Flat root box crop root growth and root character nondestructive monitoring analysis system

Abstract

The invention relates to the technical field of plant root system observation, in particular to a flat root box crop root system growth and root system character nondestructive monitoring and analysis system, which comprises the following modules: a flat root box module, which is used for providing a root system growth environment and comprises two parallel tempered glass plates and an adhesive tape, the tempered glass plates are arranged in the U-shaped stainless steel frame at intervals controlled by adhesive tapes, and the upper glass plate is of a three-section zigzag structure, so that a planting groove is formed in the top, and root systems are induced to grow close to the wall. According to the invention, by constructing a complete technical system composed of a flat root box module, an environment monitoring module, an image acquisition module, an image analysis module and a data management module, nondestructive monitoring and automatic analysis of phenotypic parameters in the root growth process are realized, so that the problems that destructive sampling is adopted in most traditional root character measurement methods, and the measurement accuracy is high are solved. The root system needs to be excavated and manually measured, so that the same plant cannot be continuously observed, and the root system growth dynamic data is missing.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to a non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box. Background Technology

[0002] In the fields of crop genetics and breeding and plant physiology research, roots, as key organs for water and nutrient absorption, are directly related to crop stress resistance and yield formation. However, due to the limitations of soil opacity, traditional root research methods mainly include destructive sampling techniques such as digging, core drilling, and root washing. Most traditional root trait measurement methods employ destructive sampling, which, due to the need for digging up roots and manual measurement, makes continuous observation of the same plant impossible, resulting in missing data on root growth dynamics and hindering the establishment of complete root morphology maps. Therefore, there is an urgent need to develop a technical solution that enables in-situ, continuous, and high-throughput monitoring and analysis of crop roots. Summary of the Invention

[0003] To overcome the above shortcomings, this invention provides a non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box. It aims to improve upon the problem that traditional root trait measurement methods mostly employ destructive sampling, which requires digging up the roots and manual measurement, making it impossible to continuously observe the same plant and resulting in a lack of dynamic root growth data.

[0004] In a first aspect, the present invention provides the following technical solution: a non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box, comprising the following modules: The flat root box module is used to provide a root growth environment. It includes two parallel tempered glass plates and adhesive strips. The spacing between the tempered glass plates is controlled by the adhesive strips and placed in a U-shaped stainless steel frame. The upper glass plate has a three-section zigzag structure to form a planting groove at the top, which induces the roots to grow against the wall. An environmental monitoring module, integrated on the flat root box module, is used to monitor soil environmental parameters in the root growth area in real time. It includes a soil sensor and a microcontroller, and the microcontroller is used to collect and transmit sensor data. The image acquisition module is used to acquire high-quality RGB images of the root system in the flat root box module. It includes a movable camera carriage, which is equipped with a camera fixing mechanism, an elastic mechanism for clamping and fixing the flat root box, and a light source system that provides uniform illumination. The image analysis module is communicatively connected to the image acquisition module and is used to automatically process and analyze the acquired root system images. It includes an image enhancement unit, a root system segmentation unit, and a phenotypic parameter calculation unit connected in sequence. The root system segmentation unit is implemented based on a lightweight fully convolutional neural network. The data management and interaction module is communicatively connected to the environmental monitoring module and the image analysis module. It is used to store soil environmental data, root system images and analysis results, and provides a web interaction interface for users to remotely submit analysis tasks, monitor progress in real time and visualize results.

[0005] By adopting the above technical solutions, a complete technical system consisting of five modules—root box, environmental monitoring, image acquisition, image analysis, and data management—was constructed, realizing non-destructive monitoring of root growth process and automatic analysis of phenotypic parameters. This improves upon the problem that traditional root trait measurement methods mostly employ destructive sampling, which requires digging up roots and manual measurement, making it impossible to continuously observe the same plant and resulting in missing dynamic data on root growth.

[0006] Preferably, the flat root box includes: A triangular planting frame constructed from L-shaped steel welded together is provided; Place the flat root box module inside the planting frame; The inclined structure of the triangular planting frame is used to keep the flat root box module at a 45° angle to the horizontal plane. The anti-slip grooves at the bottom of the planting frame increase the friction with the placement surface, preventing the flat root box from sliding.

[0007] Preferably, the environmental monitoring includes: The soil moisture sensor is inserted into the box through a hole drilled in the glass plate on the flat root box, and then sealed with a soft rubber strip. An Arduino development board is used as a microcontroller and connected to the soil moisture sensor. Configure the Arduino development board to collect sensor data at a fixed frequency of once every thirty minutes; The collected data can be stored locally via the SD module integrated into the Arduino development board, or sent to the data management and interaction module via its wireless communication module.

