Cotton root phenotype detection and analysis system and method

Abstract

The invention discloses a cotton root system phenotype detection and analysis system and method. The method comprises the following steps: acquiring a cotton root system image by using a root system image acquisition device; processing the cotton root system image by using the trained root system semantic segmentation model to obtain a cotton root system feature map; making a scale card graph; detecting the cotton root feature map by using a rhizome parameter quantitative detection model based on an XOR residual structure and a scale chart; calculating cotton root system phenotype parameters according to the detection result, and calculating dynamic evolution parameters of the cotton root system phenotype according to the sampling time sequence; establishing a cotton root growth dynamic evolution model based on the root phenotype; and detecting and analyzing the new cotton root system image by using the cotton root system growth dynamic evolution model, and predicting the growth state of the cotton plant. The intelligent estimation of the root system target growth parameters based on the learning model under the conditions of precise segmentation of the cotton root system under the complex background and no manual intervention is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of crop root morphology detection, and in particular to a cotton root phenotype detection and analysis system and method. Background Art

[0002] Cotton is a deep-rooted crop, and its roots are the primary organ for water and nutrient absorption and transport. Scientific monitoring and management of root growth directly impacts cotton plant growth and pest resistance, boll yield and quality, and reproductive breeding. During its peak root growth phase, the taproot grows up to 2.5 cm daily, reaching depths exceeding 1 meter, and lateral roots can extend up to 50 cm laterally. Deep tillage during this phase effectively controls nutrient growth, promotes root development, and prolongs the root's functional life. During its peak root absorption phase, the root network is essentially established, absorption capacity reaches peak levels, but rooting capacity declines. Strengthening water and fertilizer management during this phase can maintain root vitality and delay root aging. Therefore, dynamic monitoring and analysis of cotton plant root growth holds important research and application implications for cotton growth management and breeding.

[0003] Currently, research techniques and methods for studying plant root systems remain at the forefront and a hot topic in related fields both domestically and internationally. However, traditional research methods, such as excavation, drilling, and root cutting, can cause varying degrees of damage to plants. Aeroponics, hydroponics, and paper-based culture methods are suitable for laboratory environments and are difficult to apply to cotton fields. Meanwhile, advanced technologies are being applied to non-destructive root research. For example, Keyes et al. used synchrotron X-ray tomography (SRXCT) in 2017 to achieve three-dimensional root images accurate to 1 μm, and Pflugfelder et al. used magnetic resonance imaging (MRI) to accurately model soybean roots. However, these methods share common challenges: expensive and complex testing equipment, and high soil moisture content can lead to large errors and even hinder imaging. Micro-root canal analysis, combining a small color camera with a 360° multi-layer rotating scanner, can capture dynamic changes in root morphology. However, this technique is limited by equipment installation and sampling time, which can affect root growth paths and image clarity.

[0004] Furthermore, the background in cotton root images varies during growth. Given the slenderness, small size, and high background noise of root objects, commonly used image semantic segmentation models such as Mask R-CNN and DeepLab struggle to achieve the required segmentation accuracy. To address this, Dr. He Kaiming integrated the ResNet residual architecture with an FPN feature detection network, constructing a Mask R-CNN model using a Faster RCNN backbone and a deconvolutional mask branch. This model achieves pixel-level semantic segmentation of image objects, achieving an average precision (AP) of approximately 38%. With the emergence and continuous improvements of the one-stage detection model YOLO, the YOLO v8 model currently achieves an average precision (AP) of over 50% for semantic segmentation of image objects. However, the datasets used for these detection models are mostly large objects in everyday scenes, where the proportion of interior pixels is higher than that of outline pixels. However, the characteristics of crop root images are the opposite—the proportion of outline pixels is higher and the proportion of interior pixels is relatively lower. Therefore, simply applying these classic models to transfer learning of cotton root objects is difficult. Xie Chenxi's team at Beihang University and Pengcheng Laboratory proposed a one-stage Pyramid Grafting Network (PGNet) for salient object detection (SOD). This model extracts features from images of different resolutions and performs feature grafting. The attention matrix generated is supervised by an attention-guided loss (AGL), helping the network better interact with attention from different models. The model achieved promising results in detecting the contours of steel structures and feathers. Chen Guowei et al. proposed a high-precision natural image matting network, PPMatting, which consists of a semantic context branch (SCB) and a high-resolution detail branch (HRDB). The interaction between the two branches improves the network's ability to perceive detail, outperforming other models in extracting the root details of a person's hair. Su Jiayi et al. proposed a semantic segmentation network for grape roots based on a modified U-Net architecture. Yan Jingkun et al. added a global attention mechanism module to the OCRNet network to strengthen the focus on root targets, achieving good results in the semantic segmentation of micro-root canal images.

[0005] The detection of root targets mainly analyzes the growth of the root system, including the length and diameter of the main root, the number, length and diameter of lateral roots and root hairs, the amount of new root growth and the amount of root ablation, in order to predict the growth status and nutritional status of the plant. However, there are currently no mature high-throughput detection technologies and methods for root phenotypes. The development of high-precision acquisition and quantitative analysis methods and models for root phenotypes is a hot topic and difficulty in research at home and abroad. Combining high-throughput imaging technology and deep image semantic segmentation models to quantitatively detect and analyze the growth status of cotton roots in soil is at the forefront of research in this field. This study provides technical and methodological support for exploring the laws of root changes during cotton growth and analyzing the correlation between root changes and plant growth. It is of great significance to the growth management and cotton breeding of Xinjiang cotton, and will help to improve the scientific and technological level of Xinjiang cotton industry, and help to improve the level of agricultural science and technology and agricultural modernization in Xinjiang. Summary of the Invention

