A simple RGB-D image recognition device and method for automatically capturing rapeseed phenotypes

By using a simple RGB-D image recognition device and an automated control system, the problems of low efficiency and high cost in rapeseed phenotypic surveys have been solved, and high-precision, non-destructive, and automated collection of rapeseed phenotypic parameters has been achieved.

CN123118535APending Publication Date: 2026-11-17江西省 中国科学院庐山植物园
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611331167.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-31
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

The current rapeseed planting density is low and over-planting leads to shade avoidance response. Traditional manual phenotypic surveys are inefficient, costly, and difficult to guarantee accuracy, while automated ingestion platforms are expensive and complex to operate.

Method used

Design a simple RGB-D image recognition device, including a shooting box, a visible light source module, a stepper motor drive system, a multi-vision camera module, a plant ID recognition system, and a Raspberry Pi core control system, to achieve automated acquisition of rapeseed phenotypic characteristics. High-precision image data is obtained by using a stepper motor to drive rotation and an RGB-D camera to capture images, combined with Mask QC quality control.

Benefits of technology

It enables low-cost, non-destructive, and automated rapeseed phenotypic data collection, reducing manual labor costs, improving collection efficiency, and ensuring data repeatability and accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118535A_ABST
    Figure CN123118535A_ABST
Patent Text Reader

Abstract

This invention provides a simple RGB-D image recognition device and method for automatically capturing rapeseed phenotypic traits, relating to the fields of agricultural informatization and plant phenomics. The device includes an aluminum alloy casing with internal light-absorbing materials, a uniform visible light source, a stepper motor drive system, and a multi-view camera module containing two RGB-D cameras (side-view and top-view). A Raspberry Pi centrally controls barcode scanning and recognition, while the stepper motor rotates in stages and the dual cameras capture images synchronously. Color correction is first performed using a Colorchecker 24-color chart. Then, side and top views of each plant from multiple angles are captured and stored after Mask QC quality control. Finally, color, shape, and texture features are extracted in batches to obtain phenotypic parameters such as plant height, number of leaves, petiole length, leaf area index, canopy area, and projected area. This invention is low-cost, portable, simple to operate, and enables non-destructive, high-throughput, and automated acquisition of rapeseed shade-tolerant / dense-planting-tolerant phenotypic traits, lowering the barrier to entry for phenotyping technology.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of agricultural informatization and plant phenomics, specifically to a simple RGB-D image recognition device and method for automatically acquiring rapeseed phenotypes. Background Technology

[0002] Rapeseed is an important oilseed crop in my country, and its yield is directly related to the country's food and oil supply security. Increasing rapeseed planting density and optimizing plant structure are effective ways to increase rapeseed yield per unit area. At present, the planting density of rapeseed planting areas in my country is relatively low. However, over-planting will lead to different canopy layers of rapeseed, which will change the light quality received by the plants and easily cause shade avoidance response, resulting in changes in plant structure. In severe cases, it will limit the increase of rapeseed yield. How to weaken the rapeseed shade avoidance response, select suitable germplasm and cultivate new varieties with reasonable plant structure to adapt to the high-density planting environment has become the key to solving the problem.

[0003] With the improvement of the level of intelligence in modern agriculture, automated phenomics platforms are emerging. These platforms are equipped with various sensor detection devices and image recognition technologies, and are widely used to detect the growth phenotypes or stress resistance phenotypes of various crops in fields or greenhouses. Current investigations into the plant architecture phenotype of rapeseed also rely heavily on precise phenotypic identification. Compared to traditional phenotypic investigation methods that primarily rely on manual measurement, there are still technical difficulties and shortcomings. 1. Most shade-tolerant / dense-planting-tolerant phenotypic traits are complex quantitative traits regulated by multiple genes, such as plant height, number of branches, leaf morphology, and canopy structure, involving many plant type morphology-related indicators. Detecting complex quantitative phenotypic traits through manual surveys has disadvantages such as high labor costs, low efficiency, and destructive and damaging nature. 2. Traditional manual phenotyping surveys are difficult to guarantee in terms of measurement accuracy. They rely on subjective judgment, which can easily lead to measurement errors. The data is sometimes not repeatable, data verification is difficult, and the evaluation criteria for manual survey data vary from person to person. 3. In addition, existing greenhouse or field crop phenotypic batch collection devices adopt the form of gantry frames or drones and are equipped with commercial hyperspectral, thermal infrared imaging units and lidar units and other multi-sensor detection equipment. The construction price is relatively expensive, the relative cost is extremely high, the operation procedure is relatively complicated, the detection equipment occupies a lot of space, and is greatly affected by the size of the greenhouse.

