-assisted diagnosis interface, which can mark areas requiring fertilization with a nitrogen content of <2%, improving fertilization accuracy by more than 30%.

[0055] The present invention solves the defects of efficiency, accuracy and single function of traditional measurement methods through the collaborative innovation of technical performance breakthroughs and multi-scenario adaptability, providing accurate and efficient solutions for agricultural scientific research. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0057] Figure 1 It is a front view of the corn plant multi-parameter measurement system of the present invention;

[0058] Figure 2 for Figure 1 The schematic diagram of the structure after the A area is enlarged as shown in FIG.

[0059] Figure 3 It is a schematic diagram of the three-dimensional structure of the corn plant multi-parameter measurement system of the present invention;

[0060] Figure 4 It is a schematic diagram of the installation structure of the frame and wheels in the corn plant multi-parameter measurement system of the present invention;

[0061] Figure 5 The figure is a schematic diagram of the installation structure of the frame and the wheels in the corn plant multi-parameter measurement system of the present invention, in which the lower part of the wheel is extended outward at a certain angle. DETAILED DESCRIPTION

[0062] The present invention will be described in detail below with reference to the accompanying drawings:

[0063] Example 1: A corn plant multi-parameter measurement system

[0064] like Figures 1 to 5 As shown, the corn plant multi-parameter measurement system in this embodiment includes:

[0065] A multimodal sensing unit is used for synchronously collecting data on corn plant height, corn cob size and internal structure of corn cobs; the multimodal sensing unit comprises a multispectral camera array 1, a millimeter wave radar 2 and a micro CT probe 3, and the multispectral camera array 1 is used for capturing crop growth status images in real time.

[0066] A ground intelligent mobile platform is equipped with the multimodal sensing unit and has autonomous navigation and obstacle avoidance capabilities; the ground intelligent mobile platform includes an automatic driving car, the automatic driving car is equipped with a high-precision GNSS positioning system, the automatic driving car includes a frame 4, wheels 5 arranged on both sides of the frame, a power supply arranged in the middle of the frame and a control system 6 connected to the power supply circuit, and a liftable mechanical arm 7 arranged on the top of the frame, the liftable mechanical arm 7 includes an end, a forearm and a rear arm, the end, forearm and rear arm are hinged in sequence through machine joints, the multispectral camera array 1 and the micro CT probe 3 are arranged at the end of the mechanical arm, the millimeter wave radar 2 is arranged on the forearm, and the multispectral camera array, millimeter wave radar and micro CT probe are respectively connected to the control system circuit.

[0067] Specifically, the millimeter wave radar 2 uses a 40GHz frequency-modulated continuous wave radar. The millimeter wave radar 2 is set on the forearm and can scan the road conditions ahead and the corn stalk profile data according to the forearm movement, and can achieve 360° scanning of the stalk profile. The radar signal processing unit integrates a Doppler effect compensation algorithm to eliminate environmental noise interference such as raindrops and stalks, and the stalk diameter measurement accuracy reaches ±0.3mm.

[0068] The micro CT probe 3 uses a liquid metal alloy flexible probe with a diameter of 0.8 mm, and the surface is covered with a silicone layer or silicone plate with a hardness of 20HA, and a piezoelectric ceramic driver is used to achieve an adaptive contact pressure of 0.1-0.5 N. The X-ray source power is ≤3mW, and a cone beam CT scanning method is used. The single spike scanning time is ≤15 seconds, and the three-dimensional point cloud resolution is 0.05mm.

[0069] Multispectral camera array with 2 visible light channels (450-700nm) and 2 near-infrared channels

[0070] (700-900nm), equipped with a fisheye lens to achieve ±90° wide-angle imaging, and supports HDR dynamic range adjustment.

[0071] The frame 4 includes a chassis 41 and wheel frames 42 arranged on both sides of the chassis, the wheel frames include a swing plate 43, the wheels use hub motors, the hub motors include a stator and a rotor, the stator is fixedly linked to the lower part of the swing plate, the top of the swing plate is hinged to the chassis through a hinge shaft, the frame also includes an electric telescopic cylinder 44, one end of the electric telescopic cylinder is hinged to the chassis, and the other end of the electric telescopic cylinder is hinged to the middle part of the swing plate, the electric telescopic cylinder and the hub motor are respectively connected to the control system circuit.

[0072] The specific parameters are as follows:

[0073] 1. Self-driving car

[0074] The overall width of the car's frame plus the wheels on both sides is between 35-25cm. The frame adopts a modular aluminum chassis, and the wheels use hub motors with a torque of 10Nm and a speed of 60rpm. The hydrogen fuel cell pack has a battery life of ≥10 hours and supports working in an environment of -10℃ to 45℃.

[0075] The electric telescopic cylinder has a stroke of ±60mm and can achieve ±1° angle adjustment accuracy through a PID controller. A dynamic obstacle avoidance SLAM system is built in conjunction with RTK-GPS.

