A method and system for estimating bearing capacity of block-filled foundation based on image recognition
Through image recognition technology and full neural network model, combined with mechanical curves, the convenience of traditional foundation bearing capacity determination in artificial island and reef scenarios is solved, and fast and convenient foundation bearing capacity estimation is achieved, which is suitable for marine and terrestrial environments.
Patent Information
- Application Number
- CN202210411657.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-04-19
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-04-19
AI Technical Summary
Traditional foundation bearing capacity measurement methods are difficult to achieve convenience and efficiency in artificial islands and reefs, especially the inconvenient use of on-site measurement instruments, which cannot easily obtain foundation bearing capacity.
Using image recognition-based methods, image recognition and area calculation are performed through computers, and the foundation bearing capacity is estimated by combining mechanical curves, including single-particle mechanical strength test, image annotation, image segmentation, area processing and foundation bearing capacity calculation, using full nerves The U-Net model of the network performs image segmentation and feature extraction.
It realizes the rapid and convenient estimation of foundation bearing capacity without complex measurement processes and equipment, and is suitable for marine environments and land foundations, simplifying workflows and improving efficiency.
Smart Images

Figure CN114972186B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the fields of engineering technology and computer technology, and in particular to a method and system for estimating the bearing capacity of a block-filled foundation based on image recognition, which are used for estimating the bearing capacity of the foundation. Background Art
[0002] With the implementation of the marine strategy, more and more artificial islands and reefs have emerged. Many buildings need to be built on the islands and reefs, but the natural foundation cannot meet the bearing capacity requirements, so the natural foundation needs to be treated. The traditional foundation treatment method is to transport soil and rocks from other places far away from the project site to improve the foundation. This method has high transportation costs and a long cycle. There are even more transportation restrictions for offshore buildings. Therefore, it is more convenient to use blocks such as the seabed or surrounding reefs for foundation treatment. For the treated artificial foundation, it is necessary to have a preliminary prediction of its bearing capacity. At present, the calculation method of the bearing capacity of the rock foundation is still to use the traditional on-site measurement or the laboratory to use a uniaxial compressor to measure the saturated uniaxial compressive strength of the rock f γk Then, based on the integrity of the rock mass and the spacing, width, occurrence, and combination of the structural planes, the reduction factor ψ is determined based on regional experience. r , and then according to the characteristic value of rock foundation bearing capacity f a =ψ r ×f γk The bearing capacity of the rock foundation is calculated. The on-site measurement method requires carrying a large-scale foundation bearing capacity measurement instrument. For artificially filled islands and reefs, the on-site measurement instrument is not convenient to use and the foundation bearing capacity cannot be obtained more conveniently. Summary of the invention
[0003] In view of the above-mentioned problems, the present invention proposes a method and system for estimating the bearing capacity of a block-filled foundation based on image recognition, which obtains the foundation bearing capacity by performing image recognition and area calculation on a computer and combining mechanical curves.
[0004] To achieve the above purpose, the present invention is designed to estimate the bearing capacity of block-filled foundation based on image recognition and its system, which is special in that the method includes:
[0005] Step 1: Conduct a single particle mechanical strength test on the test soil and rock material category, draw the probability distribution curve of the compressive strength σ of the test soil and rock material category, and obtain the representative value of the particle compressive strength σ c ;
[0006] Step 2: Take a picture of the test soil and rock stockpile area, annotate the soil and rock particles in the picture, and outline the contours of the soil and rock particles;
[0007] Step 3: Use the U-Net model based on the full neural network to segment the labeled image;
[0008] Step 4: Perform area processing on the segmented image and calculate the area fraction of large particles
[0009] Step 5: Based on the area fraction of large particles Calculate the void fraction Calculate the corresponding force transmission coefficient η according to the pore fraction w of the image;
[0010] Step 6: According to the force transmission coefficient η and safety factor k, and compressive strength σ c The foundation bearing capacity of the test soil and rock pile area is calculated as f = η·σ c / k.
