Method for monitoring sediment characteristics of sediment-bearing water flow based on image recognition

Abstract

The invention provides a method for monitoring sediment characteristics of sand-bearing water flow based on image recognition. The method comprises the following steps: acquiring a sand-bearing water flow image; preprocessing the sand-containing water flow image to obtain a preprocessed image; performing image enhancement on the preprocessed image to obtain an enhanced image; segmenting the enhanced image, wherein the segmentation comprises morphological operation particle segmentation, edge extraction and contour drawing; sediment feature extraction: drawing an external rectangular frame of sediment particles, extracting side length information of the rectangular frame, extracting quantity information of the rectangular frame and calculating concentration information; and calibrating system parameters. According to the device and the method for acquiring the sand-containing water flow image, the problem that shooting is difficult due to the characteristics of three-dimensional occurrence, tiny property, rapid flowability and the like of particles in water-containing sand can be solved; the invention discloses a machine-based sand-bearing water flow image recognition processing method, which can solve the difficult problems of sand-bearing water flow image particle recognition, overlapped particle segmentation and the like.

Description

Technical Field

[0001] The invention relates to the technical field of water conservancy project monitoring and image recognition, and in particular to a method for monitoring sediment characteristics of a sand-containing water flow based on image recognition. Background Art

[0002] In the flow of rivers, especially turbines in hydropower stations, the characteristics of sediment in the water flow are one of the more critical data indicators. The particle size and concentration information in the sediment characteristics information are of great significance to the normal operation and efficient operation of the turbines in hydropower stations. In sediment monitoring, the commonly used traditional methods include screening method, sedimentation method, resistance method, microscopy method, etc. The above methods have the disadvantages of complex operation, time-consuming and labor-intensive, low degree of automation or high detection cost.

[0003] Image recognition algorithm is a new method introduced in recent years. It has the advantages of simple operation, convenience, and high degree of information automation. However, the key problem is that the three-dimensional occurrence, tinyness, and rapid fluidity of sediment particles in water make it difficult to capture and acquire images; the characteristics of sandy water flow image particle recognition and overlapping particle segmentation make processing difficult.

[0004] Therefore, there is an urgent need for a method for monitoring the sediment characteristics of sand-laden water flows based on image recognition to achieve real-time, rapid and accurate monitoring of sediment characteristics. Summary of the invention

[0005] In view of the above technical problems, the purpose of the present invention is to provide a method for monitoring the sediment characteristics of sand-containing water flow based on image recognition, so as to solve the problems of sampling and shooting images of sand-containing water flow and image recognition and analysis.

[0006] To achieve the above purpose, the technical solution provided by the present invention is as follows:

[0007] In a first aspect, the present invention provides a device for acquiring images of a sand-containing water flow, comprising:

[0008] Water flow channel, light source and high-speed camera;

[0009] The water flow channel includes a bottom plate, a side plate and a slice flow channel;

[0010] The slice flow channels are connected in parallel to the bottom of the bottom plate along the water flow direction and form an inclination angle with the bottom plate; the slice flow channels are flat cubes with upper and lower openings, so that the sand-containing water flow expands after entering;

[0011] The high-speed camera is set parallel to the wide surface of the slice flow channel;

[0012] The light source is arranged on one side of the high-speed camera;

[0013] The wide surface of the slicing channel close to the high-speed camera is highly light-transmissive, and the other wide surface opposite to it serves as a background plate.

[0014] In a second aspect, the present invention provides a method for monitoring sediment characteristics of a sand-containing water flow based on image recognition using the sand-containing water flow image acquisition device according to claim 1, comprising the following steps:

[0015] Acquire images of sandy water flows;

[0016] Preprocessing the sand-containing water flow image to obtain a preprocessed image;

[0017] Performing image enhancement on the preprocessed image to obtain an enhanced image;

[0018] Segment the enhanced image, which includes morphological operations of particle segmentation, edge extraction, and contour drawing;

[0019] Sediment feature extraction, including drawing a rectangular frame around sediment particles, extracting information about the length of the rectangular frame, extracting information about the number of rectangular frames, and calculating sediment particle size and concentration information;

[0020] System parameter calibration.

