Transparent water tank test sediment space concentration measuring method based on multispectral reflectivity difference

Abstract

The invention relates to a transparent water tank test sediment space concentration measuring method based on multispectral reflectivity difference, and belongs to the technical field of water conservancy projects. The method comprises the following steps: preparing uniform sand-containing water body samples with different concentrations according to a certain concentration gradient, and shooting images of the samples with different concentrations by using a multispectral camera; extracting N wave band pixel values in the image by using OpenCV-python, and constructing a pixel value sequence of uniform sand-containing water body images with different concentrations; then adopting a machine learning algorithm, taking the sample sand content as a label sequence, and taking different wavelength reflectivity sequences as feature vectors to construct a water body sand content interpretation model; and finally, carrying out a transparent water tank disturbance experiment, shooting a sedimentary sediment disturbance moment image, and carrying out inversion on the sediment content space distribution by adopting the interpretation model, so as to realize the application of the model.

Description

Technical Field

[0001] The present invention relates to a method for measuring the spatial distribution of sediment concentration in a transparent flume test, specifically to a method for measuring the spatial concentration of sediment in a transparent flume test based on the difference in multi-spectral reflectance, and belongs to the field of hydraulic engineering. Background Art

[0002] The determination of sediment content is widely applied in hydraulic engineering, such as in the fields of reservoir silt cleaning, the starting mechanism of river sediment particles, and the movement mechanism of river suspended sediment. In the experimental study on the movement mechanism of sediment particles in reservoir silt cleaning, the impact of high-pressure water flow causes the bottom sediment of the riverbed to mix with the water flow. At the moment of impact, the mixing of suspended sediment and water flow is not uniform, and the spatial concentration distribution of suspended particulate sediment at this time is a very important research index. In the experimental study on the starting mechanism of river sediment particles, the rapid flow of water drives the bottom sediment to move by bed load and suspension. During the movement of suspended sediment, its spatial distribution in the river is not uniform. Therefore, it is also very necessary to measure the spatial distribution of suspended sediment particles in the river.

[0003] The traditional measurement methods of sediment content are mainly divided into two categories: contact type and non-contact type. The contact type measurement methods mainly include sampling and weighing method, vibration method, etc. The non-contact methods mainly include isotope method, infrared ray method, ultrasonic method, etc. Among them, the sampling and weighing method has the characteristics of high accuracy and high credibility. However, as a contact type measurement, it will inevitably affect the target to be measured, and the measurement takes a long time and has low efficiency. It is also difficult to measure the dynamic change process of sediment content concentration. The non-contact measurement method has a clear theoretical basis, believing that there is a certain correlation between light intensity or sound wave and sediment content concentration, and realizing the quantitative expression of sediment concentration by establishing the correlation between the two. However, most of the existing means based on this principle are single-point observations, and it is difficult to measure the spatial distribution of sediment content.

[0004] In recent years, with the development of optical remote sensing technology, many scholars have carried out research on the quantitative inversion of water body suspended sediment concentration based on remote sensing multi-band information. However, the relevant research mainly focuses on the inversion of large-scale sediment content information. Limited by the accuracy of remote sensing images, it is difficult to achieve fine measurement of the spatial distribution of sediment content, and due to the shooting angle, it is difficult to achieve cross-sectional concentration measurement. Summary of the Invention

[0005] The present invention is proposed to solve the measurement of the spatial distribution state of suspended sediment concentration, aiming to enable experiments such as reservoir silt scouring and the research on the movement mechanism of sediment particles in pipelines to more intuitively, accurately, and quickly obtain the spatial distribution state of sediment concentration. Compared with the method for inverting the spatial concentration of suspended sediment based on the difference in visible light transmittance of ordinary single-lens reflex cameras, it has advantages such as a high measurement upper limit and accurate data.

