A soil compaction degree automatic identification method, system, device and medium

Abstract

The application relates to a soil compaction degree automatic identification method, system, device and medium, and relates to the soil quality inspection technical field, which comprises the following steps: acquiring a soil test sample and images of the soil test sample under different light intensities, and constructing a sample training set and an image training set; through pretreatment of the image training set, soil features under different light conditions are extracted, combined with water content data of the soil in the sample training set, and a compaction degree prediction model is generated by using a machine learning algorithm; the model can predict the compaction degree of the soil according to the features of the on-site soil under specific light conditions; the application can adapt to different light conditions, and thus accurate and reliable soil compaction degrees can be obtained. The application can adapt to different light conditions, and thus accurate and reliable soil compaction degrees can be obtained.

Description

Technical Field

[0001] This application relates to the field of soil quality testing technology, and in particular to an automatic identification method, system, device and medium for soil compaction degree. Background Technology

[0002] Soil compaction is one of the important indicators for assessing the stability of foundations in civil engineering projects such as roads and buildings. During construction, ensuring that the soil achieves sufficient compaction is crucial for the safety and durability of subsequent structures.

[0003] Currently, to address the problems inherent in traditional manual methods for measuring soil compaction, the industry typically employs physical methods, such as soil nuclear density meters and ring cutter methods. These methods primarily estimate soil compaction by directly measuring the soil's mass, volume, and moisture content. In addition, some studies have attempted to utilize chemical and biological methods. While these methods can improve accuracy to some extent, they still cannot solve the problems of complex operation and excessive time consumption.

[0004] However, the existing technologies mentioned above are easily affected by changes in lighting conditions during soil compaction testing at construction sites. Consequently, traditional physical and chemical methods are insufficient to accurately characterize parameters such as soil particle composition, gradation, and moisture content, leading to inaccurate and unreliable soil compaction measurement results. Therefore, there is an urgent need to develop an automated soil compaction identification method that can overcome the shortcomings of existing technologies and adapt to different lighting conditions. Summary of the Invention

[0005] The purpose of this application is to provide an automatic soil compaction identification method that can adapt to different lighting conditions, thereby obtaining accurate and reliable soil compaction.

[0006] Firstly, this application provides an automatic soil compaction degree identification method, which adopts the following technical solution: An automatic soil compaction degree identification method includes: Obtain soil test samples and corresponding images of soil test samples under several different light intensities; divide the soil test samples into a sample training set and an image training set. Several of the image training sets are preprocessed to obtain several soil features based on different lighting conditions; The moisture content of each soil test sample in the training set is obtained through machine learning algorithms; Based on the soil characteristics of each soil test sample under different light intensities and the moisture content of the soil test samples, a compaction prediction model is generated using machine learning algorithms; The soil compaction degree of the soil in the field is obtained through the compaction degree prediction model.

[0007] By adopting the above technical solution, soil samples are visually identified under different light intensities. The obtained soil characteristics and soil moisture content are combined to generate a compaction degree prediction model. The generated compaction degree prediction model can adapt to the actual soil under different light conditions, thereby improving the accuracy of soil compaction degree prediction.

[0008] In a preferred embodiment, this application can be further configured such that: the soil features include particle size, distribution features, and shape features; and the step of preprocessing several of the image training sets to obtain several soil features based on different lighting conditions includes: The image training set is preprocessed, including noise reduction, image enhancement, and normalization. Using a convolutional neural network, particles of different sizes, distribution characteristics, and shape characteristics are identified and classified from each soil test sample in the preprocessed image training set; Based on the particle size, distribution characteristics, and shape characteristics of each soil test sample in the image training set, a soil feature model is established to obtain several soil features based on different lighting conditions.

[0009] By adopting the above technical solutions, the image training set is preprocessed, which improves the quality and consistency of the images and helps with subsequent analysis. The convolutional neural network identifies and classifies the features of the particles, providing data on soil features under different lighting conditions for subsequent compaction degree prediction, which improves the accuracy and reliability of soil feature extraction, thereby enhancing the accuracy of compaction degree prediction.