[0006] In view of the shortcomings of the prior art, the present invention aims to provide a method for detecting and analyzing cotton root phenotypes.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] In a first aspect, the present invention provides a method for detecting and analyzing cotton root phenotypes, the key of which is that it comprises the following steps:

[0009] Step 1: Acquire a cotton root system image using a root system image acquisition device;

[0010] Step 2: Use the trained root semantic segmentation model to process the cotton root image to obtain a cotton root feature map;

[0011] Step 3: preparing a scale card for quantitatively detecting the structure and distribution of cotton root systems;

[0012] Step 4: Detect the cotton root system characteristic map using the root and stem parameter quantitative detection model based on the XOR residual structure and the ruler card map;

[0013] Step 5: Calculate cotton root phenotypic parameters based on the test results, and calculate the dynamic evolution parameters of the cotton root phenotype based on the sampling time series;

[0014] Step 6: Based on the cotton root phenotypic parameters, dynamic evolution parameters and cotton growth conditions, a cotton root growth dynamic evolution model based on root phenotype is established;

[0015] Step 7: Use the cotton root growth dynamic evolution model to detect and analyze the new cotton root system image to predict the growth status of the cotton plant.

[0016] Furthermore, in step 2, the cotton root system image is processed using the trained root system semantic segmentation model to obtain the cotton root system feature map. The specific steps are as follows:

[0017] Step 2.1: Preprocess the obtained cotton root images, annotate the image objects, and construct training and test sets;

[0018] Step 2.2: Establish a root semantic segmentation model based on the attention mechanism;

[0019] Step 2.3: Use the training set to train the root semantic segmentation model;

[0020] Step 2.4: Use the test set to test the trained root semantic segmentation model to obtain a trained root semantic segmentation model;

[0021] Step 2.5: Use the trained root semantic segmentation model to process the segmented root image to obtain the cotton root feature map.

[0022] Furthermore, the root semantic segmentation model includes a first feature extraction module, a second feature extraction module, an attention module, a feature detection module, and a feature fusion module, wherein the first feature extraction module is used to extract multi-scale regional features of the main root from the cotton root system image; the second feature extraction module is used to extract multi-scale regional features of lateral roots and root fibers from the cotton root system image; the attention module is used to use the attention mechanism to process the bottom-level feature map in the multi-scale feature map output by the first feature extraction module and the second feature extraction module; the feature detection module is used to perform feature detection on the regional features of different scales output by the first feature extraction module, the second feature extraction module, and the attention module; the feature fusion module is used to splice and fuse the feature detection module and the output regional features through multi-scale feature cross-grafting to obtain a cotton root system feature map.

[0023] Furthermore, the attention module includes a first feature fusion unit, a first activation function unit, a first convolution unit, a second activation function unit, a first multiplication processing unit, a second multiplication processing unit, a second convolution unit, a second feature fusion unit, a third activation function unit and a third feature fusion unit;

[0024] The first feature fusion unit adds the bottom-level feature map output by the first feature extraction module and the bottom-level feature map output by the second feature extraction module, and then inputs the sum into the first activation function unit. The first activation function unit strengthens the root features in the feature map and then inputs the sum into the first convolution unit. The first convolution unit performs convolution processing on the feature map and then inputs the sum into the second activation function unit. The second activation function unit suppresses the background response in the feature map and then inputs the sum into the first multiplication processing unit. The first multiplication processing unit multiplies the sum with the bottom-level feature map output by the first feature extraction module and then outputs the sum.

[0025] The second multiplication processing unit multiplies the lowest-level feature map output by the first feature extraction module and the lowest-level feature map output by the second feature extraction module, and then inputs the result into the second convolution unit. The second convolution unit convolves the feature map and then inputs the result into the second feature fusion unit. The second feature fusion unit adds the result and the lowest-level feature map output by the second feature extraction module, and then inputs the result into the third activation function unit. The third activation function unit outputs the result after strengthening the root features in the feature map.

[0026] The third feature fusion unit performs addition processing on the feature map output by the first multiplication processing unit and the feature map output by the third activation function unit.

[0027] Furthermore, the first feature extraction module adopts a transformer network, the second feature extraction module adopts a CNN network, and the feature detection module adopts an FPN network.

[0028] Furthermore, the ruler card diagram includes a reference card diagram and a slope detection set diagram. The reference card diagram is composed of horizontal and vertical ruler line segments, which are used to detect the length of the rhizome in the horizontal and vertical directions, and at the same time to determine the diameter of the rhizome to be detected; the slope detection set diagram is a set of unit line segments with a certain slope and a fixed interval, which have the same width as the unit line segments of the reference card diagram, and are used to detect the length of the rhizome in different extension directions.

[0029] Furthermore, the specific steps of detecting the cotton root system characteristic map using the root and stem parameter quantitative detection model based on the XOR residual structure and the ruler card map in step 4 are as follows:

[0030] Step A1: Detecting the rhizome diameter using a reference card, performing XOR residual processing using a rhizome parameter quantitative detection model, and retaining the reference card with two standard units as the selected detection scale for the corresponding diameter rhizome;

[0031] Step A2: selecting a card diagram series from the slope detection set for detection according to the selected detection scale;

[0032] Step A3: performing full slope detection on the possible extension direction of the rhizome according to the selected card series;

[0033] Step A4: splicing the standard units generated by the full slope detection, and counting the number of standard line segments to approximate the rhizome length.