[0021] In a possible implementation, the method for obtaining the image of the sand-containing water flow includes: setting the frame rate and magnification of a high-speed camera, recording a video after the sand-containing water flow stabilizes, and obtaining the image of the sand-containing water flow frame by frame from the video.

[0022] In a possible implementation, the preprocessing method includes the following steps:

[0023] Smoothing the sand-containing water flow image and generating a grayscale image;

[0024] Convert the grayscale image into a binary image;

[0025] Use edge algorithm to extract contours of binary images, including main contours and specific contours;

[0026] The mask is extracted from the original image according to the specific contour, and the original sand-containing water flow image is cropped using the mask to obtain a preprocessed image.

[0027] In a possible implementation manner, the specific outline is a square frame with a specified side length.

[0028] In a possible implementation, the image enhancement includes the following steps:

[0029] Perform filtering operation on the preprocessed image to obtain a denoised color image;

[0030] Convert color images to grayscale images;

[0031] Set the pixel threshold and convert the grayscale image into a binary image.

[0032] In a possible implementation, the segmentation method includes the following steps:

[0033] Perform morphological operations on the enhanced image, including dilation, erosion, opening and closing operations;

[0034] Edge detection and contour extraction to obtain the edges, number and circumscribed rectangular contours of sediment particles;

[0035] Draw the circumscribed rectangular outline of the sediment particles.

[0036] In a possible implementation, the method for calculating sediment particle size and sediment concentration information is as follows:

[0037] The calculation formula of sediment particle size is as follows:

[0038] r=a×α

[0039] Among them, the calculated value of sediment particle size is r, the length of the circumscribed rectangle of the sediment particle is a, and the particle size correction coefficient is α; the calculation formula for sediment concentration is as follows:

[0040]

[0041] Among them, the calculated value of water flow sediment concentration is ρ, the number of sediment particles in the image is n, and r i is the particle size of the i-th sediment particle, the concentration correction coefficient is β, the specific contour side length is A, and the slice flow channel thickness is d.

[0042] Further, the method for calibrating the system parameters includes:

[0043] Design a calibration experiment, add a certain weight of sediment particles into a certain volume of water flow as a standard sediment-laden water flow; calculate the concentration of the standard sediment-laden water flow; use a laser particle size analyzer to analyze and obtain the sediment particle size distribution curve and median particle size as the standard value of sediment particle size;

[0044] Set the initial value of the correction coefficient and calculate the sediment particle size and concentration;

[0045] Compare the calculated sediment particle size and concentration with the measured values ​​of the standard sediment-laden water flow and calculate the error coefficient;

[0046] Obtain the correction coefficient that minimizes the error coefficient to calibrate the system parameters.

[0047] Furthermore, the correction coefficient includes a particle size correction coefficient α and a concentration correction coefficient β;

[0048] The error coefficient includes the error of sediment particle size and the error of sediment concentration.

[0049] The beneficial effects of the present invention are as follows:

[0050] 1. The present invention provides a device for acquiring images of a sand-containing water flow. The device can well capture the flow process of sediment in a sand-containing water flow through the design of a sliced ​​flow channel and the arrangement of a high-speed camera and a light source, so as to obtain a clear image that is easy for subsequent image recognition.

[0051] 2. The present invention provides a method for monitoring sediment characteristics of sand-containing water flows based on image recognition. The method can realize fast, accurate and real-time processing and analysis of sediment images through image preprocessing, image enhancement and segmentation, combined with feature extraction and system parameter calibration, and thus obtain highly robust sediment characteristic output results. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of a device for acquiring images of sand-containing water flow;

[0053] Figure 2 are front views, wherein the upper view is a schematic diagram of only one slicing flow channel, and the lower view is a schematic diagram of three slicing flow channels;

[0054] Figure 1 , 2 Middle: 110-water flow channel, 111-bottom plate, 112-side plate, 113-slice flow channel, 120-light source, 130-high-speed camera, 140-water inlet, 150-water outlet;

