Intelligent spectrum image detection system and method for multiple stress resistance of cotton

Abstract

The invention is applied to the technical field of cotton stress resistance detection, and discloses a cotton multi-stress resistance intelligent spectral image detection system, which comprises a spectral image acquisition module, a data transmission module, a data processing unit, a stress resistance analysis model construction module and a result output module. According to the intelligent spectral image detection system for the multiple stress resistance of the cotton, spectral images can reflect deep physiological structure and chemical component changes in the cotton, and multi-dimensional spectral feature information such as spectral reflectivity features, spectral absorption features, spectral slope features and spectral texture features is extracted; compared with a traditional detection method based on appearance phenotypes or single physiological and biochemical indexes, the method has the advantages that subtle differences of the cotton in different stress resistance states can be captured more comprehensively and accurately, so that the accuracy of evaluation of multiple stress resistance properties such as drought resistance, salt resistance, high temperature resistance, low temperature resistance and waterlogging resistance of the cotton is remarkably improved, and a reliable basis is provided for cotton stress resistance identification.

Description

Technical Field

[0001] The present invention relates to the technical field of cotton stress resistance detection, and in particular to an intelligent spectral image detection system and method for cotton multi-stress resistance. Background Art

[0002] Cotton occupies a pivotal position in the global agricultural economy and is an indispensable raw material for many fields such as the textile industry. However, its growth and development process faces many challenges. Adverse conditions such as drought, salinization, high temperature, low temperature and floods frequently occur, which have a serious negative impact on the yield and quality of cotton, and thus threaten the stability and sustainable development of the entire cotton industry. Traditional cotton stress resistance detection methods have many limitations. Field phenotypic identification is a more common traditional method. Technicians need to observe and record the appearance of cotton at different growth stages in the field for a long time, such as plant height, leaf morphology, flowering and fruiting, etc., to infer the stress resistance of cotton. This method not only consumes a lot of manpower and material resources, but also has a long process cycle. More importantly, the field environment is complex and changeable. Differences in factors such as light, temperature, humidity and soil fertility will interfere with the identification results, resulting in a significant reduction in accuracy and stability. For example, when evaluating the drought resistance of cotton, if local precipitation or uneven irrigation happens to occur during the identification period, the drought-sensitive varieties may show tolerance that is inconsistent with the actual situation, thereby misleading subsequent variety selection and planting decisions;

[0003] The determination of physiological and biochemical indicators starts from the physiological changes and chemical composition inside the cotton plant, and measures its stress resistance by testing indicators such as proline content, antioxidant enzyme activity, and cell membrane permeability. Although this method can theoretically reflect the stress resistance mechanism of cotton more deeply, the actual operation is extremely complicated. It requires professional laboratory equipment and skilled technicians. From the collection and preservation of cotton samples to the precise determination of various indicators, each link has strict operating specifications and quality control requirements, which undoubtedly increases the cost of testing. Moreover, due to the complex experimental process and sample processing involved, it is difficult to achieve large-scale and rapid testing, and cannot meet the urgent needs of the modern cotton industry for efficient and accurate testing technology;

[0004] In this context, intelligent spectral image detection technology came into being, bringing new ideas and solutions to cotton multi-stress resistance detection. Spectral images can capture the reflection and absorption characteristics of cotton in different spectral ranges from visible light to near-infrared bands. These characteristics are not simple appearances, but are closely related to the cell structure, photosynthetic pigment content, moisture status, nutrients and various stress resistance-related physiological and biochemical processes inside the cotton. For example, when suffering from drought stress, the moisture content of cotton leaves decreases and the cell structure changes, which will cause regular changes in its reflectivity and absorption characteristics in specific spectral bands; under salinity stress, the ion balance in the cotton body is broken, and it will also show a unique response pattern on the spectral image. The spectral image data of cotton in different growth periods and under different stress resistance treatments are collected through high-precision spectral cameras, and these data are transmitted quickly and stably with the help of advanced data transmission technology. The data is transmitted to the data processing unit in a fixed location, and the storage, preprocessing, screening, feature extraction and other modules in the data processing unit are used to deeply mine and analyze the data. Finally, combined with the precise model constructed by the stress resistance analysis model construction module, the drought resistance, salt resistance, high temperature resistance, low temperature resistance, waterlogging resistance and other stress resistance of cotton can be quickly and accurately evaluated. This not only greatly improves the detection efficiency and reduces the labor and time costs, but also effectively eliminates the interference of external environmental factors, provides a more reliable basis for the selection and breeding of cotton varieties, and helps growers to adjust the planting management strategies in time to achieve high-yield and high-quality cultivation of cotton. At the same time, it also provides strong technical support for the cotton industry in responding to climate change and adverse stress, and promotes the entire cotton industry to move towards intelligence and precision. Summary of the invention