[0010] In a preferred embodiment, this application can be further configured such that the step of obtaining the moisture content of each soil test sample in the sample training set via a machine learning algorithm includes: Collect spectral data from the training set of samples; Based on the spectral data, the moisture content of each soil test sample in the sample training set is obtained through machine learning, and a moisture content model is established to obtain the moisture content of each soil test sample.

[0011] By adopting the above technical solution, spectral data of the sample training set is collected, and the moisture content is obtained from the spectral data through machine learning. The corresponding moisture content is obtained from a large number of samples in the sample training set, and then a moisture content model is established. Using the moisture content model, the moisture content of the soil samples on site can be quickly calculated based on the obtained spectral data, providing moisture content-based data for subsequent compaction degree prediction.

[0012] In a preferred embodiment, this application can be further configured as follows: the step of generating a compaction prediction model using a machine learning algorithm based on the soil characteristics of each soil test sample under different light intensities and the moisture content of the soil test sample includes: The remaining soil test samples are used to obtain the particle size, distribution characteristics, shape characteristics, and moisture content of the soil test samples through soil characteristic models and moisture content models. Based on the particle size, distribution characteristics, shape characteristics, and moisture content of each soil test sample, and combined with soil physics principles, a machine learning algorithm is used to obtain a compaction degree prediction model under different light conditions.

[0013] By adopting the above technical solution, the remaining sample data of the soil test samples contain features that the soil characteristic model and moisture content model have not encountered in the training phase. This helps the generated compaction prediction model learn a wider range of data distributions, improves its prediction ability, and the soil features and corresponding moisture contents generated by combining soil physics principles are more diverse. This helps the compaction prediction model capture complex and comprehensive data in the soil, thereby improving the accuracy and reliability of soil compaction prediction.

[0014] In a preferred embodiment, this application can be further configured such that the step of obtaining the soil compaction degree at the site using the compaction degree prediction model includes: Acquire images of the soil at the site and the soil itself; The soil image is divided into several test areas based on the light intensity in different regions of the soil image; The soil compaction degree of each test area is obtained by using a compaction degree prediction model to analyze the soil compaction degree of several test areas and the corresponding field soil in those test areas.

[0015] By adopting the above technical solution, soil images and soil samples are obtained from the site. Based on the light intensity of different areas in the soil images, the soil is divided into multiple test areas for detailed analysis. The soil compaction degree is then analyzed in each test area using a compaction degree prediction model. The obtained soil compaction degree can be accurate to test areas with different light intensities.

[0016] In a preferred embodiment, this application can be further configured as follows: the step of dividing the soil image into several test areas based on the light intensity of different regions in the soil image includes: Determine whether there is a sudden change in the light intensity of adjacent test areas; Among them, each method can effectively determine whether there is a sudden change in the light intensity of different regions in the soil image. If a sudden change occurs, the expected light intensity of the region with the sudden change is calculated by linear interpolation based on the light intensity of the adjacent test regions. The brightness of the image in the mutation region is adjusted according to the expected light intensity, and the adjusted image of the mutation region is then updated in the test region.

[0017] In this embodiment, a soil image and the soil itself are acquired at the site, and the soil image is divided into several test areas based on the light intensity of different regions within the soil image. First, it is determined whether there is a sudden change in light intensity between adjacent test areas. If a sudden change occurs, linear interpolation is performed based on the light intensity of adjacent test areas to calculate the expected light intensity of the area with the sudden change. The image brightness and contrast of the area with the sudden change are then adjusted according to the expected light intensity and updated in the test areas. The image of the area with the adjusted light intensity is then updated in the test areas, thereby improving the accuracy of soil compaction prediction.

[0018] In a preferred embodiment, this application may be further configured as follows: after the step of obtaining the soil compaction degree of each test area by using a compaction degree prediction model to obtain the soil compaction degree of the plurality of test areas and corresponding soil samples of the test areas, the application may further include: Acquire soil images, soil samples, and corresponding soil compaction values ​​collected on-site to establish a validation set; The validation set and the corresponding field soil were validated using the compaction degree prediction model, and the parameters of the compaction degree prediction model were optimized.