[0008] Preferably, the image acquisition includes: Before shooting, the camera car was completely wrapped with a light-blocking cloth to block out ambient light. Activate the LED light strips installed on both sides inside the camera car to provide uniform lighting for the shooting surface; The flat root box module is pushed into the root box placement area of ​​the photography vehicle and clamped and fixed by an elastic mechanism composed of springs. Under conditions of shading and uniform illumination, the camera was operated to acquire images and obtain RGB images of the root system.

[0009] Preferably, the image enhancement unit includes: The original root system RGB image is logarithmically transformed to decompose it into two additive components: illuminance and reflectance. Perform a Discrete Fourier Transform on the image after the above logarithmic transformation to convert it from the spatial domain to the frequency domain; In the frequency domain, a Gaussian filter is selected to filter the image; Perform an inverse Fourier transform on the filtered frequency domain image to convert it back to the spatial domain; The result of the inverse Fourier transform is subjected to an exponential operation to obtain the processed enhanced image.

[0010] Preferably, the root system segmentation unit includes: A compression path is constructed, and downsampling operations are performed through convolutional and pooling layers to extract multi-scale features of the image; Construct an extended path to gradually restore the spatial resolution of the image through upsampling operations and convolutional layers; By using a skip connection structure, feature maps extracted at different stages in the compressed path are passed to the corresponding stages in the extended path for feature fusion.

[0011] Preferably, the lightweight fully convolutional neural network implementation includes: An RGB three-channel root system image with a size of 512×512 pixels is input into the neural network; The neural network performs forward propagation calculations on the input image; The network outputs a binary segmented image of the same size as the input image. Pixels identified as roots are marked in white, and pixels identified as background are marked in black.

[0012] Preferably, the phenotypic parameter calculation unit includes: Single-pixel skeleton extraction is performed on the binary segmentation image of the root system. A 3×3 pixel window is used to traverse the image, and contour pixels that meet the conditions are removed through multiple iterations until only the skeleton pixels representing the central axis of the root system are retained. The total pixel length of the skeleton lines is counted using the eight-neighbor chain code tracking method for the extracted root skeleton. Calculate the root projection area and count the number of pixels with a value of 255 in the binary image; Calculate the pixel conversion factor CF, where CF = actual physical length of the scale in the image / pixel length occupied by the scale in the image; Multiply the obtained root length in pixels by the CF (Cross Flow) to get the physical root length; multiply the obtained number of pixels by the CF. 2 Multiply them to obtain the projected area of ​​the physical root.

[0013] Preferably, the data management and interaction include: Users upload root system images through a form on the front-end page; during the upload process, the form submit button on the front-end page is set to an unclickable state; The front-end uses Ajax technology to periodically send requests to the server back-end to check whether the image has been uploaded. When the back-end confirms that the image has been uploaded, the front-end receives the response and restores the form submit button to the clickable state. After a user submits an analysis task, the webpage redirects to the progress query page. On this page, the frontend uses Ajax to periodically send requests to the server to query the analysis progress of the task. The backend returns the current task progress information, and the frontend receives it and updates the progress display module on the webpage in real time. Once the task analysis is complete, the webpage automatically redirects to the results display page; this page queries the server for the final result data via Ajax, and the server returns the data in JSON format. The front end receives JSON data and renders and displays the analysis results in the taskbar, table bar, and image bar of the results page.

[0014] Secondly, the present invention provides the following technical solution: a non-destructive monitoring and analysis method for crop root growth and root traits using a flat-root box, comprising the following steps: S1. Provide a root growth environment, including two parallel tempered glass plates and adhesive strips. The spacing between the tempered glass plates is controlled by the adhesive strips and placed in a U-shaped stainless steel frame. The upper glass plate has a three-section zigzag structure to form a planting groove at the top, inducing the roots to grow against the wall. S2. Integrated into the flat root box module, used for real-time monitoring of soil environmental parameters in the root growth area, including a soil sensor and a microcontroller, wherein the microcontroller is used to collect sensor data and transmit it. S3. Obtain a high-quality RGB image of the root system in the flat root box module, which includes a movable camera carriage. The camera carriage is equipped with a camera fixing mechanism, an elastic mechanism for clamping and fixing the flat root box, and a light source system that provides uniform illumination. S4. The communication is connected to the image acquisition module for automated processing and analysis of the acquired root system images. It includes an image enhancement unit, a root system segmentation unit, and a phenotypic parameter calculation unit connected in sequence. The root system segmentation unit is implemented based on a lightweight fully convolutional neural network. S5. The communication connection is established with the environmental monitoring module and the image analysis module to store soil environmental data, root system images and analysis results, and to provide a Web interactive interface for users to remotely submit analysis tasks, monitor progress in real time and visualize results.

[0015] The present invention has the following beneficial effects: 1. In this invention, by constructing a complete technical system consisting of five major modules—a flat root box, environmental monitoring, image acquisition, image analysis, and data management—non-destructive monitoring of root growth processes and automatic analysis of phenotypic parameters are achieved. This improves upon the problem that traditional root trait measurement methods mostly employ destructive sampling, which requires digging up roots and manual measurement, making continuous observation of the same plant impossible and resulting in missing dynamic data on root growth.