[0005] The object of the present invention is to provide an intelligent spectral image detection system and method for cotton multi-stress resistance, so as to solve the problem of poor detection method proposed in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions: an intelligent spectral image detection system for cotton multi-stress resistance, comprising a spectral image acquisition module, a data transmission module, a data processing unit, a stress resistance analysis model construction module and a result output module;

[0007] Spectral image acquisition module: used to collect spectral image data of cotton at different growth stages and under different stress resistance treatments. The spectral image acquisition device includes a bracket with adjustable angle and height, a spectral camera, and a controller for controlling the shooting parameters of the spectral camera;

[0008] Data transmission module: used to transmit the spectral image data collected by the spectral image collection device to the data processing unit;

[0009] Data processing unit: including a data storage module, a data preprocessing module and a feature extraction module, wherein the data storage module is used to store spectral image data and related cotton variety information, the data preprocessing module is used to perform denoising, correction and normalization processing on the spectral image data, and the feature extraction module is used to extract spectral feature information related to cotton stress resistance from the preprocessed spectral image data;

[0010] Stress resistance analysis model building module: used to build a cotton multi-stress resistance analysis model based on the extracted spectral feature information and cotton sample data with known stress resistance. The stress resistance analysis model can analyze and evaluate the drought resistance, salt resistance, high temperature resistance, low temperature resistance and waterlogging resistance of cotton;

[0011] Result output module: used to output and display the analysis and evaluation results of the stress resistance analysis model.

[0012] Preferably, the spectral range of the spectral camera covers the characteristic absorption spectral range of cotton from visible light to near-infrared band, and its resolution is not less than 500 pixels.

[0013] The above technical solution can realize rapid, non-destructive and accurate detection of cotton's multiple stress resistance, providing strong data support for cotton variety selection, planting management and stress resistance research.

[0014] Preferably, the data processing unit further comprises a data screening module, and the data screening module is used to screen out effective spectral image data for subsequent processing according to preset cotton growth stages and stress resistance treatment conditions.

[0015] The above technical solution can effectively screen out spectral image data that meets specific needs, reduce the interference of invalid data on subsequent analysis, and improve the pertinence and effectiveness of data processing.

[0016] Preferably, the spectral feature information extracted by the feature extraction module includes spectral reflectance features, spectral absorption features, spectral slope features and spectral texture features.

[0017] By adopting the above technical solution, the multi-dimensional spectral feature information helps to comprehensively and accurately reflect the stress resistance characteristics of cotton, laying the foundation for building a high-precision stress resistance analysis model.

[0018] Preferably, the stress resistance analysis model construction module uses a machine learning algorithm to construct a stress resistance analysis model, and the machine learning algorithm includes but is not limited to a support vector machine, a random forest, and a neural network algorithm.

[0019] By adopting the above technical solutions and using advanced machine learning algorithms, we can fully explore the complex relationship between spectral characteristics and cotton stress resistance and improve the predictive ability and reliability of the model.

[0020] Preferably, the detection system also includes a sample expansion module, which is used to expand cotton sample data with known stress resistance through data enhancement technology to improve the accuracy and generalization ability of the stress resistance analysis model.

