An image recognition-based thickener pulp concentration measurement system

Abstract

The application discloses a kind of based on image recognition's thickener slurry concentration measuring system, including thickener slurry image acquisition system and its matched slurry image processing and identification system.Through image acquisition system, the image and displacement of the slurry to be measured in thickener are obtained, and a large number of known concentration slurry images are collected as standard samples for similarity identification.The image features of the sample to be measured and the standard sample are compared and identified, and the concentration of the slurry to be measured can be determined.The system designed by the application can continuously measure the concentration of the slurry at different depths in the thickener, and then determine the concentration distribution state of the slurry.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mine filling, and particularly provides a thickener slurry concentration measuring system based on image recognition. BACKGROUND

[0002] The thickener is mainly used for improving the tailings concentration, so as to be stored in the tailings pond or filled in the underground well. Due to the gravity sedimentation, the slurry concentration is different in different areas of the thickener, and accurately grasping the slurry concentration distribution is of great significance to improve the operation efficiency of the thickener. Therefore, it is necessary to develop a reliable and accurate thickener slurry concentration testing device.

[0003] Due to the complexity of the thickener, the slurry concentration is difficult to accurately measure at present, which restricts the understanding and understanding of the internal operation state of the thickener. The existing thickener slurry concentration is measured by using a radiation concentration meter, and this measurement method has problems of complex testing process and low efficiency. Other measurement methods only consider local measurement and cannot well understand and understand the overall operation state of the thickener. Therefore, it is urgent to propose a method for overall determination and analysis of the slurry concentration in the thickener. SUMMARY

[0004] The present application provides a thickener slurry concentration measuring system based on image recognition, which aims to accurately measure the slurry concentration at different positions in the thickener by image recognition method, and analyze and determine the overall distribution of the slurry concentration. Considering that the slurry concentration in the thickener changes with the depth, and as the slurry concentration increases, the density increases accordingly, the surface texture and speckle of the slurry will show different patterns, and the absorption and reflection ability of light will also differ. Therefore, the slurry concentration can be indirectly determined by collecting the slurry image in the thickener and the slurry image with known concentration, and by analyzing and comparing the feature similarity of the images.

[0005] In order to achieve the above purpose, the present application adopts the following technical scheme:

[0006] A thickener slurry concentration measuring system based on image recognition, characterized in that it comprises a slurry image acquisition system and a slurry image processing and recognition system, the image acquisition system acquires the thickener slurry image by using optical imaging principle, and the slurry image processing and recognition system sequentially performs image noise reduction processing, gray processing, feature extraction and contrast recognition, so as to realize the determination of the slurry concentration;

[0007] The slurry image acquisition system comprises a camera probe, a fixed pulley, a direct current brushless motor, a winding wheel and a PLC controller, the camera probe is connected with the PLC controller through the fixed pulley and the winding wheel via wires to transmit images and sensor data, the direct current brushless motor is driven by the PLC controller, the direct current brushless motor drives the winding wheel to control the wire outgoing and winding through the fixed pulley, and then the two-dimensional movement of the camera probe in the vertical direction is realized, the camera probe comprises an upper shell, a PCB board, an acceleration sensor, an LED ring, a light shield, a glass cover, a convex mirror, a lower shell, a counterweight, a CMOS sensor, a filter, a lens base and a concave lens, the PLC controller transmits the image and acceleration data to a slurry image processing and identification system in a computer through an expansion RJ45 port module for calculation and analysis.

[0008] Further, the glass cover of the camera probe is connected with the upper shell and the lower shell through two pairs of radial bolts, the lower shell is connected with the counterweight, the PCB board is arranged in the upper shell, the acceleration sensor and the CMOS sensor are welded to the PCB board through a wire harness, the CMOS sensor is sequentially provided with the filter, the lens base and the concave lens at the lower end, the LED ring is arranged on both sides of the lens base and connected to the PCB board pin through a DuPont wire, and the PCB board supplies power to the entire camera probe and transmits signals to the PLC controller.

[0009] Further, the lower end of the LED ring is further provided with a light shield to avoid that the light supplement source emitted by the LED ring directly irradiates the surface of the slurry.

