CN116343002A - Image analysis sewage treatment sludge area sedimentation velocity prediction method - Google Patents
Image analysis sewage treatment sludge area sedimentation velocity prediction method Download PDFInfo
- Publication number
- CN116343002A CN116343002A CN202310127946.3A CN202310127946A CN116343002A CN 116343002 A CN116343002 A CN 116343002A CN 202310127946 A CN202310127946 A CN 202310127946A CN 116343002 A CN116343002 A CN 116343002A
- Authority
- CN
- China
- Prior art keywords
- sludge
- sedimentation
- liquid level
- image
- sedimentation velocity
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume, or surface-area of porous materials
- G01N15/04—Investigating sedimentation of particle suspensions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
- G06V10/422—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation for representing the structure of the pattern or shape of an object therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/1444—Selective acquisition, locating or processing of specific regions, e.g. highlighted text, fiducial marks or predetermined fields
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/16—Image preprocessing
- G06V30/168—Smoothing or thinning of the pattern; Skeletonisation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02W—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO WASTEWATER TREATMENT OR WASTE MANAGEMENT
- Y02W10/00—Technologies for wastewater treatment
- Y02W10/10—Biological treatment of water, waste water, or sewage
Abstract
A method for predicting sedimentation velocity of an image analysis sewage treatment sludge area relates to an intelligent sewage detection method. Collecting an activated sludge sample sedimentation process video; based on the error problem generated by deformation of the liquid level of the measuring cylinder, a method based on the combination of a local threshold and a least square fitting curve is provided, the liquid level of settled sludge is fitted, and the position information of the liquid level is corrected; dividing a digital region on a measuring cylinder by adopting a region dividing and contour extracting method, and sending the digital region into an improved convolutional neural network (LeNet-5) for identification; the method for calculating the scale value of the sludge liquid level surface by combining the recognized characters with the position coordinates of the liquid level surface is provided, so that the scale value of the sludge liquid level surface is obtained; the invention provides a more accurate prediction method for sewage treatment by adopting an activated sludge process.
Description
Technical Field
The invention relates to a sewage detection method, in particular to a method for predicting sedimentation velocity of a sewage treatment sludge region by image analysis.
Background
At present, various industries in China are developing at a high speed, but still face a plurality of serious problems to be solved urgently. The sewage discharged by human factors in daily life contains various microelements which cause water eutrophication, so that the water environment of China is seriously polluted and the water ecology is seriously destroyed, and the serious problem of shortage of available water resources is caused by a series of adverse chain reactions. In the production and life, the problem of water resource shortage brings a lot of adverse effects and harm to people. As is well known, china is a large population country, the annual population number is continuously increased, water is a source of life, and people cannot leave water in daily life, so that the monitoring and treatment of sewage are an important way for improving the water resource shortage of China.
The activated sludge method is one of the common treatment processes of sewage treatment plants, the sedimentation performance of the sludge affects the operation of the whole treatment process, and the sedimentation index can reflect the sedimentation performance of the sludge, so that the efficient and accurate sedimentation index measurement is very important. However, for the measurement of sludge sedimentation performance indexes such as sludge sedimentation ratio, sludge sedimentation speed, sludge volume index and the like, the traditional method is often completely dependent on manual implementation. Because of the existence of subjective judgment factors, the error of the measurement result is larger, the artificial intelligence is continuously developed, and the unique functional characteristics of accuracy, high efficiency, stability and the like are widely applied to various fields, so that the intelligent measurement method has a certain practical research significance in the sewage treatment direction.
Disclosure of Invention
The invention aims to provide a method for predicting sedimentation velocity of an image analysis sewage treatment sludge region, which comprises the steps of carrying out sedimentation experiments on an acquired sludge-water mixture, providing a method for combining a local threshold value with a least square method, fitting the shape of a liquid level surface and determining liquid level coordinate information; dividing characters on a measuring cylinder by using an image preprocessing method, providing an improved convolutional neural network LeNet-5 for recognizing the characters, combining a liquid level coordinate to obtain a scale value of a mud-water interface, and drawing a curve of the integral sedimentation stage of the sludge; and analyzing the obtained index of the sedimentation performance of the reaction sludge, and predicting the sedimentation velocity of the sludge region by adopting a deep random configuration network (deep SCN).