[0076] 2. Robotic arm control system

[0077] The end of the six-degree-of-freedom robotic arm is equipped with an integrated torque sensor (range 0-5N·m), and the RRT* algorithm is used to plan the scanning path of the ear. Obstacle avoidance is based on the fusion data of the multispectral camera array and the millimeter-wave radar. That is, during walking, the multispectral camera array and the millimeter-wave radar can be used for obstacle avoidance and navigation, while during corn data collection, the multispectral camera array and the millimeter-wave radar are used to collect crop growth status data.

[0078] The method for measuring growth parameters of corn plants by using a corn plant multi-parameter measurement system of the present invention comprises the following steps:

[0079] Step 1: Path planning and obstacle avoidance: Based on the ridge depth and width, the control system and electric telescopic cylinder are used to dynamically adjust the tilt angle of the swing plate, and then the sampling path of the ground intelligent mobile platform is planned; the ridge depth can be obtained through radar pre-scanning, the sampling path spacing is 60cm, and the D*Lite algorithm is used to update the obstacle avoidance path in real time. When encountering straw obstruction, switch to infrared thermal imaging mode to identify obstacles, and the obstacle avoidance speed is reduced to 1km / h.

[0080] Step 2: The ground intelligent mobile platform walks in the corn planting area according to the sampling path, and obtains the stem contour data through the millimeter wave radar to calculate the plant height and diameter; the ground intelligent mobile platform obtains road information through the multi-spectral camera array and millimeter wave radar to avoid obstacles, and when the ground intelligent mobile platform moves to the designated sampling location, it stops walking, and then obtains the stem contour data through the millimeter wave radar to calculate the plant height and diameter.

[0081] Step 3: Scan the corn ear using a micro CT probe installed on a ground intelligent mobile platform to generate a three-dimensional model and extract the parameters of ear length and ear diameter;

[0082] Step 4: The edge computing unit integrates multi-source data and outputs a visualization report and AR diagnostic suggestions. The edge computing unit uses a lightweight YOLOv9s-Plant neural network model. The YOLOv9s-Plant model uses knowledge distillation technology to compress the number of parameters to 15% of the original model, and uses the following data fusion rules:

[0083] H final =ω 1 H radar +ω 2 H CT

[0084] where ω 1 =0.6,ω 2 =0.4,ω 1 and ω 2 They correspond to the data weights of radar and CT respectively. Specifically, the millimeter-wave radar data is subjected to fast Fourier transform (FFT) to extract the stem diameter, the CT point cloud is reconstructed to generate a three-dimensional model through Poisson reconstruction, and the ear length and ear thickness parameters are fitted to the cylinder through the RANSAC algorithm. The edge computing unit deploys the YOLOv9s-Plant model with a parameter volume of 3.2MB. It combines the LSTM-Transformer hybrid network to predict the stem growth rate with an error of <4%. The data fusion weight is adjusted dynamically, and the CT weight is increased to 0.6 on rainy days. The lightweight YOLOv9s-Plant neural network model is a lightweight version optimized for target detection needs in the plant field based on the YOLOv9 architecture. It introduces multi-scale feature fusion or data enhancement strategies, such as affine transformation for leaf occlusion, to improve detection robustness. Its core goal is to reduce computational complexity while maintaining high detection accuracy, so it is suitable for deployment on mobile terminals or edge computing devices.

[0085] Example 2: Experimental verification and results

[0086] 1. Experimental Setup

[0087] Experimental location: The test was conducted in the corn experimental field of Shaanxi Yulin Academy of Agricultural Sciences, with a field area of ​​2 mu, a row spacing of 60 cm, and a simulated rain and fog weather scene.

[0088] Experimental equipment: A corn plant multi-parameter measurement system was used, in which the ground intelligent mobile platform had a speed of ≤2km / h; the ground intelligent mobile platform was equipped with two micro CT probes, covering a row spacing of 60cm.

[0089] 2. Data collection:

[0090] When it is detected that the corn plant enters the measurement range, the hydraulic lifting arm first descends to a height of 25 cm from the ground, and then gradually rises to measure the height of the corn cob. During the rising process, when the micro-CT probe approaches the corn cob, the micro-CT probe rotates to scan the ear, and simultaneously acquires the three-dimensional model and Raman spectral data.

[0091] 3. Data processing and output:

[0092] The edge computing terminal runs the YOLOv9s-Plant model, outputs plant height (±0.5cm), ear length (±0.2mm) and grain protein level, and generates an Excel file containing a three-dimensional growth model, AR diagnostic interface and carbon sink measurement report; the system automatically generates a heat map report, marks the areas requiring fertilization and uploads it to the cloud through a 5G base station.

[0093] 4. Performance indicators

[0094] Plant height measurement: The average time for a single plant is 8.2 seconds, and the plant height accuracy is ±0.3cm (compared to manual measurement of ±5cm).