[0019] By adopting the above technical solution, the validation set and the corresponding field soil samples are validated through the compaction degree prediction model, and the parameters of the compaction degree prediction model are optimized, thereby improving the prediction accuracy of the compaction degree prediction model.

[0020] Secondly, this application provides an automatic soil compaction degree identification system, which adopts the following technical solution: An automatic soil compaction degree identification system includes: Sample acquisition module: used to acquire soil test samples and corresponding images of soil test samples under different light intensities, divide the soil test samples into a sample training set, and divide the soil test samples into an image training set. Soil feature acquisition module: used to preprocess several of the image training sets to acquire several soil features based on different lighting conditions; Moisture content acquisition module: used to obtain the moisture content of each soil test sample in the sample training set through machine learning algorithms; Model generation module: Used to generate a compaction prediction model based on the soil characteristics and water content of each soil test sample under different light intensities using machine learning algorithms; Compaction degree acquisition module: used to acquire the soil compaction degree of the soil in the field through the compaction degree prediction model.

[0021] Thirdly, this application provides an electronic device that adopts the following technical solution: An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described automatic soil compaction degree identification method.

[0022] Fourthly, this application provides a computer storage medium, as follows: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described automatic soil compaction degree identification method.

[0023] In summary, this application has the following beneficial technical effects: This application is able to acquire soil test samples and their images under different light intensities, obtain soil characteristics under different light conditions from the soil test samples, obtain the moisture content from the soil test samples, and generate a soil compaction prediction model that can adapt to actual testing under different light conditions based on soil characteristics and moisture content, thereby realizing the effective extraction of soil characteristics and accurate prediction of soil compaction in the ever-changing environment of the field. Attached Figure Description

[0024] Figure 1 This is a flowchart of an automatic soil compaction degree identification method in one embodiment of this application.

[0025] Figure 2 This is a flowchart of a sub-step of step S2 in one embodiment of this application.

[0026] Figure 3 This is a flowchart of a sub-step of step S3 in one embodiment of this application.

[0027] Figure 4 This is a flowchart of a sub-step of step S4 in one embodiment of this application.

[0028] Figure 5 This is a flowchart of a sub-step of step S5 in one embodiment of this application.

[0029] Figure 6 This is a flowchart of a sub-step of step S51 in one embodiment of this application.

[0030] Figure 7 This is a flowchart of the steps added after step S52 in one embodiment of this application.

[0031] Figure 8 This is a schematic diagram of the structure of an automatic soil compaction degree identification system according to one embodiment of this application.

[0032] Figure 9 This is a schematic block diagram of an electronic device in one embodiment of this application.

[0033] Figure labels: 1. Sample acquisition module; 2. Soil characteristic acquisition module; 3. Moisture content acquisition module; 4. Model generation module; 5. Compaction degree acquisition module. Detailed Implementation

[0034] The following is in conjunction with the appendix Figure 1-9 This application will be described in further detail.

[0035] refer to Figure 1 An automatic soil compaction degree identification method, specifically including: S1. Obtain soil test samples and corresponding images of soil test samples under different light intensities. Divide the soil test samples into a sample training set and an image training set.

[0036] In this process, multiple types of soil test samples should be selected for testing to improve the versatility of the subsequently acquired soil characteristic models. Appropriate lighting systems should be used to illuminate the soil test samples, ensuring that the test surfaces are evenly and adequately illuminated. Using a high-resolution camera to capture soil test samples under different light intensities can yield clearer soil characteristics.

[0037] S2. Preprocess several image training sets to obtain several soil features based on different lighting conditions.

[0038] Among these features, preprocessed soil images can improve the accuracy of subsequent analysis of particle characteristics in soil features.

[0039] S3. Obtain the moisture content of each soil test sample in the sample training set through machine learning algorithms.

[0040] In one embodiment, the machine learning algorithm can select a random forest to calculate the moisture content of each soil test sample in the training set. In another embodiment, the machine learning algorithm can select a gradient boosting machine to calculate the moisture content of each soil test sample in the training set.

[0041] S4. Based on the soil characteristics of each soil test sample under different light intensities and the moisture content of the soil test sample, a compaction prediction model is generated using a machine learning algorithm.