[0006] A method for measuring the spatial concentration of sediment in a transparent flume experiment based on the difference in multispectral reflectance according to an embodiment of the present invention includes the following steps:

[0007] Step 1: Use sediment as the test material. First, dry the sediment in a microwave oven and then grind it into a powder. Take a specific mass of sediment samples with a measuring cup, establish a suitable sediment concentration gradient, and label each cup with numbers from 1 to n and count the mass of each cup of samples as m i (i = 1, 2,... n);

[0008] Step 2: Set up the test equipment. Create a dark and enclosed space with a light-shielding cloth, and arrange a transparent flume, a multispectral camera, and a full-spectrum halogen light source inside.

[0009] Step 3: Add sediment samples of different masses into the transparent flume, and use the camera to take pictures of the transparent flume with different sediment contents to obtain n + 1 reflection pictures.

[0010] Step 4: Calculate the sediment content corresponding to each reflection picture according to the definition of the sediment content in the liquid.

[0011] Step 5: Read the spectral data of each reflection picture and construct a three-dimensional matrix of N×H×W for storage, where N, H, and W represent the number of bands, the number of pixels in the horizontal direction, and the number of pixels in the vertical direction of the spectral image, respectively.

[0012] Step 6: Based on the sediment content corresponding to each reflection picture and the spectral data of each reflection picture, construct a sediment content interpretation model for the water body.

[0013] Step 7: After the sediment in the transparent flume has been statically settled for a long time to form a solid-liquid stratification, use a syringe filled with clear water to jet the sediment deposited at the bottom, and use a multispectral camera to take a spectral image of the spatial distribution of suspended sediment at the moment of impact disturbance.

[0014] Step 8: Classify the spectral image of the spatial distribution according to the differences in the spectral curves of some regions in the image. First, make a category determination to distinguish the sediment and clear water in the transparent flume, then select a classifier for boundary processing, and finally output the processed spectral image of the spatial distribution.

[0015] Step 9: Extract the spectral data of the processed spectral image of the spatial distribution, and based on the sediment content interpretation model for the water body and the spectral data of the processed spectral image of the spatial distribution, invert the spatial distribution state of the sediment content.

[0016] In a feasible embodiment, in Step 4, the calculation formula for the sediment content is as follows:

[0017]

[0018] Among them, is the sediment concentration of the muddy water in the transparent water tank for the nth time, with the unit of kg / m 3 ; V0 is the volume of the initial clear water in the transparent water tank, with the unit of m 3 ; m i is the mass of the sediment sample taken in the measuring cup i (i = 1, 2,... n), with the unit of kg; ρ is the density of the taken sediment sample, with the unit of kg / m 3 .

[0019] In a feasible embodiment, step 5 specifically includes:

[0020] The reflection pictures taken by the multispectral camera are viewed using ENVI software. In ENVI, the spectral curves of different pixel points in the reflection pictures can be observed. The wavelength range of the spectral curves is between 452nm - 778nm; by observing the reflection pictures, an appropriate analysis window is selected therefrom to determine the processed image range; indexing is performed through the defined spectral image window, and python is used to read the spectral data of all band channels of each pixel point in each image window, and a three-dimensional matrix of N×H×W is constructed for storage, where N, H, and W respectively represent the number of bands, the number of horizontal pixels, and the number of vertical pixels of the spectral image. Finally, the image filtering algorithm is used to perform noise reduction processing on the spectral data.

[0021] In a feasible embodiment, step 6 specifically includes:

[0022] Taking the spectral curve features as the input and the water body sediment concentration as the label, a model is constructed through various machine learning algorithms to obtain the optimal interpretation model; on the captured planar two-dimensional image, pixel-by-pixel modeling is performed for machine learning fitting. Taking the reflectance of each pixel at different wavelengths as x and the sediment concentration of the water body as y for fitting, using the random sampling method, 75% of the data samples are selected as the training set, and the remaining 25% are used as the test set.

[0023] In a feasible embodiment, in step 6, the model uses R 2 as the accuracy evaluation index, and the calculation formula is as follows:

[0024]

[0025] In the formula, R 2 is the determination coefficient of the predicted value; is the sediment concentration predicted value of 25%; y i is the true value of the sediment concentration of 25%; is the true average value of the sediment concentration of 25%.