[0034] Furthermore, the specific steps of detecting the cotton root system characteristic map using the root and stem parameter quantitative detection model based on the XOR residual structure and the ruler card map in step 4 are as follows:

[0035] Step B1: Detect the rhizome diameter using a reference card image, and perform XOR residual processing using the XOR residual module in the rhizome parameter quantitative detection model to obtain the extended length of the rhizome in each direction, and retain the reference card image larger than 1 standard unit as the selected detection scale of the corresponding diameter rhizome;

[0036] Step B2: Select a card diagram series from the slope detection set for detection based on the selected detection scale;

[0037] Step B3: using the feature fusion module in the rhizome parameter quantitative detection model to merge the features of each slope direction;

[0038] Step B4: applying the small target detection module in the rhizome parameter quantitative detection model to perform detection and counting, and counting the number of standard line segments to fit the rhizome length;

[0039] Step B5: training the adjustment network parameters and card diagram settings of the rhizome parameter quantitative detection model, optimizing the detection results, and obtaining the final detection results.

[0040] Furthermore, the specific steps of establishing the cotton root growth dynamic evolution model based on root phenotype in step 6 are as follows:

[0041] Step 6.1, based on the phenotypic parameters of the cotton root system, studying the root phenotypic growth characteristics of the control group;

[0042] Step 6.2: Based on the dynamic evolution parameters of cotton plants, study and analyze the impact of planting factors on cotton growth conditions;

[0043] Step 6.3: Combine the cotton root phenotypic growth characteristics and the impact of planting factors on cotton growth conditions to establish a cotton growth inversion model based on root phenotype.

[0044] In a second aspect, the present invention provides a cotton root phenotype detection and analysis system for implementing the method, comprising:

[0045] A root system image acquisition module is used to acquire cotton root system images using a root system image acquisition device;

[0046] A semantic segmentation module is used to process cotton root images using a trained root semantic segmentation model to obtain a cotton root feature map;

[0047] A scale card image generation module is used to create scale cards for quantitatively detecting cotton root structure and distribution;

[0048] The rhizome length detection module is used to detect the cotton root system characteristic map using the rhizome parameter quantitative detection model based on the XOR residual structure and the ruler card diagram;

[0049] An evolution parameter calculation module is used to calculate cotton root phenotypic parameters based on the detection results, and to calculate the dynamic evolution parameters of the cotton root phenotype based on the sampling time series;

[0050] An evolution model building module is used to build a cotton root growth dynamic evolution model based on root phenotype according to cotton root phenotypic parameters, dynamic evolution parameters and cotton growth conditions;

[0051] The detection and analysis module is used to detect and analyze new cotton root system images using the dynamic evolution model of cotton root system growth and predict the growth status of cotton plants.

[0052] The remarkable effects of the present invention are:

[0053] 1. The present invention establishes a root semantic segmentation model with higher precision and finer granularity. Not only does it not require artificially designed features for guidance, the model itself can learn the required features from the training data and make reasonable use of them. Therefore, it has better generalization ability and can perform stably even when facing complex scenes. At the same time, by introducing the attention mechanism, the model's sensitivity to small root features is improved, and accurate segmentation of cotton roots under complex backgrounds is achieved. By adding an attention module, the model focuses on more important features and suppresses unnecessary features. The model not only performs better, but also has stronger robustness to noisy input. Experiments have shown that the visual self-attention mechanism can not only optimize the segmentation boundaries, but also reduce the phenomena of missed segmentation and mis-segmentation.

[0054] 2. The attention module improves the network's feature discrimination and selection capabilities, thereby alleviating the problem of reduced accuracy caused by interference features. It can accurately segment cotton root images without losing too much detail information, significantly improving the accuracy of the final root image segmentation results, and providing reliable reference information for the subsequent calculation and analysis of root phenotypic characteristics. Moreover, the adopted attention structure can focus on more important channel features, reducing network complexity.

[0055] 3. The present invention provides a quantitative detection method for cotton root images based on a ruler card diagram, which converts the continuous parameter detection of image target size into a quantitative calculation of discrete targets based on standard units, and realizes the intelligent estimation of root target growth parameters based on a learning model without human intervention.

[0056] 4. This invention is based on deep learning technology and applies comprehensive technologies such as metric scale card images, XOR residual structure, pixel detection and small target counting to realize the analysis and processing of root images, which can serve as technical and data support for cotton crop breeding and improvement research. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a flow chart of the method of the present invention;

[0058] Figure 2 is a structural diagram of a root system image acquisition device;

[0059] Figure 3 It is a structural diagram of the introduction type inclined tube;

[0060] Figure 4 This is the left view of the imported inclined tube;

[0061] Figure 5 This is a schematic diagram of the working state of the imported inclined tube;

[0062] Figure 6 is a flowchart of semantic segmentation of cotton root images;

[0063] Figure 7 This is a schematic diagram of cotton root system image data processing;

[0064] Figure 8 is a network structure diagram of the root semantic segmentation model;

[0065] Figure 9 is a block diagram of the attention module;

[0066] Figure 10 This is the cotton root system image obtained by semantic segmentation;

[0067] Figure 11 It is a schematic diagram of the ruler card diagram;

[0068] Figure 12 This is a structural diagram of an implementation method of the quantitative detection model of rhizome parameters;

[0069] Figure 13 This is a structural diagram of another implementation of the rhizome parameter quantitative detection model;

[0070] Figure 14 It is a principle block diagram of the system of the present invention. DETAILED DESCRIPTION

[0071] The specific implementation manner and working principle of the present invention will be further described in detail below with reference to the accompanying drawings.