[0004] To address the issues of high cost, high price, and cumbersome setup of automated acquisition platforms, as well as the high labor costs and low efficiency of traditional manual phenotyping, this invention proposes a low-cost, easy-to-operate, non-destructive, automated, and batch acquisition device for rapeseed RGB-D images and extraction of phenotypic parameters, in order to meet the needs of high-precision and automated phenotyping analysis in greenhouse environments. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a simple RGB-D image recognition device and method for automatically acquiring rapeseed phenotypic data, solving the problems of high cost, expensive price, cumbersome automated acquisition platform construction, and high labor cost and low efficiency of traditional manual phenotypic data acquisition.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a simple RGB-D image recognition device and method for automatically capturing rapeseed phenotypic characteristics, comprising a shooting box, a visible light source module, a stepper motor drive system, a multi-vision camera module, a plant ID recognition system, and a Raspberry Pi core control system, wherein: The camera box is a hollow box made of aluminum alloy profiles, with sponge and black velvet inside to block external light interference. There is a door in front for placing the plant to be tested. The visible light source module adopts a strip LED layout and is set on the top of the shooting box to provide a uniform and stable illumination light source. The stepper motor drive system includes a stepper motor power supply, a stepper motor driver, a stepper motor body, and a rotary table. The stepper motor body supports and drives the rotary table to carry the plant under test and rotate it automatically from 0 to 360°. The multi-view camera module includes a side-view RGB-D camera and a top-view RGB-D camera. Both cameras are mounted on a movable and retractable bracket and are used to acquire side and top views of the plant under test, respectively. The plant ID identification system includes a barcode bound to the plant to be tested and a barcode scanner, used to scan and identify the plant number. The Raspberry Pi core control system includes a Raspberry Pi main control module and its power supply. It is connected to the two RGB-D cameras, the barcode scanner, and the external storage module via a USB hub, and is also connected to the stepper motor drive system interface. The Raspberry Pi main control module is configured to: respond to the plant ID obtained by scanning and the input command, automatically start the stepper motor drive system to drive the rotary table to rotate at a specific angle, and simultaneously trigger the two RGB-D cameras to complete automatic shooting, and store the acquired RGB-D images to the external storage module. The rotary table rotates once every 40 steps, for a total of 10 rotations over 400 steps, so that each plant under test can obtain 10 side views and 1 top view from different angles. The Raspberry Pi main control module is also configured to start the image preview window and the Mask QC quality control dynamic window at the same time as shooting. When the dynamic result displays "pass" or "review", the image quality is deemed to be qualified. The Mask QC image and the original image are saved together to the external storage module.

[0007] Preferably, the visible light source module includes six strip LED tubes, which are evenly arranged on the top of the shooting box to emit white light with a color temperature of 6000K; a matte diffuse reflector is laid on the top of the box below the LED tubes to eliminate shadows and highlights on the surface of the plant leaves to be tested.

[0008] Preferably, the side-view RGB-D camera is fixed on a vertically arranged movable telescopic bracket, the bottom of which is connected to a horizontal moving platform, so that the camera is located on the side of the plant to be tested; the top-view RGB-D camera is fixed on a movable telescopic horizontal bracket, so that the camera is located on the top of the plant to be tested, the bottom of which is fixed to a vertical moving base platform; the telescopic bracket is composed of two stainless steel round rods connected by rotatable threaded plastic sleeves, and is embedded in the movable track by welding it to the aluminum plate of the moving platform, so as to realize the adjustment of the camera's distance and height.