[0010] Further, the acceleration sensor senses the motion acceleration of the probe to indirectly realize high-precision sensing of the speed and displacement of the probe, the two-dimensional displacement of the probe is calculated by accumulation to determine the depth position of the probe at the imaging time, and the conversion calculation formula of acceleration, speed and displacement is as follows:

[0011] v(t)=∫a(t)dt

[0012] x(t)=∫v(t)dt

[0013] Wherein, v is the speed, a is the acceleration, x is the displacement, and t is the time.

[0014] Further, the shape of the end of the lower shell matches the shape of the two ends of the convex mirror, and the convex mirror is clamped by screwing the lower shell and the counterweight.

[0015] Further, the counterweight is connected with the lower shell through threads to pull the probe to realize two-dimensional vertical movement, and different weights of the counterweight can be conveniently replaced according to the estimated value of the measured slurry concentration to ensure good traction effect.

[0016] Further, the slurry image processing and recognition system performs image processing and recognition as follows:

[0017] S1, obtaining the slurry image of the sample to be measured in the thickener by the slurry image acquisition system, collecting the slurry image of the known concentration of 1% to 90% by experiment, establishing the standard sample database, and fitting to determine the mapping relationship between the slurry concentration and the image features;

[0018] S2, image noise reduction processing: using the wavelet transform method of frequency domain noise reduction to process the image, so as to improve the speed and accuracy of image recognition;

[0019] S3, image gray processing: the slurry image after noise reduction processing is subjected to gray processing, so as to realize efficient extraction and utilization of image feature information; the gray processing of the image is to map and transform the original RGB color space to GRAY color space, and the mapping formula is as follows:

[0020] Gray = 0.299R + 0.587G + 0.114B

[0021] Wherein, R, G, B are the pixel values of red, green and blue three channels in RGB color space, and Gray is the gray value in GRAY color space;

[0022] S4, image feature extraction: the histogram of the gray image can be obtained by pixel statistical method; the color moment of the gray image is calculated based on mathematical method to describe the gray distribution, and the color moment calculation method of the gray image is as follows:

[0023]

[0024]

[0025]

[0026] Wherein, E g is the first order moment of the image, σ g is the second order moment of the image, S g is the third order moment of the image, p j represents the gray value of the jth pixel, and N is the number of pixels.

[0027] S5, image contrast recognition: first, compare the color moments of the sample image to be measured and the standard sample image, calculate the difference value of the color moments, and sort the standard sample images according to the difference value, select the 10 samples with the smallest difference value, and then calculate the Bhattacharyya coefficient representing the similarity between the test image and the 10 standard samples according to the gray histogram of the test image and the 10 standard samples, and the formula is as follows:

[0028]

[0029] Wherein, B is the bar coefficient of two images, p, q are the probability distribution of two images at the corresponding position x of histogram respectively.

[0030] S6, slurry concentration determination: comparing the most similar standard sample, and comprehensively color moment difference and bar coefficient, substituting into the slurry concentration-image feature mapping relationship, determining the actual slurry concentration of the image to be measured, the formula is as follows:

[0031]

[0032] Wherein, C p ,C q Respectively, the concentration of the image to be measured and the maximum similar image, Δ g The color moment difference between the image to be measured and the maximum similar image.

[0033] Further, the specific method of collecting 1% to 90% of the known concentration of slurry image test in the step S1 is as follows:

[0034] S11, the slurry with a known concentration of 1% is injected into the container, and the slurry image is obtained by using the image acquisition system to shoot;

[0035] S12, referring to steps S2-S4, the slurry image obtained by shooting is sequentially subjected to noise reduction processing, gray processing and feature extraction;

[0036] S13, sequentially replacing the slurry with a concentration of 1%, 2%, 3%,..., 90%, repeating S11 and S12;

[0037] S14, the gray histogram and color moment of the obtained 90 standard images are stored in a structured manner, and a standard sample database is established;

[0038] S15, taking the standard sample database as the training data, fitting and establishing the slurry concentration-image feature mapping relationship.