[0055] Figure 3 It is a flow chart of a method for monitoring sediment characteristics of sand-laden water flow based on image recognition;

[0056] Figure 4 This is a sample image of the result after S310 processing;

[0057] Figure 5 This is the result sample image of converting the grayscale image into a binary image;

[0058] Figure 6 This is a sample image of the image segmentation result;

[0059] Figure 7 Particle size calculated for image recognition and particle size curves from laser particle size analyzer;

[0060] Figure 8 The concentration fitting curve is shown in Figure 2. DETAILED DESCRIPTION

[0061] In order to make the contents of the present invention more clearly understood, the present invention is further described below based on specific embodiments in conjunction with the accompanying drawings.

[0062] like Figure 1, 2 As shown, a device for acquiring images of a sand-containing water flow comprises:

[0063] A water flow channel 110, a light source 120 and a high-speed camera 130;

[0064] The water flow channel 110 includes a bottom plate 111, a side plate 112 and a slice flow channel 113;

[0065] The slice flow channels 113 are connected to the bottom of the bottom plate 111 in parallel along the water flow direction and form an inclination angle with the bottom plate 111; the slice flow channels 113 are flat cubes with upper and lower openings, so that the sand-containing water flow expands after entering;

[0066] The high-speed camera 130 is arranged parallel to the wide surface of the slicing flow channel 113;

[0067] The light source 120 is arranged on one side of the high-speed camera 130;

[0068] The wide side of the slicing channel 113 near the high-speed camera is highly light-transmissive, and the other wide side opposite is used as a background plate; this design is adopted to ensure sufficient light source and contrast during the photo-taking process, which is conducive to clear imaging.

[0069] In one possible implementation, the slice flow channel 113 is made of organic glass material, and the thickness of the slice flow channel is 3mm to 1cm. The slice flow channel adopts a flat cube design because the sand-laden water flow generates a planar water flow after passing through the slice flow channel, which is convenient for capturing the flow process of the sediment in the sand-laden water flow; the thickness is as small as possible to ensure that the sediment particles in the water are distributed one-sidedly during the shooting process, thereby reducing the interference of particle overlap on the recognition result. Multiple parallel slice flow channels are set to ensure the accuracy of the test, and the multiple averages are taken.

[0070] In a possible implementation, the inclination angle is 30 to 60 degrees.

[0071] It should be noted that the slicing channel 113 is inclined toward the water outlet 150 , with its upper end surface opening for water inlet and its lower end surface opening for water outlet, that is, water flows into the slicing channel 113 from the bottom plate of the water channel 110 .

[0072] Furthermore, the width of the slicing channel 113 is smaller than the width of the water channel 110 , and larger than the width of the lens of the high-speed camera 130 .

[0073] In this embodiment, the width of the slicing channel is 3 mm and the inclination angle is 45°.

[0074] In this embodiment, the sand-laden water flows evenly in the water flow channel 110, and flows out through the sliced ​​flow channel. The role of the flow channel inclination is to use gravity to make the sediment tend to be distributed in a plane while slowing down the flow rate; the focal plane of the high-speed camera coincides with the planar water flow, that is, the camera elevation angle in this embodiment is 45°, and the distance between the lens and the flow channel is the focal length of the lens at the corresponding magnification; using a strong light source can ensure sufficient exposure in the short time when the shutter is open.

[0075] like Figure 3 As shown, a method for monitoring sediment characteristics of a sand-containing water flow based on image recognition comprises the following steps:

[0076] S100, acquiring a sand-containing water flow image.

[0077] In a possible implementation, S100 includes: setting the frame rate and magnification of the high-speed camera, recording a video after the sand-laden water flow is stabilized, and acquiring the sand-laden water flow image frame by frame from the video.

[0078] Specifically, the frame rate of the high-speed camera is 500-7000 FPS.

[0079] Specifically, the magnification is 1-50 times, and the particle size range that can be monitored is 0.001-10 mm.