The invention aims at realizing the following technical scheme:
a method for predicting sedimentation velocity of an image analysis sewage treatment sludge region, the method comprising the steps of:
step 1: and (3) image acquisition: collecting an activated sludge sample, and placing the activated sludge sample into a measuring cylinder for video shooting in a sedimentation process;
step 2: image preprocessing: graying treatment, sobel operator edge detection, global threshold segmentation, vertical projection and horizontal projection are carried out on the collected sludge video to determine a measuring cylinder area;
step 3: and (3) liquid level identification: performing morphological operation, selecting scale marks and numbers by using a minimum external rectangular frame, and sending the scale marks and numbers into a neural network (LeNet-5) for recognition;
step 4: measuring a sludge sedimentation rate index: recognizing the liquid level to measure the height value of a sludge layer (SBH), and calculating the Zone Sedimentation Velocity (ZSV);
step 5: extracting sludge floc image characteristics: extracting morphological characteristic parameters of flocs and filamentous fungi;
step 6: and (3) predicting the sludge sedimentation rate: the depth random configuration network is adopted, the extracted floc characteristic is used as the input of a neural network, and the output of the neural network is the predicted regional sedimentation velocity (ZSV).
The image acquisition refers to acquisition of sludge samples at fixed points in an aerobic tank, an anaerobic tank and an anoxic tank, instantaneous water samples are acquired at a water outlet of a reaction tank, the acquired sludge samples are put into an air bath constant temperature oscillator in groups for uniform oscillation, and the sludge samples which are uniformly stirred are poured into an experiment measuring cylinder every other hour for sedimentation experiments; and taking white shading paper as a shooting background of the sedimentation measuring cylinder, and sequentially carrying out video shooting of the sludge sedimentation process with the duration of one hour.
The image preprocessing refers to loading a shot sedimentation video into a computer, cutting out RGB images according to time intervals, carrying out graying and binarization operation on an original image, and converting the cut-out color image into a gray image by using a floating point algorithm.
According to the image analysis sewage treatment sludge area sedimentation velocity prediction method, the original image is subjected to graying and binarization operation, and a global threshold method is adopted to binarize the gray image; determining the position of the ROI (measuring cylinder) by adopting a projection method and extracting the ROI; detecting a scale mark and a digital area on the measuring cylinder by Sobel edge; because the direct binarization can not divide the real liquid level surface, the processing is carried out by using the local binarization.
According to the image analysis sewage treatment sludge area sedimentation velocity prediction method, the image acquisition, the camera and the liquid level surface are not always kept in a vertical state, along with the descending of the muddy water mixed liquid, the height difference of the upper shadow and the lower shadow of the acquired image liquid level surface can accurately position the liquid level surface, the size of the final liquid level value error is influenced, and the method for fitting the shape of the liquid level surface is adopted in combination with the self-adaptive threshold method.
The method for predicting the sedimentation velocity of the sewage treatment sludge region by image analysis is characterized in that the liquid level identification is to binarize scale marks and digital regions on a measuring cylinder by adopting a local threshold value method, and morphological operation is carried out to avoid the influence of some noise.
The method for predicting the sedimentation velocity of the sewage treatment sludge region by image analysis is characterized in that the measurement of the sludge sedimentation velocity index refers to the calculation of the sludge height and the sedimentation velocity after the identification of the segmented numbers.
The method for predicting sedimentation velocity of the sewage treatment sludge region by image analysis is characterized in that the characteristic extraction of the sludge floc image refers to binary characteristics extracted from each target in the floc and filamentous fungus images in sequence, and comprises the steps of extracting morphological characteristics and defining related parameters for each target; and screening out the characteristics with higher sludge sedimentation correlation by adopting typical correlation analysis.