[0095] Ear parameters: Ear length measurement accuracy is ±0.15mm, ear diameter measurement accuracy is ±0.2mm.

[0096] Obstacle avoidance efficiency: obstacle recognition rate in rainy scenes is 92%, and interruption recovery time is ≤ 30 seconds.

[0097] 5. Comparison of experimental results

[0098] index Traditional manual measurement Existing equipment (LiDAR) This system Plant height precision ±5cm ±0.5cm ±0.3cm Three-dimensional parameters of ear Unable to obtain Destructive cutting required Contactless acquisition Operation efficiency (mu / day) 0.053 0.512 1.02 Environmental adaptability Difference Some weather conditions may apply All-weather operation

[0099] Through the above-mentioned refinements, the embodiments are significantly superior to the existing technologies in terms of hardware integration, data processing efficiency and environmental adaptability, providing accurate and efficient solutions for agricultural scientific research.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A corn plant multi-parameter measurement system, characterized in that: include: A multimodal sensing unit for synchronously collecting data on corn plant height, corn cob size, and corn cob internal structure; A ground intelligent mobile platform equipped with the multimodal sensing unit and having autonomous navigation and obstacle avoidance capabilities; Edge computing unit, integrating machine learning models to process multi-source data in real time; The digital twin visualization module generates a three-dimensional growth model of farmland and an AR-assisted diagnosis interface.

2. The system according to claim 1, characterized in that The multimodal sensing unit includes a multispectral camera array, a millimeter-wave radar and a micro CT probe. The multispectral camera array is used to capture crop growth status images in real time, the millimeter-wave radar is used to non-contact measure plant height and stem diameter, and the micro CT probe is used to detect corn ears to obtain three-dimensional size data.

3. The system according to claim 2, characterized in that The ground intelligent mobile platform includes an autonomous driving vehicle, which includes a frame, wheels arranged on both sides of the frame, a power supply arranged in the middle of the frame and a control system connected to the power supply circuit, and a liftable mechanical arm arranged on the top of the frame. The autonomous driving vehicle is equipped with a high-precision GNSS positioning system, the multi-spectral camera array and the micro CT probe are arranged at the end of the mechanical arm, the millimeter-wave radar is arranged on the mechanical arm, and the multi-spectral camera array, millimeter-wave radar and micro CT probe are respectively connected to the control system circuit.

4. The system according to claim 3, characterized in that The frame includes a chassis and wheel frames arranged on both sides of the chassis, the wheel frames include a swing plate, the wheels use hub motors, the hub motors include a stator and a rotor, the stator is fixedly linked to the lower part of the swing plate, the top of the swing plate is hinged to the chassis through a hinge shaft, the frame also includes an electric telescopic cylinder, one end of the electric telescopic cylinder is hinged to the chassis, and the other end of the electric telescopic cylinder is hinged to the middle part of the swing plate, the electric telescopic cylinder and the hub motor are respectively connected to the control system circuit.

5. The system according to claim 4, characterized in that The edge computing unit comprises: Lightweight YOLOv9s-Plant neural network model for real-time identification of plant height, ear size, and pest and disease risks; The LSTM-Transformer hybrid network predicts the plant height growth curve and the optimal pollination period based on historical data, and the training data of the LSTM-Transformer hybrid network includes at least 5000 groups of field measured plant height samples.

6. A method for measuring growth parameters of corn plants based on the system as claimed in claim 5, characterized in that: The following steps are involved: Step 1: Based on the ridge depth and width, the control system and the electric telescopic cylinder are used to adjust the inclination angle of the swing plate, and then the sampling path of the ground intelligent mobile platform is planned; Step 2: The ground intelligent mobile platform walks in the corn planting area according to the sampling path, and uses millimeter wave radar to obtain stalk profile data and calculate plant height and diameter; Step 3: Scan the corn ear using a micro CT probe installed on a ground intelligent mobile platform to generate a three-dimensional model and extract the parameters of ear length and ear diameter; Step 4: The edge computing unit integrates multi-source data and outputs visual reports and AR diagnostic suggestions.

7. The method according to claim 6, characterized in that The ground intelligent mobile platform described in step 2 obtains road information through a multi-spectral camera array and millimeter-wave radar to avoid obstacles. When the ground intelligent mobile platform moves to the designated sampling location, it stops moving and then uses the millimeter-wave radar to obtain the stem contour data and calculate the plant height and diameter.

8. The method according to claim 7, characterized in that The edge computing unit in step 4 adopts a lightweight YOLOv9s-Plant neural network model. The YOLOv9s-Plant model uses knowledge distillation technology to compress the number of parameters to 15% of the original model, and adopts the following data fusion rules. H final =ω1H radar +ω2H CT Among them, ω1=0.6, ω2=0.4, ω1 and ω2 correspond to the data weights of radar and CT respectively.

Images