[0016] 2. In this invention, the soil sensor and microcontroller integrated into the environmental monitoring module are used to collect and transmit real-time data on the root growth area, thereby realizing continuous monitoring of soil environmental parameters. This improves the problem that traditional environmental monitoring methods mostly use discrete sampling, which cannot obtain continuous environmental data due to long measurement intervals and disturbance to the root environment, thus making it difficult to analyze the correlation between root growth and environmental factors.

[0017] 3. In this invention, the movable photographing trolley of the image acquisition module, together with the elastic clamping mechanism and the uniform light source system, acquires root system images under standardized conditions, thereby ensuring the consistency of image quality. This improves the problem that traditional root system image acquisition mostly uses manual shooting, which cannot guarantee image comparability due to unstable lighting conditions and shooting angles, resulting in large errors in subsequent analysis.

[0018] 4. In this invention, remote task submission and result visualization are realized through the Web interaction interface provided by the data management and interaction module, thereby supporting multi-user collaborative work. This improves the problem that traditional data management mostly uses local storage, and due to the data being scattered and not uniformly formatted, it is impossible to achieve team collaborative analysis, resulting in low research efficiency. Attached Figure Description

[0019] Figure 1 This is a module architecture diagram of a non-destructive monitoring and analysis system for crop root growth and root traits using a flat root box, as proposed in this invention. Figure 2 This is a schematic diagram of the mechanical components of a non-destructive monitoring and analysis system for crop root growth and root traits using a flat root box, as proposed in this invention. Figure 3 This is a growth concept diagram of a non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box, as proposed in this invention. Figure 4 This is a diagram of the slope platform structure of a non-destructive monitoring and analysis system for crop root growth and root traits using a flat root box, as proposed in this invention. Figure 5 This is a wireframe diagram of the mechanical device of a non-destructive monitoring and analysis system for crop root growth and root traits using a flat root box, as proposed in this invention. Figure 6This is a root fluorescence labeling diagram of a non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box, as proposed in this invention. Figure 7 This is an experimental diagram of the root system of a crop root system non-destructive monitoring and analysis system for crop root growth and root traits using a flat root box, as proposed in this invention. Figure 8 This is a neural network architecture diagram of a non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box, as proposed in this invention. Figure 9 This invention presents a root growth statistical chart of a non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box. Figure 10 This is a flowchart of a method for non-destructive monitoring and analysis of crop root growth and root traits using a flat-root box, as proposed in this invention. Detailed Implementation

[0020] The technical solutions in 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.

[0021] Example 1: In a first embodiment of the present invention, the present invention provides a non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box, such as... Figures 1-9 As shown, it includes the following steps: The flat root box module is used to provide a root growth environment. It includes two parallel tempered glass plates and adhesive strips. The spacing between the tempered glass plates is controlled by the adhesive strips and placed in a U-shaped stainless steel frame. The upper glass plate has a three-section zigzag structure to form a planting groove at the top, which induces the roots to grow against the wall. Furthermore, the flat root box includes: A triangular planting frame constructed from L-shaped steel welded together is provided; Place the flat root box module inside the planting frame; By utilizing the sloping structure of the triangular planting frame, the flat root box module is kept at a 45° angle to the horizontal plane. The anti-slip grooves at the bottom of the planting frame increase the friction with the placement surface, preventing the flat root box from sliding.

[0022] Specifically, to enable root growth within a confined space while facilitating image acquisition, the flat root box module employs the following structural design. The module comprises two parallel tempered glass panels, with the spacing between them controlled by adhesive strips to create a sealed space. The entire glass panel is housed within a U-shaped stainless steel frame, with the U-shape providing support and fixation on the sides and bottom. The upper glass panel is designed with a three-section zigzag structure, naturally forming a planting groove at the top.

[0023] During planting, the seeds are placed in the groove. The zigzag structure creates a height difference in the vertical direction, physically restricting the volume of soil above the roots. As the roots grow downwards, they are constrained by the limited soil volume and guided by the boundaries of the glass plate, limiting their growth direction to a two-dimensional plane close to the glass plate.

[0024] To further control the root growth angle, the system is equipped with a dedicated planting frame. The planting frame is constructed from L-shaped steel welded into a triangular frame structure. After the flat root box module is placed within this frame, the hypotenuse of the triangle provides a 45-degree inclined support surface for the flat root box. Under the influence of gravity, the roots exhibit geotropic growth; the inclined placement allows the roots to adhere more closely to the inner surface of the glass plate below. The bottom of the planting frame is machined with anti-slip grooves, which generate frictional resistance when in contact with the placement platform, preventing displacement of the entire device during cultivation.