[0026] In a feasible embodiment, step 9 specifically includes:

[0027] Use Python to read the spatially distributed spectral image processed in step 8, and then use an image filtering algorithm to perform noise reduction processing on the processed spatially distributed spectral image; finally, perform different processing on sediment and clear water respectively. Perform unified assignment processing on the clear water part, and use the pixel coordinates X and Y of each pixel as indexes for the sediment part to call out the corresponding models for each pixel. Using the spectral data as input, invert the spatial distribution state of the sediment concentration to obtain the spatial distribution state of the suspended sediment concentration at the moment of jet disturbance in the transparent water tank.

[0028] In a feasible embodiment, in step 7, during the process of collecting jet disturbance images, the environmental optical conditions should be kept strictly consistent with the previous sample calibration test.

[0029] In a feasible embodiment, step 2 specifically includes:

[0030] The test equipment includes an iron rack, a light-shielding cloth, a wooden table, a transparent water tank filled with clear water, a multispectral camera, and a full-spectrum halogen light source. Use the light-shielding cloth to cover the iron rack to form a closed space. Place the wooden table in the middle of the closed space. Place the transparent water tank directly above the wooden table. Place the full-spectrum halogen light source in the front side of the transparent water tank so that the light shines on the front part of the transparent water tank for reflection, and place the multispectral camera directly in front of the transparent water tank.

[0031] In a feasible embodiment, step 3 specifically includes:

[0032] First, when no sediment sample is added, turn on the full-spectrum halogen light source and use the multispectral camera to take pictures of the transparent water tank to obtain the reflection pictures under the clear water condition; then pour the sediment sample numbered 1 into the transparent water tank, stir evenly, and use the multispectral camera to take pictures of the transparent water tank to obtain the reflection pictures under the muddy water condition; finally, perform the subsequent samples in sequence according to this step to obtain n + 1 reflection pictures;

[0033] Among them, in step 3, during the process of taking pictures of different groups, keep the camera parameters and light source conditions consistent.

[0034] In a feasible embodiment, in step 2, the multispectral camera operates according to the following steps:

[0035] 1) Connect the multispectral camera and open the operation interface;

[0036] 2) After the software is started, the multispectral camera is automatically connected. At this time, place the standard white board within the field of view of the multispectral camera and click automatic exposure;

[0037] 3) Cover the lens cap and click the dark background setting button;

[0038] 4) Click the standard plate setting button, place the standard white board within the red frame area, and ensure that the red frame area is completely filled with the white board;

[0039] 5) Click the standard plate setting button again. When the image acquisition button lights up, the standard plate setting is successful;

[0040] 6) After the image acquisition button lights up, click the image acquisition button to normally acquire images.

[0041] In summary, the present invention has the characteristics of being pioneering, non-contact, and efficient. At present, most of the measurements of sediment concentration are achieved through contact methods, which will inevitably affect the test accuracy. The present invention uses a non-contact optical method to measure the sediment concentration, fundamentally avoiding interference with the measurement target. In addition, most of the existing sediment concentration measurement technologies are single-point type, and it is difficult to obtain the spatial distribution state of the suspended sediment concentration. However, the present invention constructs the correlation between the sediment-containing water body and the spectral reflectance of different wavelengths based on the image method, and can quickly and accurately obtain the spatial distribution state of the sediment concentration, having advantages incomparable to single-point measurement methods. Description of the Drawings

[0042] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0043] Figure 1 It is a schematic structural diagram of a measurement device for a method for measuring the spatial concentration of sediment in a transparent flume experiment based on the difference in multi-spectral reflectance given in an embodiment of the present invention;

[0044] Figure 2 It is a schematic flow diagram of a method for measuring the spatial concentration of sediment in a transparent flume experiment based on the difference in multi-spectral reflectance given in a specific embodiment of the present invention;

[0045] Figure 3 It is a graph showing the relationship between the pixel values of different bands and the sample sediment concentration given in an embodiment of the present invention;

[0046] Figure 4 It is a comparison graph between the actual value and the predicted value of the sediment concentration interpretation model given in an embodiment of the present invention;

[0047] Figure 5 It is a process diagram of parameter calibration of the sediment concentration interpretation model in a machine learning algorithm given in an embodiment of the present invention;

[0048] Figure 6 This is a comparison diagram of the spectral images of clear water and turbid water given in the embodiments of the present invention;

[0049] Figure 7 This is a spatial distribution map of sediment in the disturbance test given in the embodiments of the present invention;

[0050] Figure 8 This is an inversion map of the spatial distribution of sediment concentration given in the embodiments of the present invention.