[0042] In one embodiment, the machine learning algorithm can select a random forest to calculate the soil compaction degree for each soil test sample in the training set. In another embodiment, the machine learning algorithm can select a gradient boosting machine to calculate the soil compaction degree for each soil test sample in the training set.

[0043] S5. Obtain the soil compaction degree of the field soil through the compaction degree prediction model.

[0044] In this embodiment, visual recognition is performed on soil test samples to obtain accurate and reliable soil test samples of different types and under different light intensities, and images of soil test samples under each light intensity are captured. A sample training set is then created from the soil test samples to obtain moisture content using machine learning algorithms. Similarly, an image training set is created from the soil test sample images, which are preprocessed and analyzed to obtain soil characteristics. Finally, a compaction degree prediction model is generated based on the soil characteristics under different light conditions and the corresponding moisture content of the soil test samples. By using a compaction degree prediction model that can calculate soil compaction degree in the field under different light conditions, the calculation error caused by the influence of light can be reduced.

[0045] refer to Figure 2 Furthermore, in one embodiment, the soil characteristics include particle size, distribution characteristics, and shape characteristics, and step S2 is refined into the following sub-steps: S20. Preprocess the image training set, including noise reduction, image enhancement, and standardization.

[0046] Denoising the image training set reduces noise interference with soil feature analysis and improves the quality of soil test sample images. Enhancing features in the image training set makes subsequent identification of soil particle features more accurate. Converting the soil test sample images in the image training set to a uniform format or range minimizes scale differences between images. Denoising can be performed using algorithms such as Gaussian filtering or median filtering. Image enhancement can be achieved by adjusting contrast. Standardization is achieved through data normalization to ensure consistency of image data. Then, abnormal data points in the sample training set, such as outliers or obviously erroneous data, are cleaned. Through data cleaning, the accuracy and reliability of soil test samples can be improved.

[0047] S21. Using a convolutional neural network, identify and classify particles of different sizes, distribution characteristics, and shape characteristics from each soil test sample in the preprocessed image training set.

[0048] In this process, for soil particles, convolution operations are used to extract local features in the image training set of soil test samples. Convolution kernels are designed to identify particle edges, thereby helping to identify particle size and shape. Since the edges of soil particles usually have a large gray-level difference from the background, the ReLU activation function introduces non-linearity and learns a complex function mapping to distinguish particles from the background. Pooling operations are then performed, where max pooling helps to retain the largest features of the particles, while average pooling can smooth noise and improve the model's generalization ability. Finally, the gradient of the loss function with respect to the network parameters is calculated using the backpropagation algorithm, and the parameters are updated using the gradient descent method. In this embodiment, the loss function can be designed as the classification error. By optimizing the error, particles can be identified and classified more accurately.

[0049] S22. Based on the particle size, distribution characteristics, and shape characteristics of each soil test sample in the image training set, establish a soil feature model to obtain several soil features based on different lighting conditions.

[0050] In this process, a portion of the soil test sample images are further divided into an image test set. The image test set is used to test and optimize the soil feature model generated based on the image training set. The model parameters are optimized through multiple iterations to improve the accuracy of the soil feature model prediction.

[0051] In this embodiment, preprocessing the soil test sample images in the image training set can improve the quality and consistency of the images, which helps the subsequent convolutional neural network to analyze the soil test sample images of each type and under different lighting conditions. The convolutional neural network identifies and classifies the features of particles, providing the size, distribution characteristics, and shape characteristics of particles in the soil test samples under different lighting conditions for subsequent compaction degree prediction, thereby improving the accuracy and reliability of soil feature extraction and enhancing the accuracy of compaction degree prediction.

[0052] In addition, refer to Figure 3 Furthermore, in one embodiment, step S3 is refined into the following sub-steps: S30. Collect spectral data of the sample training set.

[0053] In this embodiment, a near-infrared spectrometer is used to detect each training set in the sample training set, so that the samples cover a wide range of soil moisture content, so that the moisture content model can accurately predict different moisture content levels.

[0054] S31. Based on the spectral data, obtain the moisture content of each soil test sample in the sample training set through machine learning, establish a moisture content model, and obtain the moisture content of each soil test sample.