[0025] During the image analysis phase, the phenotypic parameter calculation unit performs pixel statistics based on the binarized image. The root projected area is obtained by counting the total number of white pixels and multiplying it by the conversion factor. The conversion factor CF is calculated as follows: CF equals the actual length of the ruler divided by the ruler pixel length. The actual length of the ruler refers to the physical length of a standard ruler placed within the shooting area, in centimeters; the ruler pixel length refers to the number of pixels occupied by the ruler in the digital image. This conversion factor converts the image pixel dimension into the actual physical size.

[0026] The physical structure of the flat-root box controls root growth morphology, the tilting device enhances adhesion to the wall, and standardized image acquisition provides qualified input for subsequent analysis. These structural features work together to ensure clear two-dimensional images of the root system, laying the foundation for automated phenotypic analysis.

[0027] The environmental monitoring module, integrated into the flat root box module, is used to monitor soil environmental parameters in the root growth area in real time. It includes a soil sensor and a microcontroller. The microcontroller is used to collect and transmit sensor data. Furthermore, environmental monitoring includes: The soil moisture sensor is inserted into the box through a hole drilled in the glass plate on the flat root box, and then sealed with a soft rubber strip. An Arduino development board is used as a microcontroller and connected to a soil moisture sensor. Configure the Arduino development board to collect sensor data at a fixed frequency of once every thirty minutes; The collected data can be stored locally via the SD module integrated into the Arduino development board, or sent to the data management and interaction module via its wireless communication module.

[0028] Specifically, the environmental monitoring module incorporates a soil moisture sensor implanted through a hole drilled in the upper glass plate. A soft rubber strip is used to seal the sensor, preventing soil leakage and maintaining a stable internal environment within the root box. The sensor probe is in direct contact with the soil inside the root box, measuring the soil volumetric water content. This volumetric water content is calculated using the sensor calibration formula. ;in This indicates the volumetric water content of the soil, expressed in cubic centimeters per cubic centimeter. The sensor sensitivity coefficient is expressed in units of volts. To calibrate the offset; This is the output voltage of the sensor.

[0029] The microcontroller uses an Arduino development board, connected to a soil moisture sensor via analog input pins. The board's internal program is programmed to perform a data acquisition task every thirty minutes, with the acquisition cycle controlled by a program delay function. The system reads the current... The value is calculated according to the conversion formula. Soil volumetric water content is obtained through a calibration formula. Data collection is accompanied by device serial numbers and timestamps to create a complete data record.

[0030] During the data acquisition process, the development board reads the analog voltage value output by the sensor and converts the voltage signal into a digital value through the built-in analog-to-digital converter. The conversion relationship is expressed by the following formula: ; in This represents the integer value output by the analog-to-digital converter, ranging from 0 to 1023. This indicates the actual output voltage of the sensor; This indicates that the development board's operating reference voltage is 5V.

[0031] Data storage is achieved in two ways: first, the collected data is written to an SD card module plugged into the development board and stored as a CSV file, with the file records including timestamps and corresponding sensor readings; second, the data is packaged and sent to the server interface of the data management and interaction module via the wireless communication module integrated into the development board. Wireless transmission uses the HTTP protocol, and the data is encapsulated in JSON format, including the device number, acquisition time, and measurement value.

[0032] This module enables continuous monitoring of soil moisture parameters, and the collected data is used to analyze the correlation between root growth and environmental parameters. Monitoring data corresponds to image acquisition timestamps, providing environmental background data for phenotypic analysis. The module's sealed design ensures the internal environment of the root box is free from external interference, fixed-frequency acquisition ensures the continuity of the data sequence, and a dual-mode storage mechanism improves data reliability.

[0033] The image acquisition module is used to acquire high-quality RGB images of the root system in the flat root box module. It includes a movable camera carriage, which is equipped with a camera fixing mechanism, an elastic mechanism for clamping and fixing the flat root box, and a light source system that provides uniform illumination. Furthermore, image acquisition includes: Before shooting, completely wrap the camera cart with a light-blocking cloth to block out ambient light. Activate the LED light strips installed on both sides inside the camera car to provide uniform lighting for the shooting surface; Push the flat root box module into the root box placement area of ​​the photography car and use the elastic mechanism made of springs to clamp and fix it. Under conditions of shading and uniform illumination, the camera was operated to acquire images and obtain RGB images of the root system.

[0034] Specifically, the image acquisition module achieves standardized acquisition of root system images through a movable camera cart. The camera cart uses an aluminum profile frame structure with four omnidirectional wheels at the bottom. The camera fixing mechanism uses an adjustable gimbal to accommodate different camera lenses. The elastic clamping mechanism consists of two sets of symmetrically arranged compression springs with a spring travel of 30-50 mm. The light source system consists of symmetrically distributed LED light strips on both sides, with two sets of light strips on each side, and the light strips are at a 45-degree angle to the shooting plane.