[0051] Explanation of reference numerals:

[0052] 1. Multispectral camera; 2. Wooden table; 3. Transparent water tank; 4. Sediment-laden water sample; 5. Full-spectrum halogen light source; 6. Iron rack; 7. Light-shielding cloth; 8. Image processing window. Detailed implementation manners

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0054] In the description of the embodiments of the present invention, it should be noted that the orientation or positional relationships indicated by the terms "center", "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the embodiments of the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the embodiments of the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0055] In the description of the embodiments of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "connected" and "coupled" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of the present invention can be understood according to specific circumstances.

[0056] In the embodiments of the present invention, unless otherwise clearly defined and limited, the first feature being "on" or "under" the second feature may be that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, the first feature being "above", "over" and "on top of" the second feature may be that the first feature is directly above or obliquely above the second feature, or simply means that the horizontal height of the first feature is higher than that of the second feature. The first feature being "under", "beneath" and "underneath" the second feature may be that the first feature is directly below or obliquely below the second feature, or simply means that the horizontal height of the first feature is less than that of the second feature.

[0057] The present invention will be described in detail below with reference to the accompanying drawings.

[0058] First, the present invention prepares uniform sand-containing water body samples with different concentrations according to a certain concentration gradient, and uses a multispectral camera 1 to take images of samples with different concentrations; then, OpenCV-python is used to extract the spectral reflectance of all band channels of each pixel point in the image window, and the correlation between the sand-containing water body images with different concentrations and the spectral reflectance is constructed, as Figure 3 shown; then, a machine learning algorithm is adopted to build a model for each pixel point, using the spectral reflectance of different wavelengths as the input quantity and the sand content of the uniform sand-containing water body as the output quantity to build the water body sand content interpretation model corresponding to each pixel point; finally, by conducting a perturbation experiment on the transparent water tank 3, taking images of the moment when the deposited sediment is perturbed, and using the above-mentioned interpretation model, with X and Y as indices, the sand content of each pixel is inverted to achieve the measurement of the spatial distribution of suspended sediment.

[0059] As Figures 1 to 8 shown, the following will describe in detail the specific implementation scheme of a method for measuring the spatial concentration of test sediment in a transparent water tank 3 based on the difference in multispectral reflectance according to the present invention with reference to the accompanying drawings, Figure 2 which are the specific implementation steps of this method.

[0060] Step 1: Prepare samples for the experiment, including sample preparation and sampling of the test materials. The test materials used are the bottom sediment of the Yellow River. First, dry the sediment in a microwave oven, and then grind it into powder (to facilitate the full mixing of sediment and water). Use a measuring cup to take a specific mass of sediment samples to establish a suitable sediment concentration gradient. Sampling process: Adjust the analytical balance, place an empty measuring cup, zero it, and add an appropriate amount of sediment powder. A total of 1 to 286 cups of samples are taken. After preliminary experiments, it is decided that for 1 - 100 samplings, it is about 10 g, for 101 - 200 samplings, it is about 20 g, and for 201 samplings and later, it is about 30 g. Label each measuring cup, denoted as 1 to 287, and the weight of the sand weighed for each cup of sample is m i (i = 1, 2,..., 286).