[0039] Beneficial effects

[0040] The beneficial effects of the present application are as follows:

[0041] (1) According to the device and method provided by the present application, the slurry image in the thickener under the light supplementing condition can be collected, and then the standard sample database containing the slurry concentration-image feature mapping relationship is established, and the measurement of the slurry concentration is realized by comparing the feature similarity of the image to be measured and the standard image;

[0042] (2) The present application can continuously collect slurry images at specific positions of the thickener, determine the slurry concentration at different depths of the positions after image processing and recognition, synthesize the slurry concentration distribution cloud image, and further determine the slurry layering height and master the overall distribution of the slurry concentration;

[0043] (3) The present application is designed based on optical imaging and digital image processing technology, has low measurement cost, high accuracy, and can realize continuous measurement; compared with the traditional ray measurement technology, the technical scheme provided by the present application is harmless to human body and does not pollute the environment. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 The slurry concentration measurement system provided by the present application is shown in the figure;

[0045] Figure 2 The detailed structure of the camera probe provided by the present application is shown in the figure;

[0046] Figure 3 The image processing and recognition flowchart provided by the present application is shown in the figure;

[0047] Figure 4 The acceleration time history curve drawn according to the sensor data is shown in the figure;

[0048] Figure 5 The velocity time history curve obtained by integration is shown in the figure;

[0049] Figure 6 The displacement time history curve obtained by integration is shown in the figure;

[0050] Figure 7 The concentration cloud image obtained by continuously collecting slurry images and recognition is shown in the figure;

[0051] The meanings of the symbols in the figure are as follows:

[0052] 1-camera probe; 2-fixed pulley; 3-direct current brushless motor; 4-winding wheel; 5-PLC controller; 6-computer; 101-upper shell; 102-PCB board; 103-acceleration sensor; 104-LED light ring; 105-shutter; 106-glass cover; 107-convex mirror; 108-lower shell; 109-counterweight; 110-CMOS sensor; 111-filiter; 112-lens base; 113-concave lens. DETAILED DESCRIPTION

[0053] The technical solutions in the present application will be described in detail below in combination with the embodiments and the drawings.

[0054] The application provides a thickener slurry concentration measuring system based on image recognition, which comprises a hardware part and a software part, the hardware part refers to a slurry image acquisition system, and the software part refers to a matched image processing and recognition system. The image acquisition system acquires thickener slurry images by using an optical imaging principle, and the image processing and recognition system sequentially performs image noise reduction processing, grayscale processing, feature extraction and contrast recognition, so that the determination of the slurry concentration is realized.

[0055] As shown in Figure 1 and Figure 2 , the slurry image acquisition system comprises a camera probe 1, a fixed pulley 2, a DC brushless motor 3, a winding wheel 4 and a PLC controller 5, the camera probe 1 is connected with the PLC controller 5 through wires via the fixed pulley 2 and the winding wheel 4 to transmit images and sensor data, the DC brushless motor 3 is driven by the PLC controller 5, the DC brushless motor 3 drives the winding wheel 4 and controls the wire outgoing and winding via the fixed pulley 2, so as to realize the two-dimensional movement of the camera probe 1 in the vertical direction, the camera probe 1 comprises an upper shell 101, a PCB board 102, an acceleration sensor 103, an LED light ring 104, a light shield plate 105, a glass cover 106, a convex mirror 107, a lower shell 108, a counterweight 109, a CMOS sensor 110, a filter 111, a lens base 112 and a concave lens 113, the PLC controller 5 transmits the images and acceleration data to the slurry image processing and recognition system in the computer 6 for calculation and analysis by means of an expansion RJ45 port module.

[0056] Further, the glass cover 106 of the camera probe is connected with the upper shell 101 and the lower shell 108 through two pairs of radial bolts, the lower shell 108 is connected with the counterweight 109, the PCB board 102 is arranged inside the upper shell 101, the acceleration sensor 103 and the CMOS sensor 110 are welded to the PCB board 102 through a wire harness, the CMOS sensor 110 is sequentially provided with the filter 111, the lens base 112 and the concave lens 113 at the lower end, the LED light ring 104 is arranged on both sides of the lens base and is connected to the PCB board pin through a DuPont wire, and the PCB board 102 supplies power to the entire camera probe and transmits signals to the PLC controller 5.

[0057] Further, the LED light ring 104 is further provided with the light shield plate 105 at the lower end, so as to avoid that the supplementary light source emitted by the LED light ring 104 directly irradiates the slurry surface.