[0011] Preferably, the specific steps of step 1 are: preparing several test soil and stone material standard samples, and testing their uniaxial compressive strength respectively; processing the uniaxial compressive strength data, calculating the arithmetic mean standard deviation, setting the confidence level, obtaining the confidence interval, screening out the data within the confidence interval, drawing the probability distribution curve of the compressive strength σ of the test soil and stone material category, calculating the arithmetic mean of the data within the confidence interval, and obtaining the representative value of the particle compressive strength σ c .
[0012] Preferably, the specific steps of step 3 are: downsampling the labeled image five times, retaining the feature image in each downsampling process, and generating a feature image after four downsamplings; then deconvolving the feature image, combining and connecting the deconvolution result of each time with the convolution feature image of the corresponding layer, and then convolving the combined image to extract the combined features, that is, completing a cross connection; finally, feature classification is performed on the output feature image after five cross connections, that is, the segmentation process of the block filling foundation soil and stone particle image is completed.
[0013] Preferably, the large particle area fraction The calculation formula is Where n is the total number of large particles; S is the total area of the image; a and b are the major and minor axes of the circumscribed ellipse corresponding to the contour of the particle's maximum projection surface.
[0014] Preferably, in step 5, the corresponding relationship between the porosity fraction w and the force transmission coefficient η is established through indoor tests and numerical simulations: several groups of test soil and stone material force transmission coefficient tests are carried out, and the relationship function g(w) between the porosity fraction w and the force transmission coefficient η of the test soil and stone material is obtained by a linear regression method, and the force transmission coefficient η is obtained by substituting the porosity fraction w of the image into the g(w) function.
[0015] Preferably, the safety factor k is in the range of 1.2 to 1.5.
[0016] The present invention also proposes a block filling foundation bearing capacity estimation system based on image recognition, which is special in that the system includes a compressive strength calculation module, an image segmentation module, an area processing module, and a foundation bearing capacity calculation module;
[0017] The compressive strength calculation module is used to process the uniaxial compressive strength test data of the tested soil and stone materials and calculate the representative value σ of the particle compressive strength of the tested soil and stone materials. c ;
[0018] The image segmentation module is used to segment the image of soil and stone particles in the block filling foundation. First, the image is downsampled five times. In each downsampling process of the image, 2 3×3 convolutions and 1 2×2 maximum pooling are performed, and the feature image is retained in each downsampling process; after four downsamplings, the feature image is generated; then the feature image is deconvolved, and each deconvolution result is combined and connected with the convolution feature image of the corresponding layer, and then the combined image is convolved to extract the combined features, that is, a cross connection is completed; finally, the feature image output after five cross connections is feature classified, that is, the segmentation process of the block filling foundation soil and stone particle image is completed;
[0019] The area processing module is used to perform area processing on the segmented image and calculate the area fraction of large particles.
[0020] The foundation bearing capacity calculation module is used to calculate the bearing capacity of the foundation according to the force transmission coefficient η and the safety factor k, as well as the compressive strength σ c Calculate the foundation bearing capacity f of the test soil and rock pile area.
[0021] The present invention further proposes a computer-readable storage medium storing a computer program, wherein the computer program implements the above method when executed by a processor.
[0022] By adopting the above technical solution, the positive effects and advantages of the present invention are:
[0023] (1) Using image recognition technology, a multi-particle compressive strength database is established. Compared with the traditional foundation bearing capacity estimation method, the foundation bearing capacity is obtained by computer image recognition, area calculation and mechanical curve combination.
[0024] (2) The method provided by the present invention is simple and convenient, does not require complicated measurement procedures and equipment, reduces the cost of on-site or indoor tests, and is suitable for estimating the bearing capacity of foundations in marine environments, and is also suitable for estimating the bearing capacity of land foundations.
[0025] (3) The method proposed in the present invention can estimate the bearing capacity of the foundation in the photographed area by taking pictures to obtain images, without the need for in-situ tests or sampling for indoor simulation tests. This eliminates the cumbersome test procedures, greatly simplifies the workflow, and greatly improves efficiency compared to traditional test methods. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] Figure 1 The present invention is a flow chart of a method for estimating the bearing capacity of a block-filled foundation based on image recognition.
[0027] Figure 2 It is the probability density distribution diagram of multi-particle compressive strength.