[0061] Step 2: Set up the test equipment. As shown in Figure 1 , the test equipment includes an iron rack 6 covered with a light-shielding cloth 7, a transparent water tank 3 filled with 3L of clear water, a multispectral camera 1, a full-spectrum halogen light source 5, and a wooden table 2. The test is carried out in a dark enclosed space. The iron rack 6 is completely covered with the light-shielding cloth 7 to form a dark enclosed space. The wooden table 2 is placed in the middle of the enclosed space, and the transparent water tank 3 filled with 3L of clear water is placed directly above the wooden table 2. The full-spectrum halogen light source 5 is placed in the front side of the transparent water tank 3 to simulate sufficient natural light to irradiate the transparent water tank 3. The multispectral camera 1 is placed directly in front of the transparent water tank 3 so that the horizontal heights of the multispectral camera 1 and the transparent water tank 3 are kept the same for calibration, focusing, and exposure of the multispectral camera 1.

[0062] The multispectral camera 1 operates according to the following steps:

[0063] 1) Start the multispectral camera control software on the PC side;

[0064] 2) After the software is started, the camera is automatically connected. At this time, place the standard white board within the camera's field of view and click on auto exposure (when the dark background setting button lights up, it means the auto exposure is successful);

[0065] 3) Cover the lens cap and click on the dark background setting button (when the standard board setting button lights up, it means the dark background setting is successful);

[0066] 4) Click on the standard board setting button and place the standard white board within the red frame area (ensure that the red frame area is completely filled with the white board);

[0067] 5) Click on the standard board setting button again (when the image acquisition button lights up, it means the standard board setting is successful);

[0068] 6) After the image acquisition button lights up, click on the image acquisition button to normally acquire images.

[0069] When debugging the camera, it is necessary to pay attention to whether the calibration is successful. When the spectral curves of each pixel point in the captured standard white board image are a horizontal straight line, it indicates that the calibration is successful. Except for the light source, all objects in this test are wrapped with the light-shielding cloth 7 to prevent the reflection of objects from affecting the test and ensure the uniformity of the incident light source.

[0070] Step 3: Photographing experiment. Photographing experiments are carried out on the transparent flume 3 with different sediment concentrations. First, when no sample is added, turn on the full-spectrum halogen light source 5 and take a photo of the transparent flume 3 with a camera to obtain a reflection picture under the clear water condition; then pour the sample numbered 1 into the transparent flume 3, stir evenly, and take a photo of the transparent flume 3 with the multispectral camera 1 to obtain a reflection picture of the sediment-laden water sample 4; finally, perform the subsequent samples in sequence according to this step to obtain n + 1 reflection pictures. During the process of taking photos of different groups, keep the relative positions of the full-spectrum halogen light source 5, the transparent flume 3, and the multispectral camera 1 strictly consistent.

[0071] Step 4, calculation of the sediment concentration of the sample. According to the definition of the sediment concentration of the liquid, the sediment concentration corresponding to each picture is calculated by Equation (1):

[0072]

[0073] In the formula, is the sediment concentration of the muddy water in the transparent flume 3 for the nth time, kg / m 3 ; V0 is the volume of the initial clear water in the transparent flume 3, m 3 ; m i is the mass of the dry sand sample taken in the measuring cup i, kg, (i = 1, 2,... 286); ρ is the sediment density of the sample taken, kg / m 3 .

[0074] Step 5: Analysis of the pixel characteristics of the spectral image of the sediment-laden water body. The spectral image (i.e., the reflection picture) taken by the multispectral camera 1 is viewed with ENVI software, which has 18 band channels. In ENVI, the spectral curves of different pixel points in the spectral image can be observed, and the wavelength range of the spectral curves is between 452 nm and 778 nm. By observing the spectral image, an appropriate analysis window is selected to determine the processing image range. Index through the defined spectral image window, use python to read the data of each pixel point in each image window, and mark them as X and Y; then construct a three-dimensional matrix for storage. The original size of the three-dimensional matrix constructed in the present invention is 18×500×300; finally, use algorithms such as median filtering, mean filtering, and Gaussian filtering to perform noise reduction processing on the spectral image.