[0055] The process involves preprocessing the spectral data, such as smoothing, normalization, and scattering correction, to reduce noise and systematic errors and improve the predictive performance of the moisture content model. Next, feature wavelengths related to moisture content are extracted from the original spectral data. Then, a mathematical model between the spectral data and moisture content, i.e., the moisture content model, is established using machine learning algorithms such as random forest and gradient booster.

[0056] In addition, a portion of the soil test samples were divided into a sample test set. The sample test set was used to test and optimize the moisture content model generated based on the sample training set. The model parameters were optimized through multiple iterations to improve the accuracy of the moisture content model prediction.

[0057] In this embodiment, a sample training set is constructed by collecting spectral data from soil test samples. Machine learning algorithms are used to analyze and calculate the correlation between spectral data features and soil moisture content. Based on the moisture content data of a large number of test samples, a moisture content model capable of accurately predicting soil moisture content is constructed. Using this moisture content model, the soil moisture content can be predicted after obtaining the spectral data of the soil in the field, providing moisture content-based data for subsequent compaction degree prediction.

[0058] In addition, refer to Figure 4 Furthermore, in one embodiment, step S4 is refined into the following sub-steps: S40. From the remaining soil test samples, obtain the particle size, distribution characteristics, shape characteristics, and moisture content of the soil test samples through the soil characteristic model and moisture content model.

[0059] S41. Based on the particle size, distribution characteristics, shape characteristics, and moisture content of each soil test sample, and combined with the principles of soil physics, a machine learning algorithm is used to obtain a compaction degree prediction model under different light conditions.

[0060] Soil is composed of particles of different sizes, including sand, silt, and clay. The size and distribution of these particles, as well as soil moisture, can affect the compactness of soil structure and soil plasticity.

[0061] Moisture can fill the gaps between soil particles, increasing the soil's cohesion and thus affecting its compaction.

[0062] Soil porosity reflects the proportion of void volume to total volume in soil. Porosity is closely related to soil particle size, distribution, and moisture content. The size of porosity directly affects soil compaction; the smaller the porosity, the denser the soil and the higher the compaction.

[0063] Soil particle size, distribution characteristics, shape characteristics, and moisture content also affect soil density. Soil density can be divided into dry density and wet density. Dry density refers to the mass of the solid portion of soil per unit volume, while wet density includes the mass of both soil solids and water. An increase in soil density usually means an increase in soil compaction.

[0064] Knowing the size, distribution characteristics, shape characteristics, and moisture content of soil particles in the soil test sample, and combining this with soil physics, we start from soil particle structure, soil plasticity, soil moisture content, soil porosity, and soil density, and select machine learning algorithms such as random forest and gradient booster to generate a compaction rate prediction model based on soil characteristics and moisture content.

[0065] In this embodiment, the remaining samples of the soil test samples contain data that were not used in the soil characteristic model and moisture content model during the training phase, which helps the compaction prediction model to have a wide distribution of learning data, thereby enhancing the prediction performance of the compaction prediction model.

[0066] In addition, refer to Figure 5 Furthermore, in one embodiment, step S5 is refined into the following sub-steps: S50: Acquire images of the soil on site and the soil itself.

[0067] S51. Based on the light intensity in different areas of the soil image, divide the soil image into several test areas.

[0068] In this process, several soil images are converted to grayscale. By detecting the grayscale values ​​of the soil images, each soil image is divided into test areas with different grayscale values.

[0069] S52. Using a compaction prediction model, obtain the soil compaction degree of each test area by combining several test areas and the corresponding field soil in the test areas.

[0070] In this embodiment, a compaction degree prediction model is used to predict the compaction degree of soil images and soil samples from the field. This includes predicting and analyzing the soil characteristics of several test areas with different light intensities, as well as predicting and analyzing the moisture content of soil samples. After obtaining the predicted soil characteristic data and moisture content data, the data is used to perform a detailed analysis of the field soil under different light conditions using the compaction degree prediction model.

[0071] In addition, refer to Figure 6 Furthermore, in one embodiment, step S51 is refined into the following sub-steps: S510. Determine whether there is a sudden change in the light intensity of adjacent test areas.