[0009] Preferably, the stepper motor driver is a DM556 stepper motor driver, which uses an eight-bit DIP switch to set the dynamic current and microstepping accuracy; the system also has an optocoupler isolation module to isolate motor drive interference; the stepper motor operating current is 2.7A, the standby current is set to half-current mode, the microstepping accuracy is set to 2, and the pulse count is 400 / revolution.

[0010] Preferably, the side view is used to extract leaf morphology, number of branches, petiole length, plant height and color texture features, and the top view is used to obtain canopy area and projected area; the naming format of the 10 side views is "germ number_biological duplicate number_S_serial number", and the naming format of the 1 top view is "germ number_biological duplicate number_T_1", where S represents side view and T represents top view.

[0011] Preferably, the image Mask QC quality control process is as follows: while shooting, the image preview window and the Mask QC dynamic window are launched on the computer terminal. The target plant area and the background are accurately separated by the mask. When the dynamic result displays "pass" or "review", the image quality is judged to be qualified. The Mask QC image and the original image are saved together to the external storage module.

[0012] Preferably, the Mask QC quality control process is as follows: the target plant area is accurately separated from the background by masking, providing a reliable image and data foundation for subsequent phenotypic feature extraction; when the dynamic result displays "pass" or "review", it indicates that the image quality is feasible and can be used for subsequent image segmentation to extract the target object.

[0013] Preferably, the size of the shooting box can be adjusted according to the height of the crop to be tested. An exemplary size is 2.5m long × 1m wide × 1.5m high, which can be placed in a greenhouse, shed or cultivation room; the Raspberry Pi main control module is equipped with a Micro-HDMI interface to connect to a computer monitor, and the external storage module is a USB flash drive or mobile hard drive.

[0014] Preferably, the two RGB-D cameras are Intel RealSense depth cameras with an imaging resolution of no less than 1080P and support for automatic exposure and automatic white balance.

[0015] A preferred automated ingestion method includes the following steps: S1 Material and Equipment Preparation: Place the shooting box in a greenhouse or shed environment, adjust the position and angle of the side-view RGB-D camera and the top-view RGB-D camera so that the rapeseed plant to be tested is completely within the field of view of the two cameras, and complete the communication test and set the focal length and pixel parameters. S2 RGB-D image color correction: Using the standard Colorchecker 24-color test card, color, white balance, and exposure parameters are corrected at both side and top view shooting positions to achieve color standardization of plant images; S3 collects rapeseed phenotypes: Scan the barcode of the plant to be tested to obtain the plant ID and input the command. After the Raspberry Pi main control module responds, it automatically starts the stepper motor to drive the rotary table to rotate the plant step by step at a specific angle, and simultaneously triggers two RGB-D cameras to automatically take pictures, obtaining side views and top views of the same plant from multiple angles. After the images are controlled by Mask QC, they are stored in the external storage module. S4 Phenotypic Image Processing: Uploads images from the external storage module to the computer terminal, batch extracts color, shape and texture features of rapeseed phenotypic images before and after processing, and obtains at least one phenotypic parameter among plant height, number of leaves, petiole length, leaf area index, canopy area and projected area.

[0016] This invention provides a simple RGB-D image recognition device and method for automatically capturing rapeseed phenotypic characteristics. It offers the following advantages: This simple RGB-D image recognition device and method is used to automatically acquire rapeseed phenotypic data. It uses a Raspberry Pi to control a plant ID scanning system to drive a stepper motor to rotate the plant at a specific angle, thereby collecting crop phenotypic data. The device is portable, adaptable to non-destructive and automated phenotypic data acquisition of rapeseed or other plants in greenhouses or communities, and can lower the threshold for using phenotyping technology.

[0017] This simple RGB-D image recognition device and method is an automated means of capturing rapeseed phenotypes. Through Colorchecker color card correction and Mask QC image quality control, it effectively eliminates light color difference and accurately separates the target plant area from the background, ensuring image color standardization and data repeatability, and reducing subjective measurement errors.