[0075] Step 6: Construction of interpretation model for water body sediment content. The model was constructed through a variety of machine learning algorithms to obtain the optimal interpretation model, mainly using multiple linear regression (MLR) and support vector regression (SVR) algorithms for model construction. When using the multiple regression machine learning algorithm for modeling, the spectral curve characteristics are used as input and the water body sediment content is used as the label. In order to eliminate the influence of uneven light caused by external factors such as light source angle and experimental equipment, this study conducted machine learning fitting by pixel-by-pixel modeling on the captured two-dimensional images (ie, reflection images). Specifically, the reflectivity of each pixel at different wavelengths is x, and the water body sediment content is y. The fitting is performed, and a random sampling method is used to select 75% of the data samples as the training set, and the remaining 25% as the test set. The model uses R 2 As an accuracy evaluation index, R can be obtained by calculating the formula (2): 2 is 0.989:

[0076]

[0077] In the formula, R 2 is the coefficient of determination of the predicted value; is the predicted value of sand content of 25%; i The actual value of the sand content is 25%; is the true average value of sand content of 25%. Figure 5 This is a diagram of the parameter calibration process of the sediment content interpretation model in the machine learning algorithm provided in the embodiment of the present invention.

[0078] The SVR algorithm also uses spectral curve features as input and water body sediment content as label for pixel-by-pixel machine learning fitting. 75% of the samples are used as training sets and 25% of the samples are used as test sets to build the SVR model. The difference is that SVR uses parameter optimization methods such as cross-validation or network search to calibrate the best modeling parameters, and trains based on the best modeling parameters to obtain the sediment content interpretation model. The model accuracy is evaluated based on the test set, and the evaluation index R 2 The calculation formula is the same as the above formula (2). The final expression (3) of the SVR model is as follows:

[0079]

[0080] Where: α i , are all Lagrange operators, k(x i ,x) is the kernel function, and b is the estimated offset parameter. After debugging, this experiment selected the kernel function rfb and the penalty factor C=100 as the model parameters, and evaluated the model accuracy based on the test set. It can reach 0.986.

[0081] It is found through fitting the sediment concentration by two machine learning algorithms that the MLR model has a good fitting effect for linear correlation. For the SVR model, through the selection of its kernel function (rbf, poly, sigmoid, etc.) and the setting of parameters such as C (the penalty factor for error terms) and gamma (the coefficient of the kernel function), it is more suitable for non-linear correlation fitting.

[0082] Step 7: Perturbation test. After the sediment in the transparent water tank 3 has been statically settled for a long time, a solid-liquid stratification phenomenon is formed, with the sediment part at the lower layer and the nearly clear water part at the upper layer. Then, a syringe filled with clear water is used to jet the sediment deposited at the bottom. The deposited sediment after being perturbed shows a dispersed state in the transparent water tank 3, and a multi-spectral camera 1 is used to take images of the spatial distribution of suspended sediment at the moment of impact perturbation, as Figure 7 shown. During the process of collecting jet perturbation images, it is necessary to keep the environmental optical conditions strictly consistent with the previous sample calibration test to avoid affecting the optical image analysis of the pictures taken in the test.

[0083] Step 8: Remote sensing image supervised classification processing. The support vector machine (SVM) is used to classify the jet test pictures by using the different spectral curves of clear water and sediment in the image, as Figure 6 shown. First, the category is determined to distinguish sediment and clear water, then the SVM is used to divide its boundary, and finally the picture is output.

[0084] Step 9: Inversion of the spatial distribution state of sediment concentration. Use python to read the jet instant spectral image processed in Step 8, and then use an image filtering algorithm to denoise the spectral image; finally, different treatments are carried out for the two categories of sediment and clear water. The clear water part is assigned a value, and for the sediment part, the corresponding model of each pixel is retrieved with the pixel coordinates X and Y as the index through Step 6. With the spectral curve as the input, the spatial distribution state of the sediment concentration is inverted to obtain the spatial distribution state of the suspended sediment concentration at the moment of jet perturbation in the transparent water tank 3, as Figure 8 shown. Further, as Figure 4 shown, Figure 4 is the comparison chart of the actual value and the predicted value of the sediment concentration interpretation model given in the embodiment of the present invention.