[0072] Example 1:

[0073] like Figure 1 As shown, the embodiment of the present invention provides a cotton root phenotype detection and analysis method, the specific steps are as follows:

[0074] Step 1: Acquire a cotton root system image using a root system image acquisition device;

[0075] This embodiment studies the distribution of cotton roots in the soil during cotton growth in Xinjiang, and combines the research results and professional experience of a team of agricultural experts from Shihezi University and a team of experts in field robot design and development from the University of Lincoln in the United Kingdom to design an image acquisition device that can be used for long-term, all-round measurement of cotton roots. The specific structure is as follows: Figure 2 As shown, the root image acquisition device includes an import-type inclined tube, a control module, an image collector, a cleaning module and an image processing module based on edge computing; the control module is used to generate control instructions; the import-type inclined tube is used to collect cotton root systems; the image collector uses a high-resolution satellite camera to obtain cotton root system images according to control instructions. Since the cotton root system can reach 1m in length and 50cm in side diameter during the vigorous period, the depth and angle of the image acquisition point of the image collector are controlled by the control unit; the cleaning module is used to clean the import-type inclined tube according to the control instructions to ensure the long-term use of the equipment and reduce the image noise introduced by dust particles attached to the tube wall; in order to reduce the load on the system central server, the image processing module is used to process and upload the cotton root system image according to the control instructions using edge computing.

[0076] This embodiment is aimed at the sandy soil environment and irrigation operation of Xinjiang cotton fields. In order to reduce the analysis error caused by misaligned shooting, a claw-type positioning structure is provided on the inlet inclined tube, such as Figure 3-Figure 5 As shown, the claw-type positioning structure includes an oblique tube body 1, a mounting ring 2, a mounting groove 3, a fixing claw 5, a connecting rod 7 and a fixing block 8. The mounting ring 2 slides on the end of the oblique tube body 1 close to the ground surface, and a plurality of mounting grooves 3 are fixed on the outer wall of the mounting ring 2. The fixing claws 5 and the mounting grooves 3 are arranged in a one-to-one correspondence. One end of each fixing claw 5 is arranged in the mounting groove 3 and is rotatably connected to the mounting groove 3 through a rotating shaft. The middle part of the fixing claw 5 and one end of the connecting rod 7 are rotatably connected through a first pin shaft, and the other end of the connecting rod 7 and the fixing block 8 are rotatably connected through a second pin shaft. The fixing block 8 is fixed to the outer wall of the oblique tube body 1.

[0077] Based on the above structure, it can be known that in the process of moving the mounting ring 2 along the inclined tube body 1 to move it toward the fixed block 8, the fixing claw 5 will rotate around the rotating shaft under the action of the connecting rod 7, and at the same time, the connecting rod 7 will also rotate around the second pin shaft driven by the fixing claw 5, so that the multiple fixing claws 5 are opened away from the end of the mounting groove 3 and away from the outer wall of the inclined tube body 1, so that when the imported inclined tube is used to collect root system images, the inclined tube body 1 can be fixed in the soil, so that its position remains relatively fixed, thereby reducing the analysis error caused by misplaced shooting.

[0078] Furthermore, the plurality of fixing grooves 3 are evenly distributed circumferentially on the outer wall of the mounting ring 2 , so that the inclined tube body 1 can be fixed in position in all directions by the fixing claws 5 , thereby improving convenience in use.

[0079] from Figure 3 and Figure 5 It can also be seen that a pin 4 is independently installed on the mounting ring 2, and a plurality of insertion holes 6 are opened on the outer wall of the inclined tube body 1. Through the cooperation of the pin 4 and the insertion holes 6, the position of the moving mounting ring 2 can be fixed, thereby improving the stability of the position and posture of the fixing claw 5 during root system image acquisition, further improving the position stability of the imported inclined tube, and further reducing the analysis error caused by misplaced shooting.

[0080] Before entering step 2, the present embodiment is to guarantee the validity and reliability of analysis data, and is planted in field soil strictly according to standard and is no less than 5 mu of experimental cotton plants.According to research demand, build cotton planting mode reference group: topping vs no topping reference group, high-quality seed vs common seed reference group, nitrogen fertilizer application amount reference group (root growth is mainly affected by nitrogen fertilizer, and phosphorus and potash fertilizer factors are lower), bud stage root breaking vs continuous rooting reference group, boll opening period foliage fertilizer and water supply reference group etc., the sampling point of each reference group is no less than 10.Develop standard sampling process, sampling time node, sampling quantity.Record environmental parameters such as temperature, precipitation, irrigation amount, fertilizer type and consumption simultaneously, record the growth height of sampling cotton plant, branch amount, boll forming time, boll opening time, boll size, boll quality etc. as the explicit control factor of root growth.

[0081] Step 2: Use the trained root semantic segmentation model to process the cotton root image to obtain a cotton root feature map;

[0082] like Figure 6 As shown in the figure, in this example, the trained root semantic segmentation model is used to process the cotton root image. The specific steps to obtain the cotton root feature map are as follows:

[0083] Step 2.1: Preprocess the obtained cotton root images, annotate the image objects, and construct training and test sets;

[0084] In the embodiment of the present invention, about 50,000 cotton root system images were collected as a training data set for the model. The cotton root system images were subjected to edge cropping, median filtering, data cleaning and size normalization. The size of the source image was determined by the setting of the built-in lens of the micro-root tube, such as Figure 7 The original image shown in (a) has a resolution of 2550*2273. Considering the computational complexity and training speed of the system, we set the resolution of the standard processed image to 640*640 and divide the original image into 4*4 sub-images, as shown in the following example: Figure 7 (b) (edges are padded with 0 pixels). Use the Labelme annotation tool to annotate the image objects, as shown in Figure 7 (c). Save the annotation parameters into the json file, such as Figure 7 (d).

[0085] Step 2.2: Establish a root semantic segmentation model based on the attention mechanism;

[0086] The cotton root target images are different from the daily target detection images. They are collected by the high-precision lens built into the positioning oblique root canal buried in the cotton field soil. The original collected image data is as follows Figure 7 shown.