[0021] By adopting the above technical solution, the sample expansion module can increase the number and diversity of samples, so that the stress resistance analysis model can perform evaluation and analysis more stably and accurately when facing different cotton samples.

[0022] An intelligent spectral image detection method for cotton multi-stress resistance, the detection method comprising the following steps:

[0023] S1. Using a spectral image acquisition device to collect spectral image data of cotton at different growth stages and under different stress resistance treatments;

[0024] S2, transmitting the collected spectral image data to the data processing unit through the data transmission module;

[0025] S3, the data preprocessing module in the data processing unit performs denoising, correction and normalization on the spectral image data;

[0026] S4, a feature extraction module extracts spectral feature information related to cotton stress resistance from the preprocessed spectral image data;

[0027] S5, a stress resistance analysis model building module builds a cotton multi-stress resistance analysis model according to the extracted spectral feature information and cotton sample data with known stress resistance;

[0028] S6. Input the spectral image data of the cotton to be tested into the constructed stress resistance analysis model, and output the stress resistance analysis evaluation results of the cotton through the result output module

[0029] Preferably, S4 also includes the steps of performing correlation analysis and feature selection on the spectral feature information to remove redundant and low-correlation features to improve the efficiency and accuracy of model building.

[0030] By adopting the above technical solution, correlation analysis and feature selection can reduce data dimensions, avoid model overfitting, speed up model training and improve its generalization performance.

[0031] Preferably, S5 also includes a step of optimizing parameters of the machine learning algorithm, and the parameter optimization uses a cross-validation method to determine the optimal parameter combination.

[0032] Using the above technical solution and cross-validation to optimize the machine learning algorithm parameters can ensure that the model can maintain good accuracy and stability under different parameter settings, thereby improving the overall performance of the model.

[0033] Preferably, the detection system also includes the step of regularly updating the stress resistance analysis model, adding newly acquired cotton sample data with accurate stress resistance identification results to the model building process, so as to continuously optimize the model performance and adapt to the detection needs of different new cotton germplasms.

[0034] By adopting the above technical solution, regular updating of the stress resistance analysis model can enable it to adapt to the changes and development of cotton germplasm resources in a timely manner, ensuring the timeliness and adaptability of the detection system.

[0035] Compared with the prior art, the invention has the following beneficial effects: the intelligent spectral image detection system for cotton multi-stress resistance:

[0036] 1. Spectral images can reflect the changes in the deep physiological structure and chemical composition of cotton. By extracting multi-dimensional spectral feature information such as spectral reflectance characteristics, spectral absorption characteristics, spectral slope characteristics and spectral texture characteristics, compared with traditional detection methods based on appearance phenotypes or single physiological and biochemical indicators, they can more comprehensively and accurately capture the subtle differences of cotton under different stress resistance conditions, thereby significantly improving the accuracy of cotton drought resistance, salt resistance, high temperature resistance, low temperature resistance, waterlogging resistance and other stress resistance evaluation;

[0037] Furthermore, the data screening module in the data processing unit can screen out effective spectral image data according to the preset cotton growth stage and stress treatment conditions, remove invalid or interfering data, and perform correlation analysis and feature selection after feature extraction to remove redundant and low-correlation features, further focusing on key information, making the input data of the constructed stress resistance analysis model more representative and effective, greatly improving the accuracy of the model output results, and providing a reliable basis for the identification of cotton stress resistance.

[0038] 2. The intelligent spectral image detection system can realize the rapid acquisition of spectral image data of cotton in different growth periods and under different stress resistance treatments. The spectral camera with adjustable angle and height bracket can flexibly adapt to the monitoring needs of different planting layouts and cotton growth trends, and the data transmission module can timely transmit a large amount of data to the data processing unit, which greatly shortens the detection cycle. Compared with traditional long-term field observation and complex laboratory determination, it significantly improves the detection efficiency and can perform stress resistance detection on a large number of cotton samples in a shorter time;

[0039] Furthermore, machine learning algorithms are used to build stress resistance analysis models, such as support vector machines, random forests, neural network algorithms, etc. Once the model is built, only the spectral image data of the cotton to be tested needs to be input to quickly obtain the stress resistance analysis and evaluation results. There is no need to repeat the tedious measurement and analysis process for each sample as in traditional methods, thus achieving efficient automation of cotton stress resistance detection.