[0085] In summary, the present invention features innovation, non-contact, and high efficiency. Currently, most measurements of sediment concentration are achieved through contact methods, which inevitably affect the test accuracy. The present invention uses a non-contact optical method to measure sediment concentration, fundamentally avoiding interference with the measurement target. In addition, most existing sediment concentration measurement techniques are single-point type, making it difficult to obtain the spatial distribution of suspended sediment concentration. However, based on a matrix multi-spectral camera, the present invention uses an image method to establish the correlation between sediment-laden water and spectral reflectance at different wavelengths, enabling rapid and accurate acquisition of the spatial distribution of sediment concentration, with advantages incomparable to single-point measurement methods.

[0086] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for measuring the spatial concentration of sediment in a transparent flume experiment based on the difference in multispectral reflectance, characterized in that, It includes the following steps: Step 1: Use sediment as the test material. First, dry the sediment in a microwave oven and then grind it into powder. Take out sediment samples of a specific mass with a measuring cup, establish a suitable sediment concentration gradient, label each cup with serial numbers from 1 to n, and count the mass of each cup of samples as m i (i = 1, 2, … n); Step 2: Set up the test equipment. Create a dark and enclosed space with a light-shielding cloth, and arrange a transparent water tank, a multispectral camera, and a full-spectrum halogen light source inside. Step 3: Add sediment samples of different masses into the transparent water tank, and use the camera to take pictures of the transparent water tank with different sediment concentrations to obtain n + 1 reflection pictures. Step 4: Calculate the sediment concentration corresponding to each reflection picture according to the definition of the sediment concentration in the liquid. Step 5: Read the spectral data of each reflection picture and construct a three-dimensional matrix of N×H×W for storage, where N, H, and W represent the number of bands, the number of pixels in the horizontal direction, and the number of pixels in the vertical direction of the spectral image, respectively. Step 6: Based on the sediment concentration corresponding to each reflection picture and the spectral data of each reflection picture, construct a sediment concentration interpretation model for water bodies. Step 7: After the sediment in the transparent water tank has been statically settled for a long time to form a solid-liquid stratification, use a syringe filled with clear water to jet the sediment deposited at the bottom, and use a multispectral camera to take the spatial distribution spectral image of the suspended sediment at the moment of impact disturbance. Step 8: Classify the spatial distribution spectral image according to the differences in the spectral curves of some regions in the image. First, perform category determination to distinguish the sediment and clear water in the transparent water tank, then select a classifier for boundary processing, and finally output the processed spatial distribution spectral image. Step 9: Extract the spectral data of the processed spatial distribution spectral image, and based on the sediment concentration interpretation model for water bodies and the spectral data of the processed spatial distribution spectral image, invert the spatial distribution state of the sediment concentration.

2. The method for measuring the spatial concentration of sediment in a flume test based on the difference in multispectral reflectance according to claim 1, characterized in that In Step 4, the calculation formula for the sediment concentration is as follows: Wherein, is the sediment concentration of the turbid water in the transparent water tank at the nth time, with the unit of kg / m 3 ; V0 is the volume of the initial clear water in the transparent water tank, with the unit of m 3 ; m i is the mass of the sediment sample taken in the measuring cup i (i = 1, 2,... n), with the unit of kg; ρ is the density of the sediment sample taken, with the unit of kg / m 3 .

3. A method for measuring the spatial concentration of sediment in a flume experiment based on the difference in multispectral reflectance, as claimed in claim 1, wherein The specific content of Step 5 includes: The reflection pictures taken by the multispectral camera are viewed with ENVI software. In ENVI, the spectral curves of different pixel points in the reflection pictures can be observed. The wavelength range of the spectral curves is between 452nm - 778nm; by observing the reflection pictures, select an appropriate analysis window to determine the image processing range; index through the defined spectral image window, use python to read the spectral data of all band channels of each pixel point in each image window, construct a three-dimensional matrix of N×H×W for storage, where N, H, and W represent the number of bands, the number of pixels in the horizontal direction, and the number of pixels in the vertical direction of the spectral image respectively, and finally use an image filtering algorithm to denoise the spectral data.