[0058] The imaging mode of the camera probe 1 is as follows: the LED lamp ring 104 emits a light source, which is blocked by the light shield 105 to avoid direct irradiation, and the light is diffusely reflected to the surface of the slurry through the glass cover 106, thereby avoiding overexposure caused by direct irradiation on the basis of realizing the light supplement function; the slurry image supplemented with light is reflected and propagated by the convex mirror 107, and then is refracted by the concave lens 113 and captured by the CMOS sensor 110 through the optical filter 111.

[0059] Further, the acceleration sensor 103 senses the motion acceleration of the probe, indirectly realizes high-precision sensing of the speed and displacement of the probe, and determines the depth position of the probe at the imaging time by cumulatively calculating the two-dimensional displacement of the probe; the conversion calculation formula of acceleration, speed and displacement is as follows:

[0060] v(t)=∫a(t)dt

[0061] x(t)=∫v(t)dt

[0062] Wherein, v is the speed, a is the acceleration, x is the displacement, and t is the time.

[0063] As shown in Figure 4 , it is an acceleration-time curve drawn according to the acceleration sensor, and the speed-time curve can be obtained through the above formula, as shown in Figure 5 ; further, the displacement-time curve can be integrated, as shown in Figure 6 ; the depth position of the probe at the imaging time can be determined through the displacement-time curve.

[0064] Further, the end shape of the lower shell 108 matches the shape of the two ends of the convex mirror 107, and the convex mirror is clamped by screwing the lower shell 108 and the counterweight 109.

[0065] Further, the counterweight 109 is connected with the lower shell 108 through threads, and is used to pull the probe to realize vertical two-dimensional motion, and different weights of the counterweight can be conveniently replaced according to the estimated value of the measured slurry concentration, so as to ensure good traction effect.

[0066] As shown in Figure 3 , the software part provided by the application provides the following steps for the slurry image processing and recognition system to process and recognize images:

[0067] S1, obtaining the slurry image of the sample to be measured in the thickener through the slurry image acquisition system, and collecting the slurry image of known concentration of 1% to 90% through test, establishing a standard sample database, and fitting to determine the mapping relationship between the slurry concentration and the image features;

[0068] S2, image denoising processing: the wavelet transform method of frequency domain denoising is used to carry out denoising processing on the image, so as to improve the speed and accuracy of image recognition.

[0069] Because the noise generated in the image acquisition process will interfere with the extraction of image feature information, and then reduce the speed and accuracy of image recognition, it is necessary to carry out denoising processing on the image; image denoising includes two types of spatial domain denoising and frequency domain denoising, considering the fuzzy characteristics of the slurry image itself, the spatial domain denoising will make the image more fuzzy as a whole, therefore, the wavelet transform method of frequency domain denoising is used.

[0070] S3, image gray processing: the slurry image after denoising processing is subjected to gray processing, so as to realize efficient extraction and utilization of image feature information; the gray processing of the image is to map and transform the original RGB color space to GRAY color space, and the mapping formula is as follows:

[0071] Gray=0.299R+0.587G+0.114B

[0072] Wherein, R, G, B are respectively the pixel values of red, green and blue three channels in RGB color space, and Gray is the gray value in GRAY color space.

[0073] S4, image feature extraction: the histogram of the gray image can be obtained by pixel statistical method; in addition, the color moment of the gray image is calculated based on mathematical method to describe the gray distribution, and the color moment calculation method of the gray image is as follows:

[0074]

[0075]

[0076]

[0077] Wherein, E g is the first moment of the image gray, sigma g is the second moment of the image gray, S g is the third moment of the image gray, p j represents the gray value of the jth pixel, and N is the number of pixels.

[0078] S5, image contrast recognition: first, compare the color moments of each order of the sample image to be tested and the standard sample image, calculate the difference value of the color moments, and sort the standard sample images according to the difference value, select the 10 samples with the smallest difference value, and then calculate the Bhattacharyya coefficient representing the similarity according to the gray histogram of the sample image to be tested and the 10 standard samples, and the formula is as follows:

[0079]

[0080] Wherein, B is the two images of the bar coefficient, p, q are two images in the histogram at the corresponding position x probability distribution.