[0035] The blackout fabric is made of black Oxford cloth, completely wrapping the car frame to form a sealed dark chamber. The LED light strips have a color temperature of 5500K and a color rendering index greater than 90. Light source illuminance... Calculated by the following formula: ;in Illuminance of the illuminated surface, measured in lux; The intensity of an LED light source is expressed in candela. Indicates the angle of incidence of light; This indicates the distance from the light source to the shooting plane, in meters. Illumination uniformity reaches over 85%. During image acquisition, the flat box is pushed into the placement area, and the spring mechanism automatically applies clamping force. : ;in This indicates the spring constant, expressed in Newtons per meter (N / m). This indicates the spring compression, measured in meters. The clamping force ranges from 5 to 10 Newtons, ensuring the root box remains stable during exposure.

[0036] The camera imaging system satisfies the lens formula: ;in Indicates the lens focal length. Indicates object distance, The image distance is indicated in meters. Camera settings are: ISO 100, aperture f / 8, shutter speed 1 / 125 second.

[0037] The image sensor receives optical signals and converts them into digital signals. The RGB value of each pixel is determined by the following formula: ; ; ; in Indicates the spectral power distribution of the light source. , , This represents the camera's tri-color response function. Indicates the spectral reflectance of the subject. This represents the wavelength, and the integration range is the visible light band of 380-780 nanometers.

[0038] Images are saved in RGB three-channel JPEG format with a resolution of at least 20 megapixels. After acquisition, the images are transferred to the image analysis module via USB interface, with a transfer rate meeting the following requirements: ;in Indicates the image width in pixels. Indicates the image height in pixels. Indicates the number of bytes per pixel. Indicates transmission time. Indicates the transmission rate.

[0039] The mechanical structure ensures consistent geometric position for each shot, while the light-blocking design and uniform illumination eliminate ambient light interference, providing standardized input for subsequent image analysis.

[0040] The image analysis module is communicatively connected to the image acquisition module and is used to automatically process and analyze the acquired root system images. It includes an image enhancement unit, a root system segmentation unit, and a phenotypic parameter calculation unit connected in sequence. The root system segmentation unit is implemented based on a lightweight fully convolutional neural network. Furthermore, the image enhancement unit includes: The original root system RGB image is logarithmically transformed to decompose it into two additive components: illuminance and reflectance. Perform a Discrete Fourier Transform on the image after the above logarithmic transformation to convert it from the spatial domain to the frequency domain; In the frequency domain, a Gaussian filter is selected to filter the image; Perform an inverse Fourier transform on the filtered frequency domain image to convert it back to the spatial domain; The result of the inverse Fourier transform is subjected to an exponential operation to obtain the processed enhanced image.

[0041] Root system segmentation units include: A compression path is constructed, and downsampling operations are performed through convolutional and pooling layers to extract multi-scale features of the image; Construct an extended path to gradually restore the spatial resolution of the image through upsampling operations and convolutional layers; By using a skip connection structure, feature maps extracted at different stages in the compressed path are passed to the corresponding stages in the extended path for feature fusion.

[0042] Lightweight fully convolutional neural network implementations include: A 512×512 pixel RGB three-channel root system image is input into the neural network; The neural network performs forward propagation calculations on the input image; The network outputs a binary segmented image of the same size as the input image. Pixels identified as roots are marked in white, and pixels identified as background are marked in black.

[0043] The phenotypic parameter calculation unit includes: Single-pixel skeleton extraction is performed on the binary segmentation image of the root system. A 3×3 pixel window is used to traverse the image, and contour pixels that meet the conditions are removed through multiple iterations until only the skeleton pixels representing the central axis of the root system are retained. The total pixel length of the skeleton lines is counted using the eight-neighbor chain code tracking method for the extracted root skeleton. Calculate the root projection area and count the number of pixels with a value of 255 in the binary image; Calculate the pixel conversion factor CF, where CF = actual physical length of the scale in the image / pixel length occupied by the scale in the image; Multiply the obtained root length in pixels by the CF (Cross Flow) to get the physical root length; multiply the obtained number of pixels by the CF. 2 Multiply them to obtain the projected area of ​​the physical root.

[0044] Specifically, the image analysis module performs automated processing on the input root RGB image. The image enhancement unit uses a homomorphic filtering method, first processing the input image... Perform a logarithmic transformation: ; in Indicates the original image in coordinates The pixel value at that location. The image is decomposed into illumination components. and reflectivity Additive combinations: ; Perform a discrete Fourier transform on the logarithmic image: ; in Represents a frequency domain image. Image dimensions. Frequency domain filtering is performed using a Gaussian high-pass filter: ;in Represents the filter transfer function. Representing frequency point Distance to the center point The cutoff frequency, and This is the gain coefficient. It is a constant.

[0045] Perform an inverse Fourier transform on the filtered result and take the exponent operation to obtain the enhanced image.

[0046] The root segmentation unit employs a fully convolutional network with an encoder-decoder structure. The encoder extracts features through convolution and max pooling, with each convolutional kernel having a size of 3×3 and ReLU activation. The decoder performs upsampling through transposed convolution to restore spatial resolution. Skip connections concatenate the feature maps from each stage of the encoder with the corresponding layers of the decoder.