[0072] To mitigate the risk of low-brightness soil images due to obstructions during filming, which could lead to inaccurate predictions of soil characteristics, the captured grayscale soil images are divided into regions before obtaining the grayscale value for each test region. Based on a preset grayscale value change threshold, a sudden change is considered to have occurred when the grayscale value change between adjacent test regions exceeds this threshold.

[0073] S511. If a sudden change occurs, the expected light intensity of the sudden change area is calculated by linear interpolation based on the light intensity of the adjacent test areas.

[0074] In this process, other test regions connected to the abruptly changed test region are selected as reference regions for linear interpolation calculation, and the grayscale values ​​of the regions connected to the abruptly changed region are recorded. Then, a linear interpolation formula is applied to each pixel within the abruptly changed region to calculate its expected grayscale value. The calculated expected grayscale value is then filled into the corresponding pixel, and the interpolation result is checked to see if it is reasonable, i.e., whether the grayscale value of the abruptly changed test region transitions smoothly.

[0075] S512. Adjust the brightness of the image of the mutation area according to the expected light intensity, and update the adjusted image of the mutation area to the test area.

[0076] The process involves restoring the grayscale values ​​of test areas with abrupt changes to the expected values, recording the amount of grayscale change, and then using a smoothing function or gradient domain fusion technique to gradually transition the adjusted grayscale values ​​to the grayscale values ​​of surrounding pixels, thus adjusting the brightness. Furthermore, during brightness adjustment, it is crucial to ensure that the pixel grayscale values ​​remain within a reasonable range; grayscale values ​​exceeding this range require cropping or saturation processing to minimize image distortion.

[0077] In this embodiment, by determining whether there is a sudden change in the light intensity in different areas of the soil image, the image of the abrupt change area after adjusting the light intensity is then updated to the test area, so as to improve the accuracy of the compaction prediction model in predicting the compaction degree of the soil in the field.

[0078] In addition, refer to Figure 7 Furthermore, in one embodiment, after step S52, steps S520 and S521 are added: S520. Acquire soil images collected on-site, on-site soil and corresponding soil compaction degree, and establish a validation set.

[0079] S521. Validate the validation set and the corresponding field soil using the compaction degree prediction model, and optimize the compaction degree prediction model parameters.

[0080] In this embodiment, feedback from the validation set can identify which soil characteristics or conditions the compaction degree prediction model performs poorly, allowing for further adjustment of the model's parameters to improve prediction accuracy. The optimized compaction degree prediction model can better handle noise, outliers, and variations in data distribution, enhancing its robustness in the face of uncertainty and complex environments.

[0081] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0082] This application also provides an automatic soil compaction degree identification system, which corresponds one-to-one with the automatic soil compaction degree identification method in the embodiments.

[0083] refer to Figure 8 An automatic soil compaction degree identification system includes: a sample acquisition module 1, a soil feature acquisition module 2, a moisture content acquisition module 3, a model generation module 4, and a compaction degree acquisition module 5. Detailed descriptions of each functional module are as follows: Sample acquisition module 1: used to acquire soil test samples and corresponding images of soil test samples under several different light intensities, divide the soil test samples into a sample training set, and divide the soil test samples into an image training set. Soil feature acquisition module 2: used to preprocess several image training sets to acquire several soil features based on different lighting conditions; Moisture content acquisition module 3: Used to obtain the moisture content of each soil test sample in the sample training set through machine learning algorithms; Model generation module 4: Used to generate a compaction prediction model based on the soil characteristics and water content of each soil test sample under different light intensities using machine learning algorithms. Compaction degree acquisition module 5: Used to obtain the soil compaction degree of the soil on site through the compaction degree prediction model.

[0084] The system comprises several modules: Sample Acquisition Module 1 acquires soil test samples and their corresponding soil test images under different light intensities; Soil Feature Acquisition Module 2 preprocesses the image training set and extracts soil features under different light conditions; Moisture Content Acquisition Module 3 uses machine learning algorithms to acquire the moisture content of each soil test sample; Model Generation Module 4 combines the soil features acquired by Soil Feature Acquisition Module 2 and the moisture content acquired by Moisture Content Acquisition Module 3 to generate a compaction degree prediction model applicable to different light environments; and finally, Compaction Degree Acquisition Module 5 uses the compaction degree prediction model to obtain the compaction degree of the soil on site. Through the combined use of these modules, efficient and accurate real-time identification of soil compaction degree at the construction site under different light conditions is achieved.