[0080] Specifically, a specific light source is set. In this example, a Dannyyou industrial light source is used. The color of the observed particles is mostly black, and white light or cold light (blue light) is prepared as the specific light source.

[0081] S200, preprocessing the sand-containing water flow image to obtain a preprocessed image.

[0082] In one possible implementation, S200 includes:

[0083] S210, smoothing the sand-containing water flow image and generating a grayscale image;

[0084] S220, converting the grayscale image into a binary image;

[0085] S230, extracting contours of the binary image using an edge algorithm, including main contours and specific contours;

[0086] S240, performing mask extraction on the original image according to the specific contour, and using the mask to crop the original sand-containing water flow image to obtain a preprocessed image.

[0087] Specifically, the smoothing process uses a mean filtering method, which can also reduce the impact of low-level edges on the edges of the main research area.

[0088] Specifically, the edge algorithm is a Sobel edge detection algorithm, which can quickly extract main contours based on a single threshold.

[0089] Specifically, the specific contour is a square frame with a specified side length. The known length of the square frame can be used as known physical information to complete subsequent work on the one hand, and on the other hand, it can narrow the research scope and improve the model accuracy.

[0090] S300, performing image enhancement on the preprocessed image to obtain an enhanced image.

[0091] In a possible implementation, the S300 includes:

[0092] S310, preprocessing the image ( Figure 4 ) performs filtering operation to obtain a denoised color image;

[0093] S320, converting the color image into a grayscale image;

[0094] S330, setting a pixel threshold to convert the grayscale image into a binary image.

[0095] Specifically, the filtering operation is Gaussian filtering, which is used to eliminate noise and smooth the image.

[0096] Specifically, the conversion formula for converting a color image to a grayscale image is:

[0097] Gray=0.2989×R+0.5870×G+0.1140×B

[0098] Among them: the weight coefficients 0.2989, 0.5870 and 0.1140 respectively represent the sensitivity of the human eye to the three colors red (R), green (G) and blue (B).

[0099] Specifically, the method for converting a grayscale image into a binary image is:

[0100] Obtain the grayscale distribution function (grayscale histogram) of the grayscale image through the grayscale image;

[0101] Preliminarily select a suitable threshold based on the distribution function;

[0102] The pixels with grayscale values ​​greater than the threshold are set as background (white), and the pixels with grayscale values ​​less than the threshold are set as foreground (black).

[0103] The result after conversion is Figure 5 shown.

[0104] S400, segmenting the enhanced image, including morphological operations of particle segmentation, edge extraction and contour drawing.

[0105] In a possible implementation, S400 includes:

[0106] S410, performing morphological operations on the enhanced image, including dilation, erosion, opening operation and closing operation;

[0107] S420, edge detection and contour extraction, obtaining the edge, number and circumscribed rectangular contour of the sediment particles;

[0108] S430, drawing the circumscribed rectangular outline of the sediment particles.

[0109] Specifically, in S410, dilation and closing operations are used to fill small holes, and corrosion and opening operations are used to eliminate tiny spots and blocks; the size and shape of the filter in the morphological operation have a greater impact on the result, and a square filter with a side length of 5 is initially selected.

[0110] Specifically, in S420, the edge detection and contour extraction, different from the algorithm in S230, adopts the canny edge detection algorithm, including the following steps:

[0111] Perform Gaussian filtering on the image processed by S410 to calculate the gradient intensity and direction of each pixel;

[0112] Non-maximum suppression, double threshold detection of edges, and suppression of isolated weak edges should be performed to complete edge detection.

[0113] Specifically, for visualization of the process, S430 draws the circumscribed rectangular outline of the detected sediment particles.

[0114] The segmentation results are as follows Figure 6 shown.

[0115] S500, sediment feature extraction, includes drawing a rectangular frame circumscribing sediment particles, extracting information about the length of the rectangular frame, extracting information about the number of rectangular frames, and calculating sediment particle size and concentration information.