[0040] 3. Reduced human cost investment. Traditional field phenotyping requires a large number of technicians to stay in the field for a long time to observe and record, while the intelligent system only requires a small number of people to operate the spectral image acquisition device and the monitoring system. The data processing and analysis process mainly relies on automated data processing units and model building modules, reducing dependence on manual operation.

[0041] Furthermore, compared with the traditional physiological and biochemical index determination, which requires the purchase of a large number of expensive professional laboratory instruments and equipment, the spectral image acquisition device is relatively low-cost and reusable. At the same time, due to the improvement of detection efficiency, it also reduces the consumption of other related materials caused by the long detection cycle, achieving effective cost control from many aspects and improving the economic benefits of cotton stress resistance detection;

[0042] Furthermore, data enhancement technology is used to expand the cotton sample data with known stress resistance, enriching the model's training data. This enables the stress resistance analysis model to better learn the characteristic patterns of different cotton samples under various stress resistance conditions, thereby improving the model's accuracy and generalization ability. It can adapt to the stress resistance detection needs of different cotton varieties and different planting environments, enhance the model's reliability and stability in practical applications, and at the same time, can add newly acquired cotton sample data with accurate stress resistance identification results to the model construction process, continuously optimize the model performance, and effectively detect the stress resistance of new germplasm in a timely manner. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] The present invention provides a technical solution: an intelligent spectral image detection system for cotton multi-stress resistance.

[0045] Example 1: Application in breeding of new cotton varieties

[0046] In an experimental field with an area of ​​5,000 square meters, 1,000 cotton seedlings of different hybrid combinations were planted, and the row spacing was set at 0.5 meters by 0.5 meters to ensure that each cotton plant had enough growth space and was convenient for spectral image collection. The intelligent spectral image detection system was used to collect data during the seedling stage, bud stage, flowering stage and boll stage of cotton growth.

[0047] The adjustable bracket height range of the spectral image acquisition module is 0.5 meters to 2 meters, and the angle can be flexibly adjusted between 0-360 degrees in the horizontal direction and -30 degrees to 90 degrees in the vertical direction. The spectral range of the spectral camera is set to 400-1200 nanometers, accurately covering the characteristic absorption spectrum range of cotton from visible light to near-infrared bands. Its resolution is 800 pixels, which can clearly capture the subtle spectral characteristic changes of cotton leaves, stems and other parts. Each time the camera is collected, it takes 5 spectral images of each cotton plant from different angles to ensure the comprehensiveness of the data. The data transmission module transmits the collected spectral image data to the data processing unit at a speed of 100 megabytes per second;

[0048] The data screening module in the data processing unit selects valid data from the collected massive data according to the preset growth stage conditions. For example, in the seedling stage, only data related to the early growth characteristics of cotton seedlings are retained, and interference data caused by factors such as soil background are excluded. The data preprocessing module uses a filtering algorithm to denoise the data and reduce the noise signal to less than 5% of the original signal; the error of the spectral data is controlled within ±2% by comparing with the standard spectral library; the data is mapped to a numerical range of 0-1 using a normalization algorithm to facilitate subsequent processing. The feature extraction module extracts spectral reflectance features (such as the reflectance value at 700 nanometers), spectral absorption features (the depth and width of the absorption peak in a specific band), spectral slope features (such as the slope value of the 500-600 nanometer band) and spectral texture features (texture parameters such as contrast and correlation are calculated through the grayscale co-occurrence matrix).