4. A method for measuring the spatial concentration of sediment in a flume experiment based on the difference in multispectral reflectance according to claim 1, characterized in that, The specific content of Step 6 includes: Using the spectral curve characteristics as the input and the sediment concentration of the water body as the label, construct a model through various machine learning algorithms to obtain the optimal interpretation model; on the captured planar two-dimensional image, perform pixel-by-pixel modeling for machine learning fitting. Take the reflectance of each pixel at different wavelengths as x and the sediment concentration of the water body as y for fitting. Using the random sampling method, select 75% of the data samples as the training set and the remaining 25% as the test set.

5. A method for measuring the spatial concentration of sediment in a flume test based on the difference in multispectral reflectance according to claim 4, characterized in that, In step 6, the model uses R 2 as the accuracy evaluation index, and the calculation formula is as follows: where R 2 is the coefficient of determination of the predicted value; is the predicted sediment concentration value at 25%; y i is the true sediment concentration value at 25%; is the true average sediment concentration value at 25%.

6. The method for measuring the spatial concentration of sediment in a flume experiment based on the difference in multispectral reflectance according to claim 1, characterized in that, The specific content of Step 9 includes: Read the spatially distributed spectral image processed in step 8 using Python, and then use an image filtering algorithm to denoise the processed spatially distributed spectral image; finally, perform different treatments on sediment and clear water respectively. Perform unified assignment processing on the clear water part, and use the pixel coordinates X and Y of each pixel as indices for the sediment part to call out the corresponding models for each pixel. Using the spectral data as input, invert the spatial distribution state of the sediment concentration to obtain the spatial distribution state of the suspended sediment concentration at the moment of jet disturbance in the transparent water tank.

7. A method for measuring the spatial concentration of sediment in a flume experiment based on the difference in multispectral reflectance according to claim 1, characterized in that, In step 7, during the process of collecting jet disturbance images, the environmental optical conditions should be kept strictly consistent with the previous sample calibration tests.

8. A method for measuring the spatial concentration of sediment in a flume experiment based on the difference in multispectral reflectance according to any one of claims 1 to 7, characterized in that, The specific steps of step 2 are as follows: The test equipment includes an iron rack, a light-shielding cloth, a wooden table, a transparent water tank filled with clear water, a multispectral camera, and a full-spectrum halogen light source. Cover the iron rack with the light-shielding cloth to form a closed space. Place the wooden table in the middle of the closed space. Place the transparent water tank directly above the wooden table. Place the full-spectrum halogen light source in the front side of the transparent water tank so that the light shines on the front part of the transparent water tank for reflection. Place the multispectral camera directly in front of the transparent water tank.

9. A method for measuring the spatial concentration of sediment in a flume experiment based on the difference in multispectral reflectance according to any one of claims 1 to 7, characterized in that, The specific steps of step 3 are as follows: First, when no sediment sample is added, turn on the full-spectrum halogen light source and take a photo of the transparent water tank with the multispectral camera to obtain a reflection picture under the clear water condition; then pour the sediment sample numbered 1 into the transparent water tank, stir evenly, and take a photo of the transparent water tank with the multispectral camera to obtain a reflection picture under the muddy water condition; finally, perform the following steps on the subsequent samples in sequence to obtain n + 1 reflection pictures; Among them, in step 3, during the process of taking photos of different groups, keep the camera parameters and light source conditions consistent.

10. A method for measuring the spatial concentration of sediment in a flume experiment based on the difference in multispectral reflectance according to any one of claims 1 to 7, characterized in that, In step 2, the multispectral camera operates according to the following steps: 1) Connect the multispectral camera and open the operation interface; 2) After the software is started, the multispectral camera will be automatically connected. At this time, place the standard whiteboard within the field of view of the multispectral camera and click auto exposure; 3) Cover the lens cap and click the dark background setting button; 4) Click the standard board setting button, place the standard whiteboard within the red frame area, and ensure that the red frame area is completely filled with the whiteboard; 5) Click the standard board setting button again. When the image acquisition button lights up, the standard board setting is successful; 6) After the image acquisition button lights up, click the image acquisition button to normally acquire images.

Images