The characteristic parameters are selected as the ten parameters because the characteristic parameters are respectively the area of the flocs, the number of the flocs, the shape factor of the filiform bacteria, the eccentricity of the filiform bacteria, the roundness of the filiform bacteria, the radius of rotation of the filiform bacteria, the eccentricity of the flocs, the aspect ratio of the filiform bacteria, the perimeter of the filiform bacteria and the perimeter of the outline of the flocs.
The method for predicting the sedimentation velocity of the sewage treatment sludge region by image analysis is characterized in that the sludge sedimentation velocity prediction refers to predicting the sedimentation velocity of the sludge region by a depth random configuration network (deep SCN), a prediction model is trained by 180 active sludge phase-contrast microscopic images, each image is ten flocs of area, the number of flocs, the shape factor of filiform bacteria, the eccentricity of filiform bacteria, the roundness of filiform bacteria, the rotation radius of filiform bacteria, the eccentricity of filiform bacteria, the aspect ratio of filiform bacteria, the perimeter of filiform bacteria and the perimeter parameter characteristics of outline lines of the flocs and corresponds to a ZSV value; then using 10 activated sludge phase-contrast microscopic images as a group to test, and taking the average value of the output results of all sample pictures as a ZSV value of 18 days; the ZSV value (yk-1) of the previous day is also inputted as an input characteristic, and eleven characteristics are added.
The invention has the advantages and positive effects that:
1. the invention provides a sludge region sedimentation velocity prediction method based on image analysis, which is based on a method combining a convolutional neural network and an image processing technology, realizes automatic identification of a mud-water interface scale value in a sludge sedimentation video, respectively completes drawing of an integral sedimentation change curve chart and measurement of a sludge sedimentation performance index in a sludge sedimentation process according to the identified scale value, carries out corresponding analysis on floc characteristics which possibly influence the sludge sedimentation performance, takes the floc characteristics as an auxiliary variable to be input into a depth random network, and predicts the layered sedimentation velocity (Zone Settling Velocity, ZSV) of sludge. Aiming at the problem of error caused by deformation of a liquid level surface when a settled sludge video in an experimental measuring cylinder is shot, a method based on combination of a local threshold and a least square fitting curve is designed, the liquid level surface of settled sludge is fitted, and the position information of the liquid level surface is obtained; dividing a digital region on a measuring cylinder by adopting some methods of image preprocessing, and sending the digital region into an improved convolutional neural network (LeNet-5) for identification; designing a method for calculating the scale value of the sludge liquid level surface by combining the recognized characters with the position coordinates of the liquid level surface, and obtaining the scale value of the sludge liquid level surface; measuring the sedimentation ratio of the sludge, the volume index of the sludge and drawing a whole sedimentation curve of the sludge by using the obtained sludge liquid level value, and dividing different sedimentation stages according to the change trend of the curve; predicting the sedimentation velocity of the sludge region by adopting a deep random configuration network (deep SCN), taking eleven characteristic values as the input of a model, and carrying out error analysis on the predicted result, a true value and the commonly used regional sedimentation velocity model prediction.
2. The invention provides a method for predicting the sedimentation velocity of a sludge region based on image analysis, which obtains an integral sedimentation curve of activated sludge by recording a sludge sedimentation video and carrying out a batch sedimentation experiment, and divides different sedimentation stages of the sludge according to the change trend of the sedimentation curve. And measuring parameter indexes reflecting the sedimentation performance of the sludge, wherein the parameter indexes mainly comprise the sedimentation ratio of the sludge, the sedimentation speed of a mud-water interface and the sludge volume index. In order to further explore microscopic elements affecting the sedimentation performance of the sludge, a phase-contrast microscope and a CCD image acquisition system are adopted to acquire a phase-contrast microscopic image of a sludge sample. Dividing marked sludge floc images based on a threshold method, extracting characteristics of the flocs, analyzing main characteristic parameters affecting sludge sedimentation performance, and predicting the sedimentation velocity of a sludge area by adopting a deep random configuration network (deep SCN).