[0047] The network input is a 512×512×3 RGB image, and the output is a 512×512×1 binary segmentation map. During forward propagation, the feature map size is halved layer by layer in the encoder, and the number of channels is doubled; in the decoder, the size is doubled layer by layer, and the number of channels is halved.

[0048] The phenotypic parameter calculation unit calculates binary images. Morphological processing is performed. Skeleton extraction employs an iterative thinning algorithm, using 3×3 structuring elements to traverse the image and remove boundary pixels that meet the deletion criteria. The thinning process continues until no more pixels can be removed, resulting in a skeleton image with a single pixel width. .

[0049] Root length calculation employs the eight-neighbor chain code tracing method. Starting from the skeleton endpoint, the skeleton path is traced along eight directions using chain code, and the total number of chain code steps is counted. The root projected area is obtained by counting the total number of white pixels: ; Physical dimension conversion is based on a scale reference. The conversion factor is calculated as follows: ;in Indicates the actual length of the scale. Indicates the ruler pixel length. Final root length. Root projection area .

[0050] This module enables automated calculation of phenotypic parameters from raw images, providing quantitative data support for root morphology analysis.

[0051] The data management and interaction module is connected to the environmental monitoring module and the image analysis module. It is used to store soil environmental data, root images and analysis results, and provides a web interaction interface for users to remotely submit analysis tasks, monitor progress in real time and visualize results. Furthermore, data management and interaction include: Users upload root system images through a form on the front-end page; during the upload process, the form submit button on the front-end page is set to an unclickable state; The front-end uses Ajax technology to periodically send requests to the server back-end to check whether the image has been uploaded. When the back-end confirms that the image has been uploaded, the front-end receives the response and restores the form submit button to the clickable state. After a user submits an analysis task, the webpage redirects to the progress query page. On this page, the frontend uses Ajax to periodically send requests to the server to query the analysis progress of the task. The backend returns the current task progress information, and the frontend receives it and updates the progress display module on the webpage in real time. Once the task analysis is complete, the webpage automatically redirects to the results display page; this page queries the server for the final result data via Ajax, and the server returns the data in JSON format. The front end receives JSON data and renders and displays the analysis results in the taskbar, table bar, and image bar of the results page.

[0052] Specifically, the data management and interaction module is built on a B / S architecture and uses a MySQL database to store three types of data: soil environmental data, root system image files, and analysis result data. The database design includes three main tables: user, task, and result, with relationships established through user_id and task_id.

[0053] System storage capacity requirements are calculated using the following formula: ;in Indicates the total storage capacity requirement. Indicates the first The size of each image file, Indicates the first The size of the JSON data for each analysis result. This indicates the total number of tasks managed by the system.

[0054] When a user submits a root image via a web form, the front-end performs the following steps: sets the disabled attribute of the form submission button to true, and uploads the file using an XMLHttpRequest object in multipart / form-data encoding format. The file upload time is estimated using the following formula: ; in This indicates the actual upload time. Indicates the size of the image file to be uploaded. Indicates network upload bandwidth. This represents the server's disk write speed. During the upload process, the frontend initiates a timed query mechanism, with the polling interval dynamically adjusted according to the following formula: ; in Indicates the actual polling interval. Set to 2 seconds. To adjust the coefficient, This indicates the task has been running for a certain period of time. The frontend sends query requests to the / api / upload_status interface at this interval. When the backend returns the status field value as complete, the disabled attribute of the submit button is restored to false.

[0055] After a task is submitted, the system generates a 32-bit task identifier (task_id) and creates a new record with a status field of "pending" in the task table. The database connection pool size is configured according to the following formula: ; in Indicates the actual number of connections. Indicates the maximum number of connections. As a growth factor, For system uptime, This is the reference time parameter.

[0056] The front-end page redirects to the progress query interface, which makes a timed Ajax request to the ` / api / task_progress` API. System throughput is evaluated using the following formula: ;in This indicates the number of requests processed per second. This indicates the total number of requests within the statistical period. Indicates the duration of the statistical period.

[0057] After the task is completed, the database record status is updated to "completed". The front-end automatically redirects to the results display page, retrieving the analysis results in JSON format via Ajax. The data export function uses streaming processing, and the export time meets the following requirements: ;in This indicates the time taken for the export operation. This indicates the total amount of data to be exported. This indicates the disk I / O read / write speed.

[0058] After parsing the JSON data, the client uses the DataTables plugin to render the tabular data, draws statistical charts using the Canvas API, and decodes and displays Base64-encoded image data. All operation records are written to the system log to ensure traceability of the operation process.