[0081] In this embodiment, the difference of color moment is obtained by comparing the difference of each order color moment respectively, and then taking the arithmetic mean. If the gray value of the test image is larger, the difference between it and the standard image color moment is positive, and the concentration is greater than the standard image; if the gray value of the test image is smaller, the difference between it and the standard image color moment is negative, and the concentration should be smaller than the standard sample.

[0082] S6, slurry concentration determination: comparing the standard sample with the largest similarity, and comprehensively considering the color moment difference and the bar coefficient, substituting into the slurry concentration-image feature mapping relationship to determine the actual slurry concentration of the test image, the formula is as follows:

[0083]

[0084] Wherein, C p ,C q Respectively, the concentration of the test image and the largest similar image, Δ g The color moment difference between the test image and the largest similar image.

[0085] Further, the specific method of collecting 1% to 90% of the known concentration of slurry image test in step S1 is as follows:

[0086] S11, the slurry with a known concentration of 1% is injected into the container, and the slurry image is obtained by shooting with the image acquisition system;

[0087] S12, referring to steps S2-S4, the slurry image obtained by shooting is sequentially subjected to noise reduction processing, gray processing, and feature extraction;

[0088] S13, sequentially replacing the slurry with a concentration of 1%, 2%, 3%, …, 90%, repeating S11 and S12;

[0089] S14, the gray histogram and the color moment of the obtained 90 standard images are stored in a structured manner, and a standard sample database is established;

[0090] S15, taking the standard sample database as the training data, fitting to establish the slurry concentration-image feature mapping relationship.

[0091] It should be noted that the standard concentration slurry is obtained by adjusting, and the test container is a beaker or other small vessel. Although the slurry is a suspension, it will settle over time, but the image acquisition operation is rapid and does not require a lot of time, therefore, the influence of gravity and depth is considered negligible.

[0092] By continuously collecting slurry images at different positions in the thickener, analyzing and identifying the slurry concentration corresponding to the images according to the above steps, a slurry concentration cloud chart as shown in Figure 7 may be synthesized and drawn, and the overall distribution state of the slurry concentration in the thickener is determined.

[0093] The details of the present application are known. The above description is only to illustrate the technical concept and characteristics of the present application, the purpose is to enable those skilled in the art to understand the content of the present application and to implement it, and cannot limit the protection scope of the present application. Any equivalent changes or modifications made in accordance with the spirit and essence of the present application shall be covered within the protection scope of the present application.

Claims

1. An image recognition based thickener pulp density measurement system, characterized by, The pulp image acquisition system and the pulp image processing and identification system, the image acquisition system uses optical imaging principle to obtain the pulp image of the thickener, through the pulp image processing and identification system, the image noise reduction processing, gray processing, feature extraction and contrast identification are carried out in turn, the determination of the pulp concentration is realized; The pulp image acquisition system includes a camera probe, a fixed pulley, a DC brushless motor, a winding wheel and a PLC controller, the camera probe is connected with the PLC controller through the fixed pulley and the winding wheel by wires to transmit images and sensor data, the DC brushless motor is driven by the PLC controller, the winding wheel is driven by the DC brushless motor, and the outgoing and incoming wires of the fixed pulley control the outgoing and incoming of the wires, thereby realizing the two-dimensional motion of the camera probe in the vertical direction, the camera probe includes an upper shell, a PCB board, an acceleration sensor, an LED ring, a light shield, a glass cover, a convex mirror, a lower shell, a counterweight, a CMOS sensor, a filter, a lens base and a concave lens, the PLC controller transmits the image and acceleration data to the pulp image processing and identification system in the computer through the expansion RJ45 port module for calculation and analysis; The steps of image processing and identification of the pulp image processing and identification system are as follows: S1, obtaining the pulp image of the sample to be measured in the thickener through the pulp image acquisition system, collecting the pulp image with known concentration of 1%~90% through test, establishing the standard sample database, and fitting to determine the mapping relationship between the pulp concentration and the image features; S2, image noise reduction processing: using wavelet transform method of frequency domain noise reduction to reduce the noise of the image; S3, image gray processing: carrying out gray processing on the pulp image after noise reduction to realize efficient extraction and utilization of image feature information; S4, image feature extraction: obtaining the histogram of the gray image through pixel statistical method; Calculating the color moment of the gray image based on mathematical method to describe the gray distribution; S5, image contrast identification: firstly comparing the color moments of the sample image to be measured and the standard sample image, calculating the difference value of the color moments, sorting the standard sample images according to the difference value, selecting the 10 standard sample images with the smallest difference value, and then calculating the Bhattacharya coefficient representing the similarity of the 10 standard sample images and the sample image to be measured according to the gray histograms, the formula is as follows: wherein, are the Bhattacharyya coefficients of the two images, are the probability distributions of the two images at the corresponding positions of the histograms, respectively. are the probability distributions of the two images at the corresponding positions of the histograms, respectively. S6, determination of pulp concentration: comparing and determining the standard sample with the largest similarity, and comprehensively considering the color moment difference value and the Bhattacharya coefficient, substituting into the mapping relationship between the pulp concentration and the image features to determine the actual pulp concentration of the sample image, the formula is as follows: wherein, , respectively are the concentration of the image to be measured and the maximum similar image, is the color moment difference value of the image to be measured and the maximum similar image.