[0018] This simple RGB-D image recognition device and method is an automated acquisition device for rapeseed phenotypic data. It uses a stepper motor to drive a disc to rotate at a fixed number of steps (once every 40 steps, for a total of 400 steps 10 times) to stably obtain a standardized image sequence of 10 side views and 1 top view for each plant. This facilitates the subsequent batch extraction of phenotypic parameters such as plant height, number of leaves, petiole length, leaf area index, canopy area, and projected area, achieving rapid, automatic, and non-destructive data acquisition. Attached Figure Description

[0019] Figure 1 This is a simplified schematic diagram of the automated phenotypic acquisition device of the present invention; Figure 2 This is a flowchart of the automated phenotypic acquisition process of the present invention.

[0020] The components include: 10. Camera housing; 11. Door; 20. Visible light source module; 30. Stepper motor drive system; 31. Stepper motor power supply; 32. Stepper motor driver; 33. Rotary disc stage; 34. Stepper motor body; 35. Optical coupler isolation module; 40. Multi-view camera module; 41. Side-view RGB-D camera; 42. Top-view RGB-D camera; 43. Vertical telescopic bracket; 44. Horizontal telescopic bracket; 45. Horizontal moving platform; 46. Vertical moving base platform; 50. Plant ID recognition system; 60. Raspberry Pi core control system; 61. Power supply; 62. Raspberry Pi main control module; 70. Plant under test; 80. Computer terminal. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example

[0022] like Figure 1 - Figure 2This invention provides a simple RGB phenotyping device, primarily comprising a shooting box 10, a visible light source module 20, a stepper motor drive system 30, a multi-vision camera module 40, a plant ID recognition system 50, a Raspberry Pi core control system 60, and a computer terminal 80. The shooting box 10 is constructed with an aluminum alloy frame, and six strip LED tubes are fixed to the top of the box. A matte diffuser plate is placed under the strip LEDs to eliminate shadows and highlights on the surface of rapeseed seedling leaves, ensuring uniform illumination of the plant surface. The shooting box can be placed in greenhouses, sheds, cultivation rooms, or other facilities where phenotyping is required. To accommodate different types of crops, such as rapeseed plants at different stages, corn, and rice, the shooting box can be automatically adjusted in size according to the crop height. In this example, the shooting box dimensions are 2.5m × 1m × 1.5m (length × width × height).

[0023] like Figure 1 - Figure 2 As shown, in this invention, both the top-view camera RGB-D and the side-view camera RGB-D are mounted on movable and retractable brackets. The side-view depth camera RGB-D41 is fixed to a vertically placed movable and retractable bracket 43, the bottom of which is connected to a horizontal moving platform 45. This camera is located on the side of the rapeseed. The top-view depth camera RGB-D42 is fixed to a movable horizontal retractable bracket 44, located on top of the rapeseed. The bottom of this bracket is fixed to a vertical moving base platform 46. The movable brackets 43 and 42 are retractable stainless steel rods, consisting of two stainless steel round rods connected by a rotatable threaded plastic sleeve. Tightening the threads allows the two retractable brackets to extend and retract. The bottom of the retractable rod is welded and fixed to an aluminum plate on the horizontal 43 or vertical moving base platform 46. The aluminum plate is movable and its tightness is adjusted by specific screws. The aluminum plate is embedded in the movable track of the horizontal 43 or vertical moving base platform 46 to allow the retractable rod to move closer or further and higher. Both cameras are fixed to the retractable rods using fixing knobs. The telescopic support and movable base platform are freely adjustable to be suitable for phenotypic collection of different types of crops.

[0024] like Figure 1 - Figure 2As shown, the phenotypic acquisition device of this invention is uniformly controlled by a Raspberry Pi core control system 60, which controls a stepper motor drive system 30, two depth cameras 41 / 42, a plant ID recognition system 50, and an external storage module. The Raspberry Pi core control system 60 is equipped with a Micro-HDMI interface to connect to a computer monitor 80. The Raspberry Pi main control module 62 stores pre-programmed commands. Specific instructions are input from the computer terminal to complete image acquisition. This invention sends plant ID commands through the plant ID scanning system. The Raspberry Pi core control system 60 responds to the image acquisition control command, thereby automatically starting the stepper motor body 34, the rotary table 33, and the two camera systems 41 and 42 to complete the shooting. The images are synchronously stored in the external storage module. The background below the rotary table is removed using an RGB-D depth map. The leaf mask is extracted by combining the Lab space green threshold. The mask occupies the viewport ratio ∈ [0.05, 0.95] and the maximum connected component area ratio is >60% to pass; otherwise, it is reviewed.