[0059] Example 2: In a wheat variety drought resistance screening experiment, researchers needed to continuously observe the root growth dynamics of 200 different varieties during a 21-day stress period, and screen superior germplasm by analyzing root architecture parameters. Traditional root research methods have the following limitations: each observation requires destructive sampling, making continuous tracking of the same plant impossible; manual measurements take 3-5 working days and are subject to significant subjective errors; there is a lack of real-time soil moisture monitoring methods in root growth areas, making it difficult to establish the correlation between root response and environmental stress; manually captured root images suffer from uneven lighting and angle shifts, severely affecting image quality; manual analysis of a single image takes 15-20 minutes, resulting in low efficiency for large-scale population analysis; experimental data is stored in a scattered manner, hindering collaboration among researchers and making data tracing and analysis difficult. These factors severely restrict the efficiency and accuracy of large-scale root phenotypic studies. To address these problems, this invention provides a non-destructive monitoring and analysis method for crop root growth and root traits using a flat-root box, the structure of which is as follows... Figure 10 As shown. The specific implementation process of this method is as follows: The growth space is formed by two parallel tempered glass plates, with the spacing between the glass plates controlled by adhesive strips to create a sealed structure. The entire structure is fixed within a U-shaped stainless steel frame. The upper glass plate adopts a three-section zigzag design, forming a planting groove at the top. This structure physically restricts root growth between the glass plates, creating a two-dimensional root distribution suitable for observation.

[0060] The soil moisture sensor is implanted inside the root box through pre-drilled holes in the upper glass plate and sealed using a soft rubber strip. The microcontroller collects sensor data at fixed time intervals. The collected data is stored locally or wirelessly transmitted to a data processing center to record parameters such as soil volumetric water content.

[0061] The mobile photography trolley is equipped with a camera fixing mechanism, an elastic clamping mechanism, and a lighting system. A light-blocking cloth is used to isolate ambient light during image acquisition, ensuring consistent shooting conditions. Camera parameters are fixed to specific values ​​to guarantee that the acquired root system images have a uniform quality standard.

[0062] The image enhancement unit uses a specific algorithm to process images and improve image quality. The root segmentation unit uses a fully convolutional neural network with a specific structure to convert the input RGB image into a binary segmentation result. The phenotypic parameter calculation unit extracts the root skeleton using a specific method, calculates parameters such as root length, and combines them with an actual size conversion factor to calculate the physical length.

[0063] The system is built around a database for storage and utilizes server-side and client-side technologies for task management. Users upload images and query processing status via a web interface, and the system returns analysis results for data visualization on the front-end interface. Data is transferred between each step through standard interfaces, forming a complete processing flow.

[0064] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box, characterized in that, Includes the following modules: The flat root box module is used to provide a root growth environment. It includes two parallel tempered glass plates and adhesive strips. The spacing between the tempered glass plates is controlled by the adhesive strips and placed in a U-shaped stainless steel frame. The upper glass plate has a three-section zigzag structure to form a planting groove at the top, which induces the roots to grow against the wall. An environmental monitoring module, integrated on the flat root box module, is used to monitor soil environmental parameters in the root growth area in real time. It includes a soil sensor and a microcontroller, and the microcontroller is used to collect and transmit sensor data. The image acquisition module is used to acquire high-quality RGB images of the root system in the flat root box module. It includes a movable camera carriage, which is equipped with a camera fixing mechanism, an elastic mechanism for clamping and fixing the flat root box, and a light source system that provides uniform illumination. The image analysis module is communicatively connected to the image acquisition module and is used to automatically process and analyze the acquired root system images. It includes an image enhancement unit, a root system segmentation unit, and a phenotypic parameter calculation unit connected in sequence. The root system segmentation unit is implemented based on a lightweight fully convolutional neural network. The data management and interaction module is communicatively connected to the environmental monitoring module and the image analysis module. It is used to store soil environmental data, root system images and analysis results, and provides a web interaction interface for users to remotely submit analysis tasks, monitor progress in real time and visualize results.

2. The non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box according to claim 1, characterized in that, The flat root box includes: A triangular planting frame constructed from L-shaped steel welded together is provided; Place the flat root box module inside the planting frame; The inclined structure of the triangular planting frame is used to keep the flat root box module at a 45° angle to the horizontal plane. The anti-slip grooves at the bottom of the planting frame increase the friction with the placement surface, preventing the flat root box from sliding.

3. The non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box according to claim 1, characterized in that, The environmental monitoring includes: The soil moisture sensor is inserted into the box through a hole drilled in the glass plate on the flat root box, and then sealed with a soft rubber strip. An Arduino development board is used as a microcontroller and connected to the soil moisture sensor. Configure the Arduino development board to collect sensor data at a fixed frequency of once every thirty minutes; The collected data can be stored locally via the SD module integrated into the Arduino development board, or sent to the data management and interaction module via its wireless communication module.