2. The image recognition based thickener pulp density measurement system according to claim 1, wherein, The glass cover of the camera probe is connected with the upper shell and the lower shell through two pairs of radial bolts, the lower shell is connected with the counterweight, the PCB board is arranged in the upper shell, the acceleration sensor and the CMOS sensor are welded to the PCB board through the wire harness, the CMOS sensor is sequentially provided with the filter, the lens base and the concave lens at the lower end, the LED ring is arranged on both sides of the lens base and connected to the PCB board pin through the Dupont wire, the PCB board supplies power to the entire camera probe and transmits signals to the PLC controller.

3. The image recognition based thickener pulp density measurement system according to claim 2, wherein, The lower end of the LED light ring is also provided with a light shield to avoid the light source emitted by the LED light ring directly irradiating the surface of the slurry.

4. The image recognition based thickener pulp density measurement system according to claim 1, wherein, The acceleration sensor senses the motion acceleration of the probe, indirectly realizes high-precision sensing of the probe speed and displacement, calculates the two-dimensional displacement of the probe through accumulation, and determines the depth position of the probe at the imaging time; the conversion calculation formula of acceleration, speed and displacement is as follows: wherein, is the velocity, is the acceleration, is the displacement, is the time.

5. The image recognition based thickener pulp density measurement system according to claim 2, wherein, The shape of the lower shell end matches the shape of the two ends of the convex mirror, and the convex mirror is clamped by screwing the lower shell and the counterweight.

6. The image recognition based thickener pulp density measurement system according to claim 2, wherein, The counterweight is connected with the lower shell through threads, used to pull the probe to realize vertical two-dimensional motion, and different weights of the counterweight can be easily replaced according to the estimated value of the measured slurry concentration, so as to ensure good traction effect.

7. The image recognition based thickener pulp density measurement system according to claim 1, wherein, The gray scale processing of the image is to map and transform the original RGB color space to GRAY color space, and the mapping formula is as follows: wherein, are the pixel values of the red, green, and blue channels in the RGB color space, respectively, is the gray value in the GRAY color space.

8. The image recognition based thickener pulp density measurement system according to claim 1, wherein, The color moment calculation method of the gray image is as follows: ； ； ； wherein, is the first moment of the gray scale of the image, is the second moment of the gray scale of the image, is the third moment of the gray scale of the image, denotes the gray scale value of the pixel, is the number of pixels.

9. The image recognition based thickener pulp density measurement system according to claim 1, wherein, The specific method of collecting 1%-90% known concentration slurry image test in step S1 is as follows: S11, the slurry with a known concentration of 1% is injected into the container, and the slurry image is obtained by shooting with the slurry image acquisition system; S12, referring to steps S2-S4, the obtained slurry image is sequentially subjected to noise reduction processing, gray scale processing and feature extraction; S13, replace the slurry with a concentration of 1%, 2%, 3%, …, 90% in turn, and repeat S11 and S12; S14, the gray scale histogram and color moment of the obtained 90 standard images are stored in a structured manner, and a standard sample database is established; S15, using the standard sample database as training data, fitting and establishing the slurry concentration-image feature mapping relationship.

Images