[0085] Specific limitations regarding the automatic soil compaction degree identification system can be found in the context of the automatic soil compaction degree identification method, and will not be repeated here. Each module in the aforementioned automatic soil compaction degree identification system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in an electronic device, or stored in the memory of the electronic device as software, so that the processor can call and execute the operations corresponding to each module. In one embodiment, an electronic device is provided, which is a user terminal. (Reference) Figure 9 The electronic device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores test data tables. The network interface allows communication with external terminals via a network connection. When the processor executes the computer program, it implements an automatic soil compaction degree identification method.

[0086] In one embodiment, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: S1. Obtain soil test samples and corresponding images of soil test samples under different light intensities. Divide the soil test samples into a sample training set and an image training set.

[0087] S2. Preprocess several image training sets to obtain several soil features based on different lighting conditions.

[0088] S3. Obtain the moisture content of each soil test sample in the sample training set through machine learning algorithms.

[0089] S4. Based on the soil characteristics of each soil test sample under different light intensities and the moisture content of the soil test sample, a compaction prediction model is generated using a machine learning algorithm.

[0090] S5. Obtain the soil compaction degree of the field soil through the compaction degree prediction model.

[0091] In one embodiment, soil characteristics include particle size, distribution characteristics, and shape characteristics, and the sub-steps of refining step S2 include: S20. Preprocess the image training set, including noise reduction, image enhancement, and standardization.

[0092] S21. Using a convolutional neural network, identify and classify particles of different sizes, distribution characteristics, and shape characteristics from each soil test sample in the preprocessed image training set.

[0093] S22. Based on the particle size, distribution characteristics, and shape characteristics of each soil test sample in the image training set, establish a soil feature model to obtain several soil features based on different lighting conditions.

[0094] In one embodiment, the sub-steps of step S3 refinement include: S30. Collect spectral data of the sample training set.

[0095] S31. Based on the spectral data, obtain the moisture content of each soil test sample in the sample training set through machine learning, establish a moisture content model, and obtain the moisture content of each soil test sample.

[0096] In one embodiment, the sub-steps of step S4 refinement include: S40. From the remaining soil test samples, obtain the particle size, distribution characteristics, shape characteristics, and moisture content of the soil test samples through the soil characteristic model and moisture content model.

[0097] S41. Based on the particle size, distribution characteristics, shape characteristics, and moisture content of each soil test sample, and combined with the principles of soil physics, a machine learning algorithm is used to obtain a compaction degree prediction model under different light conditions.

[0098] In one embodiment, the sub-steps of step S5 refinement include: S50: Acquire images of the soil on site and the soil itself.

[0099] S51. Based on the light intensity in different areas of the soil image, divide the soil image into several test areas.

[0100] S52. Using a compaction prediction model, obtain the soil compaction degree of each test area by combining several test areas and the corresponding field soil in the test areas.

[0101] In one embodiment, the sub-steps of step S51 include: S510. Determine whether there is a sudden change in the light intensity of adjacent test areas.

[0102] S511. If a sudden change occurs, the expected light intensity of the sudden change area is calculated by linear interpolation based on the light intensity of the adjacent test areas.

[0103] S512. Adjust the brightness of the image of the mutation area according to the expected light intensity, and update the adjusted image of the mutation area to the test area.

[0104] In one embodiment, the additional steps following step S52 include: S520. Acquire soil images collected on-site, on-site soil and corresponding soil compaction degree, and establish a validation set.

[0105] S521. Validate the validation set and the corresponding field soil using the compaction degree prediction model, and optimize the compaction degree prediction model parameters.