[0028] Figure 3 This is the original schematic diagram of the block foundation.
[0029] Figure 4 Schematic diagram for labeling bulk particles.
[0030] Figure 5 Schematic diagram of the ground-based image segmentation process for filling in the U-Net model block.
[0031] Figure 6 Schematic diagram of block particle segmentation.
[0032] Figure 7 This is the relationship diagram between the force transmission coefficient and the porosity fraction.
[0033] Figure 8 Figure 2 is a graph of porosity fraction and force transmission coefficient for a granite example. DETAILED DESCRIPTION
[0034] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.
[0035] like Figure 1 As shown, the implementation steps of a method for estimating the bearing capacity of a block filling foundation based on image recognition proposed by the present invention are as follows:
[0036] Step 1: Conduct a single particle mechanical strength test on the test soil and rock material category, draw the probability distribution curve of the compressive strength σ of the test soil and rock material category, and obtain the representative value of the particle compressive strength σ c ;
[0037] Several standard samples of soil and rock materials are prepared for testing, and their uniaxial compressive strength is tested respectively; the uniaxial compressive strength data is processed, the arithmetic mean standard deviation is calculated, the confidence level is set, the confidence interval is obtained, the data within the confidence interval is screened out, and the probability distribution curve of the compressive strength σ of the soil and rock material category is plotted, as shown in Fig. Figure 2 As shown, the arithmetic mean of the data within the confidence interval is calculated to obtain the representative value of the particle compressive strength σ c .
[0038] Step 2: Take a picture of the test soil and rock pile area, and use LabelMe to annotate the soil and rock particles in the picture to outline the contours of the soil and rock particles, such as Figure 3 , Figure 4 As shown;
[0039] Step 3: Use the U-Net model based on the full neural network to segment the annotated image. First, the image is downsampled five times. During each downsampling process, 2 3×3 convolutions and 1 2×2 maximum pooling are performed. The feature image is retained in each downsampling process. After four downsamplings, the feature image is generated; then the feature image is deconvolved, and each deconvolution result is combined with the convolution feature image of the corresponding layer, and then the combined image is convolved to extract the combined features, completing a cross connection; finally, the feature image output after five cross connections is feature classified, completing the segmentation process of the block filling foundation soil and stone particle image, as shown in Figure 1. Figure 5 , Figure 6 shown. Figure 5 In the figure, step ① indicates a 3×3 convolution; step ② indicates a 2×2 pooling; step ③ indicates a deconvolution; step ④ indicates an image combination; and step ⑤ indicates a cross connection. Figure 6 This is the block particle segmentation map output by the U-Net model.
[0040] Step 4: Perform area processing on the segmented image and calculate the area fraction of large particles
[0041] Assuming that each segmentation unit is approximately elliptical, calculate the area fraction of large particles (particle size greater than or equal to 4.75 mm) Where n represents the total number of large particles; S represents the total image area; a and b represent the major and minor axes of the circumscribed ellipse corresponding to the contour of the particle's maximum projection surface.
[0042] Step 5: Based on the area fraction of large particles Calculate the void fraction Calculate the corresponding force transmission coefficient η according to the pore fraction w of the image;
[0043] The corresponding relationship between the porosity fraction w and the force transmission coefficient η is established through indoor tests and numerical simulations: several groups of test soil and rock material force transmission coefficient tests are carried out, and the relationship function g(w) between the porosity fraction w and the force transmission coefficient η of the test soil and rock material is obtained through the linear regression method, such as Figure 7 As shown, the transmission coefficient η is obtained by substituting the pore fraction w of the image into the g(w) function.
[0044] Step 6: According to the force transmission coefficient η and safety factor k, and compressive strength σc The foundation bearing capacity of the test soil and rock pile area is calculated as f = η·σ c / k.