[0087] from Figure 7 It can be seen that cotton root images have complex and noisy backgrounds, slender and widely extended targets, variable and irregular root shapes, multiple contour features, varying sizes of main roots, lateral roots, and root fibers, and a small number of pixels occupied in the diameter direction. Common semantic segmentation detection models such as Mask R-CNN and DeepLab are unable to meet the detection accuracy requirements for root systems. The Transformer-based Segmenter model, the PGNet (Pyramid Grafting Network) model, and the HRDB (High Resolution Detail Branch)-based PP-Matting model have all achieved good evaluation scores in the field of high-precision segmentation of detail images. PGNet-UH (trained on the UH dataset) achieved a MAE of 0.020 and an F-score of 0.945. PP-Matting achieved an MSE of 0.005 when trained on the high-precision images of Composition-1k and 0.009 when trained on the low-precision Distinctions-646 dataset. However, a common problem is that they rely on experimental datasets and have limited model generalization capabilities. For image targets such as cotton roots, which are basically composed of edges and branch outlines, the processing effect is not ideal.

[0088] In response to the above features, this embodiment establishes a high-precision small target segmentation model suitable for high-resolution cotton root images - PCSegRoot. This model uses a transformer network and a CNN network to extract regional features of the main root, lateral roots, and root hairs respectively. It uses an FPN structure to detect features of regional targets of different scales, splices and fuses the features of each branch through multi-scale feature cross-grafting, and uses the AGL mechanism to improve the model's sensitivity to small features, forming a pyramid cross-model root semantic segmentation model based on the attention mechanism. Specifically:

[0089] like Figure 8 As shown, the root semantic segmentation model in this example includes a first feature extraction module, a second feature extraction module, an attention module, a feature detection module, and a feature fusion module, wherein the first feature extraction module is used to extract multi-scale regional features of the main root from the cotton root system image; the second feature extraction module is used to extract multi-scale regional features of lateral roots and root hairs from the cotton root system image; the attention module is used to use the attention mechanism to process the bottom-level feature map in the multi-scale feature map output by the first feature extraction module and the second feature extraction module; the feature detection module is used to perform feature detection on the regional features of different scales output by the first feature extraction module, the second feature extraction module, and the attention module; the feature fusion module is used to splice and fuse the feature detection module and the output regional features by cross-grafting of multi-scale features to obtain a semantic segmentation result;

[0090] The root semantic segmentation model based on the attention mechanism established in the present invention not only does not require artificially designed features for guidance, but the model itself can learn the required features from the training data and make reasonable use of them. Therefore, it has better generalization ability and can perform stably even when facing complex scenes. Moreover, by introducing the attention mechanism, the model's sensitivity to fine root features is improved, and accurate segmentation of cotton roots under complex backgrounds is achieved.

[0091] In this embodiment of the present invention, the structure of the attention module is as follows: Figure 9 As shown, it includes a first feature fusion unit, a first activation function unit, a first convolution unit, a second activation function unit, a first multiplication processing unit, a second multiplication processing unit, a second convolution unit, a second feature fusion unit, a third activation function unit and a third feature fusion unit;

[0092] The first feature fusion unit adds the bottom-level feature map T1 output by the first feature extraction module and the bottom-level feature map T2 output by the second feature extraction module, and then inputs the sum into the first activation function unit. The first activation function unit strengthens the root features in the feature map and then inputs the feature map into the first convolution unit. The first convolution unit performs convolution processing on the feature map and then inputs the sum into the second activation function unit. The second activation function unit suppresses the background response in the feature map and then inputs the sum into the first multiplication processing unit. The first multiplication processing unit multiplies the sum with the bottom-level feature map T1 output by the first feature extraction module and then outputs the sum.

[0093] The second multiplication processing unit multiplies the bottom feature map T1 output by the first feature extraction module and the bottom feature map T2 output by the second feature extraction module, and then inputs the result into the second convolution unit. The second convolution unit performs convolution processing on the feature map and then inputs the result into the second feature fusion unit. The second feature fusion unit adds the result and the bottom feature map T2 output by the second feature extraction module, and then inputs the result into the third activation function unit. The third activation function unit outputs the result after strengthening the root features in the feature map.

[0094] The third feature fusion unit adds the feature map output by the first multiplication processing unit and the feature map output by the third activation function unit and outputs a feature map T.

[0095] In this example, the first activation function unit and the third activation function unit both use the ReLU activation function, the second activation function unit uses the Sigmoid activation function, and the first convolution unit and the second convolution unit are both 1×1×1 convolutions.

[0096] The root semantic segmentation model described in this embodiment improves the sensitivity of the model to small root features by introducing the attention mechanism, and realizes the accurate segmentation of cotton roots under complex backgrounds. By adding the attention module, the model focuses on more important features and suppresses unnecessary features. The model not only performs better, but also has stronger robustness to noise input. Experiments have shown that the visual self-attention mechanism can not only optimize the boundary of segmentation, but also reduce the phenomenon of missed segmentation and mis-segmentation. At the same time, the attention module improves the feature discrimination and selection ability of the network, thereby alleviating the problem of reduced accuracy caused by interference features, and the adopted attention structure can focus on more important channel features, reducing the complexity of the network.

[0097] Step 2.3: Use the training set to train the root semantic segmentation model;

[0098] In this example, the loss function for training the root semantic segmentation model is:

[0099] L fl =-α(1-y'*y)*γ*log(y'),

[0100] Among them, α and γ represent weight coefficients, y' is the model output result, and y is the model input result.

[0101] Step 2.4: Use the test set to test the trained root semantic segmentation model to obtain a trained root semantic segmentation model;

[0102] The trained root semantic segmentation model should achieve the following segmentation effect on high-resolution images (2550*2273, slightly different depending on the acquisition device and parameter settings): mPA (Mean Pixel Accuracy) is not less than 85%, MAE (Mean absolute Error) is not higher than 15%, and F-score index is not less than 90%.

[0103] Step 2.5: Use the trained root semantic segmentation model to process the root image to obtain Figure 10 The cotton root system characteristics diagram is shown.