[0116] In one possible implementation, the method for calculating sediment particle size and concentration information in S500 includes:

[0117] The calculation formula of sediment particle size is as follows:

[0118] r=a×α

[0119] Among them, the calculated value of sediment particle size is r, the length of the circumscribed rectangle of the sediment particle is a, and the particle size correction coefficient is α; the calculation formula for sediment concentration is as follows:

[0120]

[0121] Among them, the calculated value of water flow sediment concentration is ρ, the number of sediment particles in the image is n, and r i is the particle size of the i-th sediment particle, the concentration correction coefficient is β, the side length of the specific contour (square box with specified side length) is A, and the thickness of the slice flow channel is d.

[0122] It should be noted that the specific outline (square frame with specified side length) refers to the outer frame of the entire image, such as Figure 4 The frame of the entire image.

[0123] Specifically, in this embodiment, two function functions provided by OpenCV are called to extract the circumscribed rectangle.

[0124] S600, system parameter calibration.

[0125] In a possible implementation, the method for calibrating the system parameters in S600 includes:

[0126] Design a calibration experiment, add a certain weight of sediment particles into a certain volume of water flow as a standard sediment-laden water flow; calculate the concentration of the standard sediment-laden water flow; use a laser particle size analyzer to analyze and obtain the sediment particle size distribution curve and median particle size as the standard value of sediment particle size;

[0127] Set the initial value of the correction coefficient and calculate the sediment particle size and concentration;

[0128] Compare the calculated sediment particle size and concentration with the measured values ​​of the standard sediment-laden water flow and calculate the error coefficient;

[0129] Obtain the correction coefficient that minimizes the error coefficient to calibrate the system parameters.

[0130] Further, the correction coefficient includes a particle size correction coefficient α and a concentration correction coefficient β;

[0131] The error coefficient includes the error of sediment particle size and the error of sediment concentration.

[0132] It should be noted that the error coefficient of the median particle size is calculated based on the calculated value of the sediment particle size and the standard value; the error coefficient of the concentration is calculated based on the calculated value of the series of experimental concentrations and the standard value.

[0133] Furthermore, the correction method of the correction parameters can be either a manual modification method or calling the pyESMAD function to calibrate the parameters.

[0134] The following are related tests of the method described in this embodiment.

[0135] The particle size test and concentration test data of the standard sediment solution are shown in Table 1 and Table 2 respectively.

[0136] Table 1 Particle size test data of standard sediment solution

[0137]

[0138] Table 2 Concentration test data of standard sediment solution

[0139]

[0140]

[0141] The main parameters for the system parameter calibration described in this embodiment include: particle size correction coefficient α=1.118 and concentration correction coefficient β=2.335.

[0142] After the coefficient parameters are calibrated, it is used for sediment characteristics monitoring. The image recognition results are as follows: Figure 7 As shown, from Figure 7 It can be seen that the test results of image recognition are generally highly matched with the measured results, and can relatively accurately reflect the characteristics of cement-containing sand. Figure 8 is the concentration fitting curve, where the designed concentration is the measured value of the standard sediment solution, and the calculated concentration is the calculated value of image recognition. Figure 8 It can be seen that the method described in this embodiment is basically close to the measured value, and the two are highly consistent.

[0143] The technical solution provided by the present invention solves the problems of sampling and shooting images of sand-containing water flow and image analysis, and can realize real-time image capture of the flow state of sand in sand-containing water flow, as well as image recognition and analysis of sand characteristics, and then output high-precision measurement results.

[0144] The above description is only a preferred specific implementation manner of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention should be included in the scope of protection of the invention.

Claims

1. A device for acquiring images of sand-containing water flow, characterized in that: include: Water flow channel, light source and high-speed camera; The water flow channel includes a bottom plate, a side plate and a slice flow channel; The slice flow channels are connected in parallel to the bottom of the bottom plate along the water flow direction and form an inclination angle with the bottom plate; the slice flow channels are flat cubes with upper and lower openings, so that the sand-containing water flow expands after entering; The high-speed camera is set parallel to the wide surface of the slice flow channel; The light source is arranged on one side of the high-speed camera; The wide surface of the slicing channel close to the high-speed camera is highly light-transmissive, and the other wide surface opposite to it serves as a background plate.