[0045] This embodiment uses granite block foundation as the test soil and stone material type, and the specific implementation process is as follows:
[0046] a. Make 100 granite standard specimens and test their uniaxial compressive strength respectively. The data obtained are as follows (unit: MPa):
[0047] 106.9 103.0 94.7 97.9 99.0 118.7 107.8 80.2 88.2 117.1 108.0 102.2 88.6 86.6 100.4 107.8 101.8 106.8 86.2 84.4 102.4 89.5 117.1 107.7 100.0 107.8 83.0 95.0 104.5 111.0 91.2 90.3 80.0 107.0 99.3 114.1 116.5 113.6 83.5 86.1 98.6 116.0 120.1 116.2 96.7 114.0 117.4 103.8 100.2 90.5 88.9 109.9 111.5 110.6 89.4 97.6 116.6 120.6 118.3 95.4 86.2 82.4 97.0 90.7 116.3 114.6 117.4 89.8 90.0 95.5 117.1 92.7 87.1 104.6 104.3 87.6 80.7 99.4 98.1 120.4 107.6 119.9 87.1 108.4 91.0 108.8 98.3 84.9 91.5 102.8 86.2 111.9 110.6 102.9 84.0 92.1 108.8 84.4 86.0 100.7
[0048] After processing this set of data, the arithmetic mean is 100.6, the standard deviation is 11.69153, the significance level is set to 10%, that is, the confidence level is 90%, and the confidence interval is (98.7, 102.5). The data in the above table that are greater than or equal to 98.7 and less than or equal to 102.5 are screened out as shown in the following table (unit: MPa):
[0049] 102.4 102.2 99.0 100.4 100.0 99.3 101.8 99.4 100.2 100.7
[0050] Calculate the arithmetic mean of the above data and obtain the representative value of compressive strength σc as 100.5;
[0051] b. Process the captured image through the above steps 2), 3), and 4) to obtain the area score The porosity is 91%.
[0052] c. Test data of pore fraction and force transmission coefficient (as shown in the following table)
[0053] Porosity fraction Transmission coefficient 0 1 5% 0.93 8% 0.89 12% 0.84 17% 0.80 20% 0.77 23% 0.74 26% 0.72 28% 0.70 31% 0.68
[0054] like Figure 8 As shown in the figure, the linear regression method can be used to obtain the formula of the porosity w and the force transmission coefficient η of the granite block foundation:
[0055] η=0.0108×(10w) 2 -1.378w+1.002
[0056] Since the porosity fraction w = 9% is not included in the test data, substituting it into the above formula, we can obtain the force transmission coefficient η = 0.89;
[0057] d. According to the bearing capacity of the foundation f = η·σ c / k, η=0.89,σ c Substituting 100.5 into the previous formula and taking k = 1.2, we can obtain that the bearing capacity of the granite block foundation where the photo was taken is 74.5MPa.
[0058] The present invention also proposes a block filling foundation bearing capacity estimation system based on image recognition, which is deployed in a computer and is used to implement a block filling foundation bearing capacity estimation method based on image recognition. The system includes a compressive strength calculation module, an image segmentation module, an area processing module, and a foundation bearing capacity calculation module;
[0059] Compressive strength calculation module: used to process the uniaxial compressive strength test data of the tested soil and stone materials, and calculate the representative value σ of the particle compressive strength of the tested soil and stone materials c ;
[0060] Image segmentation module: used to segment the block filling foundation soil and stone particle image (the image has been annotated). First, the image is downsampled five times. In each downsampling process, 2 3×3 convolutions and 1 2×2 maximum pooling are performed. The feature image is retained in each downsampling process. After four downsamplings, the feature image is generated. Then the feature image is deconvolved. Each deconvolution result is combined and connected with the convolution feature image of the corresponding layer, and then the combined image is convolved to extract the combined features, that is, a cross connection is completed. Finally, the feature image output after five cross connections is feature classified, that is, the segmentation process of the block filling foundation soil and stone particle image is completed.
[0061] Area processing module: used to process the segmented image and calculate the area fraction of large particles
[0062] Foundation bearing capacity calculation module: used to calculate the bearing capacity of the foundation according to the force transmission coefficient η, safety factor k, and compressive strength σ c Calculate the foundation bearing capacity f of the test soil and rock pile area.
[0063] The present invention further proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned method for estimating the bearing capacity of a block-filled foundation based on image recognition.
[0064] Although the preferred embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms of specific changes without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the scope of protection of the present invention.