[0104] It can be seen that step 2 of this embodiment establishes a root semantic segmentation model with higher precision and finer granularity. It not only does not require artificially designed features for guidance, but the model itself can learn the required features from the training data and make reasonable use of them. Therefore, it has better generalization ability and can perform stably even in complex scenes. It improves the model's sensitivity to small root features and realizes accurate segmentation of cotton roots under complex backgrounds, thereby providing data support for root detection and analysis, and solves the problem that the existing model has limited generalization ability due to its reliance on experimental dataset models, and has unsatisfactory processing effect on image targets such as cotton roots that are basically composed of edges and branch outlines.

[0105] Step 3: preparing a scale card for quantitatively detecting the structure and distribution of cotton root systems;

[0106] This embodiment designs a set of metric scale charts suitable for quantitatively testing cotton root structure and distribution. The scale charts include a reference chart and a slope detection chart set. The reference chart, composed of horizontal and vertical scale segments, is used to detect the horizontal and vertical lengths of rhizomes and also to determine the diameter of the rhizomes being tested. The slope detection chart set consists of a set of regularly spaced unit segments with the same width as the reference chart's unit segments and a certain slope, used to detect the lengths of rhizomes in different extension directions.

[0107] Specifically: The ruler card is composed of unit length line segments of different widths. The width of the unit line segment is used to measure rhizomes of different diameters. The straight lines composed of the unit line segments are parallel and evenly distributed on the ruler card at certain intervals. The straight line intervals and the intervals of the unit line segments can be adjusted according to the model training loss. According to the distribution characteristics of the cotton root system image, the card image with the same width includes two reference card images in the horizontal and vertical directions, which are used to determine the diameter of the rhizome to be measured and the length of the rhizome growing in these two directions; with the center point of the reference card image as the axis point, the ruler line of the card image is rotated clockwise to 180 degrees at intervals of 10 degrees, forming a set of 18 slope detection images covering 360 degrees of full slope, which are used to detect rhizomes in different extension directions. Figure 11 shown.

[0108] During implementation, the size of the ruler card is the same as the size of the cotton root system image to be detected. In this embodiment, it is set to 640*640. The ruler card is composed of standard unit line segments with fixed intervals in the same direction. Among them, the reference card is composed of horizontal and vertical ruler line segments, which are used to detect the length of the rhizome in the horizontal and vertical directions, and at the same time to determine the diameter of the rhizome to be detected, such as Figure 11 (a) with Figure 11 (b) As shown; the slope detection set is a set of unit line segments with the same width as the reference card and a certain slope, which are used to detect the length of the root in different extension directions, such as Figure 11 (c) shown.

[0109] Step 4: Detect the cotton root system characteristic map using the root and stem parameter quantitative detection model based on the XOR residual structure and the ruler card map;

[0110] This embodiment designs a learning model based on an XOR residual structure, namely the rhizome parameter quantitative detection model. It takes a cotton root segmentation image and reference card images of different unit widths as input. After model processing, XOR residual card image targets are generated. Targets larger than 1 to 2 standard units are retained, and the reference card image with the smallest unit width is used as the measured rhizome diameter. A sequence of card images corresponding to the unit width is selected for length measurement. The rhizome to be measured is subjected to XOR residual processing with 18 pairs of attached ruler cards in different slope directions through the learning model to measure the growth length of the rhizome. A residual card image feature of at least one unit length is generated. The 18 residual card image features are channel-fused, and discrete targets are eliminated to achieve the purpose of replacing continuous rhizome targets with standard unit lengths. Finally, the number of unit lengths on the residual card image feature map is measured, and the continuous rhizome length values ​​are fitted. Through model learning training, the structure is optimized, parameters are adjusted, and the card image design standards are improved to approximate the actual growth parameters of the rhizome.

[0111] In the specific implementation process, the root and stem parameter quantitative detection model has two implementation methods, specifically:

[0112] like Figure 12 As shown, in the first implementation, the specific steps of detecting the cotton root system characteristic map using the rhizome parameter quantitative detection model and the ruler card map are as follows:

[0113] Step A1: Detecting the rhizome diameter using a reference card, performing an XOR calculation using a rhizome parameter quantitative detection model, and retaining the reference card with two standard units as the selected detection scale for the corresponding diameter rhizome;

[0114] Step A2: selecting a card diagram series from the slope detection set for detection according to the selected detection scale;

[0115] Step A3: performing full slope detection on the possible extension direction of the rhizome according to the selected card series;

[0116] Step A4: splicing the standard units generated by the full slope detection, and counting the number of standard line segments to approximate the rhizome length.

[0117] like Figure 13 As shown, in the second implementation, the specific steps of using the rhizome parameter quantitative detection model and the scale card diagram to detect the cotton root system characteristic map are as follows:

[0118] Step B1: Detect the rhizome diameter using a reference card image, and perform XOR calculation using the XOR residual module in the rhizome parameter quantitative detection model to obtain the extended length of the rhizome in each direction, and retain the reference card image with a value greater than 1 standard unit (pixel ratio greater than 80% is 1) as the selected detection scale for the corresponding diameter rhizome;

[0119] Step B2: Select a card diagram series from the slope detection set for detection based on the selected detection scale;

[0120] Step B3: using the feature fusion module in the rhizome parameter quantitative detection model to merge the features of each slope direction;

[0121] Step B4: applying the small target detection module in the rhizome parameter quantitative detection model to perform detection and counting, and counting the number of standard line segments to fit the rhizome length;

[0122] Step B5: training the adjustment network parameters and card diagram settings of the rhizome parameter quantitative detection model, optimizing the detection results, and obtaining the final detection results.