[0049] The stress resistance analysis model construction module uses the existing sample data of 500 cotton varieties with known stress resistance. These data are expanded to 2000 sample data through the sample expansion module using data enhancement techniques such as mirroring, rotation, and adding noise. The random forest algorithm is used to construct a cotton multi-stress resistance analysis model, and the algorithm parameters are optimized through a 5-fold cross-validation method. When the stress resistance of cotton in the process of new variety breeding is tested, the spectral image data of the cotton to be tested is input into the constructed model. The results show that one of the cotton hybrid combinations numbered C-12 has a model evaluation score of 85 points (out of 100 points) in terms of drought resistance and a salt tolerance score of 78 points, which are significantly higher than other combinations. The combination is further verified in the field, and C-12 and other control varieties are planted in test areas simulating drought (soil moisture content controlled at 5%-10%) and saline soil environments (soil salt content of 0.3%-0.5%). After one growing season of observation, the yield of C-12 cotton was 18% higher than that of the control variety on average, the cotton fiber length was 3 mm longer than that of the control variety, and the micronaire value was within the appropriate range and more stable, proving the effectiveness of the intelligent spectral image detection system in screening stress resistance in the breeding of new cotton varieties, greatly improving breeding efficiency and shortening the breeding cycle by about 20%.

[0050] Example 2: Application in cotton planting field management

[0051] A large cotton farm with a planting area of ​​1,000 mu is divided into 20 plots, each with an area of ​​about 50 mu. During the cotton growing period, the intelligent spectral image detection system is used to monitor the cotton in each plot every 7 days;

[0052] The spectral image acquisition device is carried by an electric trolley and can move quickly between plots. The spectral range of its spectral camera is 450-1100 nanometers and the resolution is 600 pixels. In each plot, collection points are set according to a 5m×5m grid. Each collection point takes spectral images of cotton plants from four directions. Each collection acquires about 400 images in total. The data transmission module transmits data to the data processing unit at a speed of 80 megabytes per second.

[0053] After being processed by the data processing unit, the data is input into the stress resistance analysis model based on the support vector machine algorithm. The model is built based on 1,000 sets of data accumulated from the farm planting the same cotton variety under different climate and soil conditions in the past five years, and is improved by combining 2,000 sets of general data from other regions in the industry, and is updated regularly every month.

[0054] The test results showed that in a test in March, the cotton of Plot A, which is located at the edge of the farm near the hillside and covers an area of ​​48 mu, showed an early warning of decreased low-temperature resistance. The model-assessed low-temperature resistance index was only 30 (out of 100). Based on this result, Plot A was promptly covered with a 0.03-meter-thick mulch and 50 kg of thermal fertilizer (such as wood ash) was added per mu to prevent cold and keep warm. In Plot B, which covers an area of ​​52 mu, it was detected that the cotton had a potential risk of waterlogging in June, and its waterlogging resistance assessment value was 40 (out of 100). Therefore, the farm cleared the surrounding drainage channels in advance, making the channel depth reach 0.8 meters and the width reach 1.5 meters to ensure timely drainage when the rainy season comes.

[0055] Through continuous monitoring and targeted management throughout the growing season, the farm's cotton yield increased by about 15% compared with the previous year under similar climatic conditions. In terms of cotton fiber quality, the fiber strength increased by an average of 1.2 cm / tex, and the short fiber rate decreased by 3%. This shows that the cotton multi-stress resistance intelligent spectral image detection system can accurately detect changes in cotton stress resistance during cotton field planting management, providing a basis for timely and effective management measures, thereby ensuring high-yield and high-quality production of cotton, improving agricultural production efficiency, and reducing yield losses caused by natural disasters by about 25%.

[0056] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific embodiments of the present invention and do not limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are included in the protection scope of the present invention.