Claims
1. A method for estimating the bearing capacity of block-filled foundation based on image recognition, Features: The method comprises: Step 1: Conduct a single particle mechanical strength test on the test soil and rock material category, draw the probability distribution curve of the compressive strength σ of the test soil and rock material category, and obtain the representative value of the particle compressive strength σ c ; Step 2: Take a picture of the test soil and rock stockpile area, annotate the soil and rock particles in the picture, and outline the contours of the soil and rock particles; Step 3: Use the U-Net model based on the full neural network to segment the annotated image; the specific steps are: downsample the annotated image five times, retain the feature image in each downsampling process, and generate the feature image after four downsampling; then deconvolve the feature image, combine and connect the deconvolution result with the convolution feature image of the corresponding layer, and then convolve the combined image to extract the combined features, that is, complete a cross connection; finally, perform feature classification on the output feature image after five cross connections, that is, complete the segmentation process of the block filling foundation soil and stone particle image; Step 4: Perform area processing on the segmented image and calculate the area fraction of large particles Step 5: Based on the area fraction of large particles Calculate the void fraction Calculate the corresponding force transmission coefficient η according to the pore fraction w of the image; The calculation formula of the large particle area fraction φ is: Where n is the total number of large particles; S is the total area of the image; a and b are the major and minor axes of the circumscribed ellipse corresponding to the contour of the particle's maximum projection surface, respectively; The corresponding relationship between the porosity fraction w and the force transmission coefficient η is established through indoor tests and numerical simulations: several groups of test methods for the force transmission coefficient of soil and rock are carried out, and the relationship function g(w) between the porosity fraction w and the force transmission coefficient η of the test soil and rock is obtained by the linear regression method. The force transmission coefficient η is obtained by substituting the porosity fraction w of the image into the g(w) function; Step 6: According to the force transmission coefficient η and safety factor k, and compressive strength σ c The foundation bearing capacity of the test soil and rock pile area is calculated as f = η·σ c / k.
2. According to the method for estimating the bearing capacity of block filling foundation based on image recognition according to claim 1, Features: The specific steps of step 1 are: making several test soil and stone material standard samples, and testing their uniaxial compressive strength respectively; processing the uniaxial compressive strength data, calculating the arithmetic mean standard deviation, setting the confidence level, obtaining the confidence interval, screening out the data within the confidence interval, drawing the probability distribution curve of the compressive strength σ of the test soil and stone material category, calculating the arithmetic mean of the data within the confidence interval, and obtaining the representative value of the particle compressive strength σ c .
3. According to the method for estimating the bearing capacity of block-filled foundation based on image recognition according to claim 1, Features: The safety factor k ranges from 1.2 to 1.
5.
4. A system for estimating the bearing capacity of a block-filled foundation based on image recognition, which is implemented based on the method for estimating the bearing capacity of a block-filled foundation based on image recognition as claimed in any one of claims 1 to 3. Features: The system includes a compressive strength calculation module, an image segmentation module, an area processing module, and a foundation bearing capacity calculation module; The compressive strength calculation module is used to process the uniaxial compressive strength test data of the tested soil and stone materials and calculate the representative value σ of the particle compressive strength of the tested soil and stone materials. c ; The image segmentation module is used to segment the image of soil and stone particles in the block filling foundation. First, the image is downsampled five times. In each downsampling process of the image, 2 3×3 convolutions and 1 2×2 maximum pooling are performed, and the feature image is retained in each downsampling process; after four downsamplings, the feature image is generated; then the feature image is deconvolved, and each deconvolution result is combined and connected with the convolution feature image of the corresponding layer, and then the combined image is convolved to extract the combined features, that is, a cross connection is completed; finally, the feature image output after five cross connections is feature classified, that is, the segmentation process of the block filling foundation soil and stone particle image is completed; The area processing module is used to perform area processing on the segmented image and calculate the area fraction of large particles. The foundation bearing capacity calculation module is used to calculate the bearing capacity of the foundation according to the force transmission coefficient η and the safety factor k, as well as the compressive strength σ c Calculate the foundation bearing capacity f of the test soil and rock pile area.
5. A computer-readable storage medium storing a computer program, It is characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 3 is implemented.