[0106] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0107] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

Claims

1. A method for automatically identifying soil compaction degree, characterized in that, include: Obtain soil test samples and corresponding images of soil test samples under several different light intensities; divide the soil test samples into a sample training set and an image training set. Several of the image training sets are preprocessed to obtain several soil features based on different lighting conditions; The moisture content of each soil test sample in the training set is obtained through machine learning algorithms; Based on the soil characteristics of each soil test sample under different light intensities and the moisture content of the soil test samples, a compaction prediction model is generated using machine learning algorithms; The soil compaction degree of the soil in the field is obtained through the compaction degree prediction model.

2. The method according to claim 1, characterized in that, The soil features include particle size, distribution characteristics, and shape characteristics. The step of preprocessing several training sets of images to obtain several soil features based on different lighting conditions includes: The image training set is preprocessed, including noise reduction, image enhancement, and normalization. Using a convolutional neural network, particles of different sizes, distribution characteristics, and shape characteristics are identified and classified from each soil test sample in the preprocessed image training set; Based on the particle size, distribution characteristics, and shape characteristics of each soil test sample in the image training set, a soil feature model is established to obtain several soil features based on different lighting conditions.

3. The method according to claim 2, characterized in that, The step of obtaining the moisture content of each soil test sample in the sample training set through a machine learning algorithm includes: Collect spectral data from the training set of samples; Based on the spectral data, the moisture content of each soil test sample in the sample training set is obtained through machine learning, and a moisture content model is established to obtain the moisture content of each soil test sample.

4. The method according to claim 3, characterized in that, The step of generating a compaction prediction model using a machine learning algorithm based on the soil characteristics of each soil test sample under different light intensities and the water content of the soil test samples includes: The remaining soil test samples are used to obtain the particle size, distribution characteristics, shape characteristics, and moisture content of the soil test samples through soil characteristic models and moisture content models. Based on the particle size, distribution characteristics, shape characteristics, and moisture content of each soil test sample, and combined with soil physics principles, a machine learning algorithm is used to obtain a compaction degree prediction model under different light conditions.

5. The method according to claim 1, characterized in that, The step of obtaining the soil compaction degree on site using the compaction degree prediction model includes: Acquire images of the soil at the site and the soil itself; The soil image is divided into several test areas based on the light intensity in different regions of the soil image; The soil compaction degree of each test area is obtained by using a compaction degree prediction model to analyze the soil compaction degree of several test areas and the corresponding field soil in those test areas.

6. The method according to claim 5, characterized in that, The step of dividing the soil image into several test areas based on the light intensity of different regions in the soil image includes: Determine whether there is a sudden change in the light intensity of adjacent test areas; If a sudden change occurs, the expected light intensity of the area of ​​the sudden change is calculated by linear interpolation based on the light intensity of the adjacent test areas. The brightness of the image in the mutation region is adjusted according to the expected light intensity, and the adjusted image of the mutation region is then updated in the test region.

7. The method according to claim 1, characterized in that, After the step of obtaining the soil compaction degree of each test area by using a compaction degree prediction model to obtain the soil compaction degree of several test areas and corresponding soil samples from the test areas, the method further includes: Acquire soil images, soil samples, and corresponding soil compaction values ​​collected on-site to establish a validation set; The validation set and the corresponding field soil were validated using the compaction degree prediction model, and the parameters of the compaction degree prediction model were optimized.

8. An automatic soil compaction degree identification system, characterized in that, include: Sample acquisition module 1: used to acquire soil test samples and corresponding images of soil test samples under different light intensities, divide the soil test samples into a sample training set, and divide the soil test samples into an image training set. Soil feature acquisition module 2: used to preprocess several of the image training sets to acquire several soil features based on different lighting conditions; Moisture content acquisition module 3: Used to obtain the moisture content of each soil test sample in the sample training set through machine learning algorithms; Model generation module 4: Used to generate a compaction prediction model based on the soil characteristics and water content of each soil test sample under different light intensities using machine learning algorithms. Compaction degree acquisition module 5: used to acquire the soil compaction degree of the soil on site through the compaction degree prediction model.

9. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as any one of the automatic soil compaction degree identification methods as claimed in claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer program is stored and can be loaded by a processor and executed as any one of the automatic soil compaction degree identification methods as described in claims 1 to 7.

Images