[0025] In this invention, the scanner is a USB barcode scanner. After scanning the barcode, the plant ID is obtained. The keyboard input enters to send a command to the Raspberry Pi and synchronously drives the stepper motor drive system 30 to work, so that the motor drives the rotating disk 33 to rotate, thereby causing the plant to rotate at a specific angle, thus realizing the collection of information of the plant under test 70. In this invention, the stepper motor body 34 includes a supporting shaft, which supports automatic rotation from 0-360°. The rapeseed potted plant is placed on the rotating disk 33, and different angles are captured by taking pictures at a specific number of steps. The stepper motor driver 32 uses an eight-bit DIP switch to set the dynamic current and subdivision accuracy to match different specifications of stepper motors, preventing the circuit from overcurrent and overheating and burning out the stepper motor and main control power supply. The optocoupler isolation module 35 prevents interference from the motor drive and improves the stability of the device. The stepper motor of this invention has an operating current of 2.7A, a standby current set to half-current mode, a subdivision accuracy set to 2, and a pulse count of 400 / revolution. A stepper motor drives the rotary table 33 to rotate once every 40 steps, for a total of 400 steps, or 10 times in total. This results in 10 side-view images and 1 top-view image per rapeseed plant. The top-view camera 42 acquires one top-view image, mainly including characteristics such as canopy area and projected area under the acquisition conditions. The side-view camera 41 acquires 10 side-view images from different angles, mainly capturing plant morphological characteristics such as the number of branches, petiole length, and plant height, as well as color and texture features. The camera supports automatic imaging with a resolution ≥1080P. After acquiring images of rapeseed plants under different shading treatments from different side views, the images are stored via USB flash drive or external hard drive.

[0026] This invention uses a simple RGB acquisition device to collect information from different side and top views of the same rapeseed seedling. It extracts features of shape, texture, and color from the phenotypic images to obtain rapeseed phenotypic parameters under different shading treatments, such as plant height, number of leaves, petiole length, and leaf area index. This invention enables batch image processing, thus achieving rapid, automatic, and non-destructive acquisition of crop phenotypic parameters.

[0027] Material Preparation: Place the imaging box 10 in the greenhouse / tunnel or other facility environment where data collection is needed. Position the portable RGB-D image acquisition system inside the imaging box near the door 11. Adjust the length of the vertical telescopic bracket 43 and the height of the horizontal telescopic bracket 44. Correspondingly adjust the positions and acquisition angles of the two depth cameras fixed on the vertical and horizontal telescopic brackets 43 and 44 on the platform to ensure that plants such as rapeseed are within the field of view of the side-view RGB-D camera 41 and the top-view RGB-D camera 42. At this time, the external USB connectors of the two cameras are plugged into the Ugreen power supply multi-hub diversity unit, and the power is off. Turn on the power, test the terminal control communication of the phenotypic acquisition device computer terminal 80 to ensure it is normal, and fine-tune the field of view of the side-view and top-view cameras again. The computer terminal 80 inputs commands to control two depth cameras 41 and 42. The user terminal computer terminal 80 starts the GUI graphical interface, i.e. the preview function, to observe whether the complete rapeseed plant to be measured appears in the GUI dynamic image window. If necessary, the cameras 41 and 42 are adjusted to a suitable angle and fixed to determine the camera focal length, pixels, image length and width and other parameters to ensure accurate acquisition.