4. The non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box according to claim 1, characterized in that, The image acquisition includes: Before shooting, the camera car was completely wrapped with a light-blocking cloth to block out ambient light. Activate the LED light strips installed on both sides inside the camera car to provide uniform lighting for the shooting surface; The flat root box module is pushed into the root box placement area of ​​the photography vehicle and clamped and fixed by an elastic mechanism composed of springs. Under conditions of shading and uniform illumination, the camera was operated to acquire images and obtain RGB images of the root system.

5. The non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box according to claim 1, characterized in that, The image enhancement unit includes: The original root system RGB image is logarithmically transformed to decompose it into two additive components: illuminance and reflectance. Perform a Discrete Fourier Transform on the image after the above logarithmic transformation to convert it from the spatial domain to the frequency domain; In the frequency domain, a Gaussian filter is selected to filter the image; Perform an inverse Fourier transform on the filtered frequency domain image to convert it back to the spatial domain; The result of the inverse Fourier transform is subjected to an exponential operation to obtain the processed enhanced image.

6. The non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box according to claim 1, characterized in that, The root system segmentation unit includes: A compression path is constructed, and downsampling operations are performed through convolutional and pooling layers to extract multi-scale features of the image; Construct an extended path to gradually restore the spatial resolution of the image through upsampling operations and convolutional layers; By using a skip connection structure, feature maps extracted at different stages in the compressed path are passed to the corresponding stages in the extended path for feature fusion.

7. The non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box according to claim 1, characterized in that, The lightweight fully convolutional neural network implementation includes: An RGB three-channel root system image with a size of 512×512 pixels is input into the neural network; The neural network performs forward propagation calculations on the input image; The network outputs a binary segmented image of the same size as the input image. Pixels identified as roots are marked in white, and pixels identified as background are marked in black.

8. The non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box according to claim 1, characterized in that, The phenotypic parameter calculation unit includes: Single-pixel skeleton extraction is performed on the binary segmentation image of the root system. A 3×3 pixel window is used to traverse the image, and contour pixels that meet the conditions are removed through multiple iterations until only the skeleton pixels representing the central axis of the root system are retained. The total pixel length of the skeleton lines is counted using the eight-neighbor chain code tracking method for the extracted root skeleton. Calculate the root projection area and count the number of pixels with a value of 255 in the binary image; Calculate the pixel conversion factor CF, where CF = actual physical length of the scale in the image / pixel length occupied by the scale in the image; Multiply the obtained root length in pixels by the CF (Cross Flow) to get the physical root length; multiply the obtained number of pixels by the CF. 2 Multiply them to obtain the projected area of ​​the physical root.

9. The non-destructive monitoring and analysis system for crop root growth and root traits using a flat-root box according to claim 1, characterized in that, The data management and interaction include: Users upload root system images through a form on the front-end page; during the upload process, the form submit button on the front-end page is set to an unclickable state; The front-end uses Ajax technology to periodically send requests to the server back-end to check whether the image has been uploaded. When the back-end confirms that the image has been uploaded, the front-end receives the response and restores the form submit button to the clickable state. After a user submits an analysis task, the webpage redirects to the progress query page. On this page, the frontend uses Ajax to periodically send requests to the server to query the analysis progress of the task. The backend returns the current task progress information, and the frontend receives it and updates the progress display module on the webpage in real time. Once the task analysis is complete, the webpage automatically redirects to the results display page; this page queries the server for the final result data via Ajax, and the server returns the data in JSON format. The front end receives JSON data and renders and displays the analysis results in the taskbar, table bar, and image bar of the results page.

10. A non-destructive monitoring and analysis method for crop root growth and root traits using a flat-root box, characterized in that, A non-destructive monitoring and analysis system for crop root growth and root traits using a flat root box as described in any one of claims 1-9 includes the following steps: S1. Provide a root growth environment, including two parallel tempered glass plates and adhesive strips. The spacing between the tempered glass plates is controlled by the adhesive strips and placed in a U-shaped stainless steel frame. The upper glass plate has a three-section zigzag structure to form a planting groove at the top, inducing the roots to grow against the wall. S2. Integrated into the flat root box module, used for real-time monitoring of soil environmental parameters in the root growth area, including a soil sensor and a microcontroller, wherein the microcontroller is used to collect sensor data and transmit it. S3. Obtain a high-quality RGB image of the root system in the flat root box module, which includes a movable camera carriage. The camera carriage is equipped with a camera fixing mechanism, an elastic mechanism for clamping and fixing the flat root box, and a light source system that provides uniform illumination. S4. The communication is connected to the image acquisition module for automated processing and analysis of the acquired root system images. It includes an image enhancement unit, a root system segmentation unit, and a phenotypic parameter calculation unit connected in sequence. The root system segmentation unit is implemented based on a lightweight fully convolutional neural network. S5. The communication connection is established with the environmental monitoring module and the image analysis module to store soil environmental data, root system images and analysis results, and to provide a Web interactive interface for users to remotely submit analysis tasks, monitor progress in real time and visualize results.

Images