Claims

1. An intelligent spectral image detection system for cotton multi-stress resistance, characterized in that: It includes a spectral image acquisition module, a data transmission module, a data processing unit, a stress resistance analysis model building module and a result output module; Spectral image acquisition module: used to collect spectral image data of cotton at different growth stages and under different stress resistance treatments. The spectral image acquisition device includes a bracket with adjustable angle and height, a spectral camera, and a controller for controlling the shooting parameters of the spectral camera; Data transmission module: used to transmit the spectral image data collected by the spectral image collection device to the data processing unit; Data processing unit: including a data storage module, a data preprocessing module and a feature extraction module, wherein the data storage module is used to store spectral image data and related cotton variety information, the data preprocessing module is used to perform denoising, correction and normalization processing on the spectral image data, and the feature extraction module is used to extract spectral feature information related to cotton stress resistance from the preprocessed spectral image data; Stress resistance analysis model building module: used to build a cotton multi-stress resistance analysis model based on the extracted spectral feature information and cotton sample data with known stress resistance. The stress resistance analysis model can analyze and evaluate the drought resistance, salt resistance, high temperature resistance, low temperature resistance and waterlogging resistance of cotton; Result output module: used to output and display the analysis and evaluation results of the stress resistance analysis model.

2. The intelligent spectral image detection system for cotton multi-stress resistance according to claim 1, characterized in that: The spectral range of the spectral camera covers the characteristic absorption spectral range of cotton from visible light to near-infrared bands, and its resolution is not less than 500 pixels.

3. The intelligent spectral image detection system for cotton multi-stress resistance according to claim 1, characterized in that: The data processing unit also includes a data screening module, which is used to screen out effective spectral image data for subsequent processing according to preset cotton growth stages and stress resistance treatment conditions.

4. The intelligent spectral image detection system for cotton multi-stress resistance according to claim 1, characterized in that: The spectral feature information extracted by the feature extraction module includes spectral reflectance features, spectral absorption features, spectral slope features and spectral texture features.

5. The intelligent spectral image detection system for cotton multi-stress resistance according to claim 1, characterized in that: The stress resistance analysis model construction module uses a machine learning algorithm to construct a stress resistance analysis model, and the machine learning algorithm includes but is not limited to a support vector machine, a random forest, and a neural network algorithm.

6. The intelligent spectral image detection system for cotton multi-stress resistance according to claim 1, characterized in that: The detection system also includes a sample expansion module, which is used to expand cotton sample data with known stress resistance through data enhancement technology to improve the accuracy and generalization ability of the stress resistance analysis model.

7. The intelligent spectral image detection method for cotton multi-stress resistance according to claim 1, characterized in that: The detection method comprises the following steps: S1. Using a spectral image acquisition device to collect spectral image data of cotton at different growth stages and under different stress resistance treatments; S2, transmitting the collected spectral image data to the data processing unit through the data transmission module; S3, the data preprocessing module in the data processing unit performs denoising, correction and normalization on the spectral image data; S4, a feature extraction module extracts spectral feature information related to cotton stress resistance from the preprocessed spectral image data; S5, a stress resistance analysis model building module builds a cotton multi-stress resistance analysis model according to the extracted spectral feature information and cotton sample data with known stress resistance; S6. Input the spectral image data of the cotton to be tested into the constructed stress resistance analysis model, and output the stress resistance analysis evaluation results of the cotton through the result output module.

8. The intelligent spectral image detection method for cotton multi-stress resistance according to claim 7, characterized in that: The S4 also includes the steps of performing correlation analysis and feature selection on the spectral feature information to remove redundant and low-correlation features to improve the efficiency and accuracy of model building.

9. The intelligent spectral image detection method for cotton multi-stress resistance according to claim 7, characterized in that: The S5 also includes a step of optimizing the parameters of the machine learning algorithm, and the parameter optimization uses a cross-validation method to determine the optimal parameter combination.

10. The intelligent spectral image detection method for cotton multi-stress resistance according to claim 7, characterized in that: The detection system also includes the step of regularly updating the stress resistance analysis model, adding newly acquired cotton sample data with accurate stress resistance identification results into the model building process to continuously optimize the model performance and adapt to the detection needs of different new cotton germplasms.