[0123] This embodiment uses the above-mentioned rhizome parameter quantitative detection model based on the ruler card diagram to convert the continuous parameter measurement of the rhizome into a quantitative measurement of standard units, perform dimensionality reduction approximate calculations, and optimize the super parameters and the ruler card diagram through model training to obtain the best estimated value. This not only realizes the automated cotton rhizome measurement, but also effectively improves the accuracy of the measurement results.

[0124] Step 5: Calculate cotton root phenotypic parameters such as root length, root diameter, and number of root branches based on the test results, and calculate the dynamic evolution parameters of the cotton root phenotypic parameters such as root growth, branching, root breaking, and ablation rate based on the sampling time series;

[0125] Step 6: Based on the cotton root phenotypic parameters, dynamic evolution parameters and cotton growth conditions, a cotton root growth dynamic evolution model based on root phenotype is established;

[0126] Step 6.1, based on the phenotypic parameters of the cotton root system, studying the root phenotypic growth characteristics of the control group;

[0127] Step 6.2: Based on the dynamic evolution parameters of cotton plants, study and analyze the impact of planting factors on cotton growth conditions;

[0128] Step 6.3: Combine the cotton root phenotypic growth characteristics and planting factors to determine the impact of cotton growth conditions such as cotton plant growth, boll quantity and quality, and cottonseed breeding, and establish a cotton growth inversion model based on root phenotype.

[0129] Step 7: Input a new cotton root system image, use the cotton root system growth dynamic evolution model to detect and analyze the new cotton root system image, and predict the growth status of the cotton plant.

[0130] Example 2:

[0131] like Figure 14 As shown, the embodiment of the present invention provides a cotton root phenotype detection and analysis system for implementing the method described in Example 1, specifically comprising:

[0132] A root system image acquisition module is used to acquire cotton root system images using a root system image acquisition device;

[0133] A semantic segmentation module is used to process cotton root images using a trained root semantic segmentation model to obtain a cotton root feature map;

[0134] A scale card image generation module is used to create scale cards for quantitatively detecting cotton root structure and distribution;

[0135] The rhizome length detection module is used to detect the cotton root system characteristic map using the rhizome parameter quantitative detection model based on the XOR residual structure and the ruler card diagram;

[0136] An evolution parameter calculation module is used to calculate cotton root phenotypic parameters based on the detection results, and to calculate the dynamic evolution parameters of the cotton root phenotype based on the sampling time series;

[0137] An evolution model building module is used to build a cotton root growth dynamic evolution model based on root phenotype according to cotton root phenotypic parameters, dynamic evolution parameters and cotton growth conditions;

[0138] The detection and analysis module is used to detect and analyze new cotton root system images using the cotton root system growth dynamic evolution model.

[0139] In summary, the present invention first establishes a root semantic segmentation model with higher precision and finer granularity to semantically segment cotton root images. This model not only does not require artificially designed features for guidance, but can also learn the required features from training data and make reasonable use of them. This model therefore has better generalization capabilities and can maintain stable performance even in complex scenes. Furthermore, by introducing an attention mechanism, the model's sensitivity to small root features is improved, achieving accurate cotton root segmentation under complex backgrounds. By adding an attention module, the model focuses on more important features and suppresses unnecessary features, resulting in better performance and greater robustness to noisy inputs. Furthermore, a quantitative detection method for cotton root images based on a scale card image is provided. This method uses a combination of techniques, including a metric scale card image, an XOR residual structure, pixel detection, and small target counting, to analyze and process the semantically segmented root images. This method converts the continuous parameter detection of image target size into a discrete target count based on standard units, achieving intelligent estimation of root target growth parameters based on a learning model without human intervention. This method can serve as a technical and data support for cotton crop cultivation and improvement research.

[0140] The technical solution provided by the present invention is introduced in detail above. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, the present invention can also be improved and modified in a number of ways, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A cotton root phenotype detection and analysis method, characterized in that: Including steps: Step 1: Acquire a cotton root system image using a root system image acquisition device; Step 2: Use the trained root semantic segmentation model to process the cotton root image to obtain a cotton root feature map; Step 3: preparing a scale card for quantitatively detecting cotton root structure and distribution; Step 4: Detect the cotton root system characteristic map using the root and stem parameter quantitative detection model based on the XOR residual structure and the ruler card map; Step 5: Calculate cotton root phenotypic parameters based on the test results, and calculate the dynamic evolution parameters of the cotton root phenotype based on the sampling time series; Step 6: Based on the cotton root phenotypic parameters, dynamic evolution parameters and cotton growth conditions, a cotton root growth dynamic evolution model based on root phenotype is established; Step 7: Use the cotton root growth dynamic evolution model to detect and analyze the new cotton root system image to predict the growth status of the cotton plant.

2. The cotton root phenotype detection and analysis method according to claim 1, characterized in that: In step 2, the cotton root system image is processed using the trained root system semantic segmentation model to obtain the cotton root system feature map. The specific steps are as follows: Step 2.1: Preprocess the obtained cotton root images, annotate the image objects, and construct training and test sets; Step 2.2: Establish a root semantic segmentation model based on the attention mechanism; Step 2.3: Use the training set to train the root semantic segmentation model; Step 2.4: Use the test set to test the trained root semantic segmentation model to obtain a trained root semantic segmentation model; Step 2.5: Use the trained root semantic segmentation model to process the segmented root image to obtain the cotton root feature map.

3. The cotton root phenotype detection and analysis method according to claim 2, characterized in that: The root system semantic segmentation model includes a first feature extraction module, a second feature extraction module, an attention module, a feature detection module, and a feature fusion module, wherein the first feature extraction module is used to extract multi-scale regional features of the main root from the cotton root system image; the second feature extraction module is used to extract multi-scale regional features of lateral roots and root hairs from the cotton root system image; the attention module is used to use the attention mechanism to process the bottom-level feature map in the multi-scale feature map output by the first feature extraction module and the second feature extraction module; the feature detection module is used to perform feature detection on the regional features of different scales output by the first feature extraction module, the second feature extraction module, and the attention module; the feature fusion module is used to splice and fuse the feature detection module and the output regional features through multi-scale feature cross-grafting to obtain a cotton root system feature map.