[0028] RGB-D Image Color Correction: Before acquiring plant phenotypic images, color charts are used to correct the colors of side-view and top-view images. A standard Colorchecker (24-color) test chart is used to perform RGB-D color correction on the images acquired by the camera module. The color chart can be set relative to the side-view RGB-D camera 41 and the top-view RGB-D camera 42. When testing the side-view camera 41, the color chart is placed vertically in the shooting area; when testing the top-view camera 42, the color chart is placed horizontally in the shooting area. After inputting instructions on the computer terminal 80, a preview window is launched, displaying the (24-color) test chart frame. Based on the instructions, the color correction software automatically identifies the 24 color blocks and adjusts the ROI size in the center area of ​​each color block. The computer terminal 80 obtains the corrected parameters, including color, white balance, and exposure parameters, thereby obtaining the optimal measured RGB-D parameters under specific lighting conditions. Color correction can minimize color differences caused by illumination, achieving color standardization of plant images and thus better extracting plant color features.

[0029] Rapeseed phenotypic data collection: After initializing the hardware and adjusting the parameters, the stepper motor power supply 31 is turned on to control the start of the stepper motor rotation system. Rapeseed seedlings with 5 leaves are used as the T0 stage for photography. They are shaded with translucent plastic bags. After 30 days, the germplasm has three biological replicates (3 seedlings). Rapeseed seedlings under normal light and those under shade (T1 stage) are placed on the rotating disc 33 respectively. The barcode scanner in the plant ID recognition system 50 identifies the plant barcode to obtain the plant ID. The ID format for unshaded plants is Bna001_01, Bna001_02, and Bna001_03; the ID format for shaded plants is Bna001_04, Bna001_05, and Bna001_06. After pressing Enter, the stepper motor system turntable is started simultaneously by sending a command to the computer terminal 80 on the Raspberry Pi main control module 62, which in turn turns on the two RGB-D cameras 41 and 42, completing initialization and triggering the cameras to take pictures. Pause for 10 seconds, then trigger and complete the photo taking within 5 seconds. The stepper motor takes a total of 400 steps, capturing 10 side-view images and 1 top-view image for each rapeseed seedling. The 10 side-view images generated for each seedling are named Bna001_1_S_1,...,Bna001_1_S_10; the 1 top-view image is named Bna001_1_T_1, where S represents side view and T represents top view. All phenotypic images are simultaneously stored on an external USB flash drive or portable hard drive via a Raspberry Pi module. At the same time, the Raspberry Pi main control module 62 sends a command to the computer terminal 80 to start the image preview window, inputs a command to present the image Mask QC quality control dynamic window, and synchronizes the dynamic results of the rapeseed image Mask QC. The mask described in this invention accurately separates the target plant area from the background, providing a reliable image and data foundation for subsequent phenotypic feature extraction. The rapeseed image dynamic Mask QC presentation process displays words such as "pass" or "review," indicating that the image quality is feasible and can be used for subsequent image segmentation and target object extraction. The Mask QC image and the image are saved to an external USB flash drive.

[0030] Phenotypic Image Processing: After collecting phenotypic data of rapeseed before and after shading treatment, the power is turned off. The data is uploaded to a computer terminal via an external USB flash drive or portable hard drive. Software is used to batch extract phenotypic parameters of the rapeseed before and after treatment, namely color, shape, and texture features, to obtain phenotypic parameters such as plant height, number of leaves, and leaf area index. This enables rapid phenotypic imaging of crops in greenhouses / tunnels or other facility environments, and rapid, non-destructive acquisition of phenotypic parameters.