2. A method for monitoring sediment characteristics of a sand-containing water flow based on image recognition using the sand-containing water flow image acquisition device according to claim 1, characterized in that: The following steps are involved: Acquire images of sandy water flows; Preprocessing the sand-containing water flow image to obtain a preprocessed image; Performing image enhancement on the preprocessed image to obtain an enhanced image; Segment the enhanced image, which includes morphological operations of particle segmentation, edge extraction, and contour drawing; Sediment feature extraction, including drawing a rectangular frame around sediment particles, extracting information about the length of the rectangular frame, extracting information about the number of rectangular frames, and calculating sediment particle size and concentration information; System parameter calibration.

3. The method for monitoring sediment characteristics of sand-containing water flow based on image recognition according to claim 2 is characterized in that: The method for obtaining the image of the sand-containing water flow comprises: setting the frame rate and magnification of a high-speed camera, recording a video after the sand-containing water flow is stable, and obtaining the image of the sand-containing water flow frame by frame from the video.

4. The method for monitoring sediment characteristics of sand-containing water flow based on image recognition according to claim 2 is characterized in that: The pretreatment method comprises the following steps: Smoothing the sand-containing water flow image and generating a grayscale image; Convert the grayscale image into a binary image; Use edge algorithm to extract contours of binary images, including main contours and specific contours; The mask is extracted from the original image according to the specific contour, and the original sand-containing water flow image is cropped using the mask to obtain a preprocessed image.

5. The method for monitoring sediment characteristics of sand-containing water flow based on image recognition according to claim 2 is characterized in that: The specific outline is a square frame with a specified side length.

6. The method for monitoring sediment characteristics of sand-containing water flow based on image recognition according to claim 2 is characterized in that: The image enhancement comprises the following steps: Perform filtering operation on the preprocessed image to obtain a denoised color image; Convert color images to grayscale images; Set the pixel threshold and convert the grayscale image into a binary image.

7. The method for monitoring sediment characteristics of sand-containing water flow based on image recognition according to claim 2 is characterized in that: The segmentation method comprises the following steps: Perform morphological operations on the enhanced image, including dilation, erosion, opening and closing operations; Edge detection and contour extraction to obtain the edges, number and circumscribed rectangular contours of sediment particles; Draw the circumscribed rectangular outline of the sediment particles.

8. The method for monitoring sediment characteristics of sand-containing water flow based on image recognition according to claim 2 is characterized in that: The method for calculating sediment particle size and sediment concentration information is as follows: The calculation formula of sediment particle size is as follows: r＝a×α Among them, the calculated value of sediment particle size is r, the length of the circumscribed rectangle of the sediment particle is a, and the particle size correction coefficient is α; the calculation formula for sediment concentration is as follows: Among them, the calculated value of water flow sediment concentration is ρ, the number of sediment particles in the image is n, and r i is the particle size of the i-th sediment particle, the concentration correction coefficient is β, the specific contour side length is A, and the slice flow channel thickness is d.

9. The method for monitoring sediment characteristics of sand-containing water flow based on image recognition according to claim 8 is characterized in that: The method for calibrating the system parameters comprises: Design a calibration experiment, add a certain weight of sediment particles into a certain volume of water flow as a standard sediment-laden water flow; calculate the concentration of the standard sediment-laden water flow; use a laser particle size analyzer to analyze and obtain the sediment particle size distribution curve and median particle size as the standard value of sediment particle size; Set the initial value of the correction coefficient and calculate the sediment particle size and concentration; Compare the calculated sediment particle size and concentration with the measured values ​​of the standard sediment-laden water flow and calculate the error coefficient; Obtain the correction coefficient that minimizes the error coefficient to calibrate the system parameters.

10. The method for monitoring sediment characteristics of a sand-containing water flow based on image recognition according to claim 9, characterized in that: The correction coefficient includes a particle size correction coefficient α and a concentration correction coefficient β; the error coefficient includes an error in sediment particle size and an error in sediment concentration.

Images