4. The cotton root phenotype detection and analysis method according to claim 3, characterized in that: The attention module includes a first feature fusion unit, a first activation function unit, a first convolution unit, a second activation function unit, a first multiplication processing unit, a second multiplication processing unit, a second convolution unit, a second feature fusion unit, a third activation function unit and a third feature fusion unit; The first feature fusion unit adds the bottom-level feature map output by the first feature extraction module and the bottom-level feature map output by the second feature extraction module, and then inputs the sum into the first activation function unit. The first activation function unit strengthens the root features in the feature map and then inputs the sum into the first convolution unit. The first convolution unit performs convolution processing on the feature map and then inputs the sum into the second activation function unit. The second activation function unit suppresses the background response in the feature map and then inputs the sum into the first multiplication processing unit. The first multiplication processing unit multiplies the sum with the bottom-level feature map output by the first feature extraction module and then outputs the sum. The second multiplication processing unit multiplies the lowest-level feature map output by the first feature extraction module and the lowest-level feature map output by the second feature extraction module, and then inputs the result into the second convolution unit. The second convolution unit convolves the feature map and then inputs the result into the second feature fusion unit. The second feature fusion unit adds the result and the lowest-level feature map output by the second feature extraction module, and then inputs the result into the third activation function unit. The third activation function unit outputs the result after strengthening the root features in the feature map. The third feature fusion unit performs addition processing on the feature map output by the first multiplication processing unit and the feature map output by the third activation function unit.

5. The cotton root phenotype detection and analysis method according to claim 3, characterized in that: The first feature extraction module adopts the transformer network, the second feature extraction module adopts the CNN network, and the feature detection module adopts the FPN network.

6. The cotton root phenotype detection and analysis method according to claim 1, characterized in that: The ruler card diagram includes a reference card diagram and a slope detection set diagram. The reference card diagram is composed of horizontal and vertical ruler line segments, which is used to detect the length of the rhizome in the horizontal and vertical directions, and is also used to determine the diameter of the rhizome to be detected; the slope detection set diagram is a set of unit line segments with a certain slope and a fixed interval, which are the same width as the unit line segments of the reference card diagram, and are used to detect the length of the rhizome in different extension directions.

7. The cotton root phenotype detection and analysis method according to claim 1, characterized in that: The specific steps of detecting the cotton root system characteristic map using the root and stem parameter quantitative detection model based on the XOR residual structure and the ruler card map in step 4 are as follows: Step A1: Detecting the rhizome diameter using a reference card, performing XOR residual processing using a rhizome parameter quantitative detection model, and retaining the reference card with two standard units as the selected detection scale for the corresponding diameter rhizome; Step A2: selecting a card diagram series from the slope detection set for detection according to the selected detection scale; Step A3: performing full slope detection on the possible extension direction of the rhizome according to the selected card series; Step A4: splicing the standard units generated by the full slope detection, and counting the number of standard line segments to approximate the rhizome length.

8. The cotton root phenotype detection and analysis method according to claim 1, characterized in that: The specific steps of detecting the cotton root system characteristic map using the root and stem parameter quantitative detection model based on the XOR residual structure and the ruler card map in step 4 are as follows: Step B1: Detect the rhizome diameter using a reference card image, and perform XOR residual processing using the XOR residual module in the rhizome parameter quantitative detection model to obtain the extended length of the rhizome in each direction, and retain the reference card image larger than 1 standard unit as the selected detection scale of the corresponding diameter rhizome; Step B2: Select a card diagram series from the slope detection set for detection based on the selected detection scale; Step B3: using the feature fusion module in the rhizome parameter quantitative detection model to merge the features of each slope direction; Step B4: applying the small target detection module in the rhizome parameter quantitative detection model to perform detection and counting, and counting the number of standard line segments to fit the rhizome length; Step B5: training the adjustment network parameters and card diagram settings of the rhizome parameter quantitative detection model, optimizing the detection results, and obtaining the final detection results.

9. The cotton root phenotype detection and analysis method according to claim 1, characterized in that: The specific steps for establishing the cotton root growth dynamic evolution model based on root phenotype described in step 6 are as follows: Step 6.1, based on the phenotypic parameters of the cotton root system, studying the root phenotypic growth characteristics of the control group; Step 6.2: Based on the dynamic evolution parameters of cotton plants, study and analyze the impact of planting factors on cotton growth conditions; Step 6.3: Combine the cotton root phenotypic growth characteristics and the impact of planting factors on cotton growth conditions to establish a cotton growth inversion model based on root phenotype.

10. A cotton root phenotype detection and analysis system, characterized in that: include: A root system image acquisition module is used to acquire cotton root system images using a root system image acquisition device; A semantic segmentation module is used to process cotton root images using a trained root semantic segmentation model to obtain a cotton root feature map; A scale card image generation module is used to create scale cards for quantitatively detecting cotton root structure and distribution; The rhizome length detection module is used to detect the cotton root system characteristic map using the rhizome parameter quantitative detection model based on the XOR residual structure and the ruler card diagram; An evolution parameter calculation module is used to calculate cotton root phenotypic parameters based on the detection results, and to calculate the dynamic evolution parameters of the cotton root phenotype based on the sampling time series; An evolution model building module is used to build a cotton root growth dynamic evolution model based on root phenotype according to cotton root phenotypic parameters, dynamic evolution parameters and cotton growth conditions; The detection and analysis module is used to detect and analyze new cotton root system images using the dynamic evolution model of cotton root system growth and predict the growth status of cotton plants.

Images