[0031] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A simple RGB-D image recognition device for automatically capturing rapeseed phenotypic characteristics, comprising a shooting box (10), a visible light source module (20), a stepper motor drive system (30), a multi-vision camera module (40), a plant ID recognition system (50), and a Raspberry Pi core control system (60), characterized in that: The inside of the shooting box (10) is lined with sponge and black velvet to block external light interference, and the front of the shooting box (10) is hinged with a door (11). The visible light source module (20) adopts a strip LED layout and is set on the top of the shooting box (10) to provide a uniform and stable illumination light source; The stepper motor drive system (30) includes a stepper motor power supply (31), a stepper motor driver (32), a stepper motor body (34), and a rotary table (33). The stepper motor body (34) supports and drives the rotary table (33) to carry the plant under test and rotate automatically from 0 to 360°. The multi-view camera module (40) includes a side-view RGB-D camera (41) and a top-view RGB-D camera (42). Both cameras are mounted on a movable and retractable bracket and are used to collect side views and top views of the plant under test, respectively. The plant ID identification system (50) includes a barcode bound to the plant to be tested and a barcode scanner for scanning and identifying the plant number. The Raspberry Pi core control system (60) includes a Raspberry Pi main control module (62) and its power supply (61). It is connected to a side-view RGB-D camera (41), a top-view RGB-D camera (42), a barcode scanner and an external storage module via a USB hub, and is also connected to the stepper motor drive system (30) interface. The Raspberry Pi main control module (62) is configured to: respond to the plant ID obtained by scanning and input commands, automatically start the stepper motor drive system (30) to drive the rotary table (33) to rotate at a specific angle, and simultaneously trigger the two RGB-D cameras (41, 42) to complete automatic shooting, and store the collected RGB-D images to the external storage module.

2. The simplified RGB-D image recognition device for automatically capturing rapeseed phenotypic characteristics according to claim 1, characterized in that: The visible light source module (20) includes 6 strip LED tubes, which are evenly arranged on the top of the shooting box (10) with white light of 6000K color temperature. A matte diffuse reflector is laid on the top of the box below the LED tubes.

3. The simplified RGB-D image recognition device for automatically capturing rapeseed phenotypic characteristics according to claim 1, characterized in that: The side-view RGB-D camera (41) is fixed on a vertically arranged vertical telescopic bracket (43), the bottom of which is connected to a horizontal moving platform (45). The top-view RGB-D camera (42) is fixed on a movable horizontal telescopic bracket (44), the bottom of which is fixed to a vertical moving base platform (46). The telescopic bracket is composed of two stainless steel round rods connected by a rotatable threaded plastic sleeve, and is embedded in the movable track by welding it to the aluminum plate of the moving platform.

4. The simplified RGB-D image recognition device for automatically capturing rapeseed phenotypic characteristics according to claim 1, characterized in that: The stepper motor driver (32) system is also equipped with an optocoupler isolation module (35) for isolating motor drive interference.

5. The simplified RGB-D image recognition device for automatically capturing rapeseed phenotypic characteristics according to claim 1, characterized in that: The Raspberry Pi main control module (62) is equipped with a Micro-HDMI interface to connect to a computer monitor (80), and the external storage module is a USB flash drive or a portable hard drive.

6. The simplified RGB-D image recognition device for automatically capturing rapeseed phenotypic characteristics according to claim 1, characterized in that: Both the side-view RGB-D camera (41) and the top-view RGB-D camera (42) are Intel RealSense depth cameras with an imaging resolution of no less than 1080P.

7. The method for automatically capturing rapeseed phenotypes using a simple RGB-D image recognition device according to claim 1, applied to the apparatus of any one of claims 1-6, comprises the following steps: S1 Material and equipment preparation: Place the shooting box (10) in a greenhouse or shed environment, adjust the position and angle of the side-view RGB-D camera (41) and the top-view RGB-D camera (42) so that the rapeseed plant to be tested is completely within the field of view of the two cameras, and complete the communication test and set the focal length and pixel parameters. S2 RGB-D image color correction: Using the standard Colorchecker 24-color test card, color, white balance, and exposure parameters are corrected at both side and top view shooting positions to achieve color standardization of plant images; S3 Collect rapeseed phenotype: Scan the barcode of the plant to be tested to obtain the plant ID and input the command. After the Raspberry Pi main control module (62) responds, it automatically starts the stepper motor to drive the rotary table (33) to rotate the plant step by step at a specific angle, and simultaneously triggers two RGB-D cameras (41, 42) to automatically take pictures, obtain side views and top views of the same plant from multiple angles, and store the images in the external storage module after passing the Mask QC quality control. S4 Phenotypic Image Processing: Uploads images from the external storage module to the computer terminal, batch extracts color, shape and texture features of rapeseed phenotypic images before and after processing, and obtains at least one phenotypic parameter among plant height, number of leaves, petiole length, leaf area index, canopy area and projected area.