Drawings
FIG. 1 is a flow chart of a method for predicting sedimentation velocity of a sludge region in image analysis according to the invention;
FIG. 2 is a graph showing the comparison of the predicted sludge settling velocity.
Detailed Description
The technical problems involved in the scheme of the invention are described below with reference to the accompanying drawings. It should be noted that the described embodiments are only for better understanding of the present invention and do not serve any limiting purpose.
As can be seen from the flow chart modeled in fig. 1, the overall process includes the following steps:
step 1: and (3) image acquisition: and collecting sludge samples at fixed points in an aerobic tank, an anaerobic tank and an anoxic tank, collecting instantaneous water samples at a water outlet of a reaction tank, putting the collected sludge samples into an air bath constant temperature oscillator in groups for uniform oscillation, and pouring the sludge samples which are uniformly stirred into an experiment measuring cylinder of 100ml every other hour for sedimentation experiments. And taking white shading paper as a shooting background of the sedimentation measuring cylinder, and sequentially carrying out video shooting of the sludge sedimentation process with the duration of one hour.
Step 2: image preprocessing: loading the shot sedimentation video into a computer, and cutting out RGB images according to a certain time interval. Graying and binarizing operation are carried out on the original image, and the truncated color image is converted into a gray image by using a floating point algorithm:
Gray＝0.299*R+0.587*G+0.114*B (1)
wherein Gray is the Gray image obtained, and R, G, B represents the three primary colors of red, green and blue of the color image, respectively. Binarizing the gray image by adopting a global threshold method:
where dst (x, y) is a binarized pixel value, 255 and 0 represent white and black, respectively, src (x, y) represents a pixel value of a gray scale image, and thresh is an experimentally derived global threshold. And determining the position of the ROI (measuring cylinder) by adopting a projection method, extracting the position, determining the range of a segmentation threshold according to the brightness value of the muddy water interface region obtained through experiments, fitting the shape of the interface, and obtaining the coordinates (x, y) of the central point. The Sobel edge detection is adopted to find out the scale mark and the digital area on the measuring cylinder,
formulas (3-5) Gx, gy represent gradient calculations of the image I in the horizontal and vertical directions, respectively, and G represents the result of the final processing. Since the direct binarization cannot divide the real liquid level surface, the processing is performed by local binarization:
where dst (x, y) represents the binarized pixel value, src (x, y) represents the pixel value of the image to be processed, and T (x, y) is the neighborhood mean of the pixel (x, y) minus the empirical constant C. The C value of the experiment is 11, and the color of the flocs floating on the uppermost layer is similar to that of the liquid level surface, and the processed results of the flocs and the liquid level surface are the same, so that the method can be also indicated to be suitable for positioning the liquid level surface.
Because the camera and the liquid level surface are not always kept in a vertical state in the image acquisition process, along with the descent of the muddy water mixed solution, the height difference of the upper shadow and the lower shadow of the acquired image liquid level surface can accurately position the liquid level surface, and the size of the final liquid level value error is affected. The invention provides a method for fitting the shape of a liquid level surface by combining an adaptive threshold method, which comprises the steps of determining the center coordinate of the liquid level surface, identifying a character N closest to the liquid level by combining the distance between h representing the center coordinate and an adjacent complete character, and obtaining a final liquid level value N, wherein the formula is as follows:
since the liquid level surface basically presents an elliptical shape, the shape of the liquid level surface is fitted by adopting a direct least squares elliptical fitting algorithm proposed by AndrewW et al. The algorithm only obtains regularized solution 4ac-b under regularization 2 =1, minimizing the sum of the squares of algebraic distances of points to circles.
The algorithm principle is as follows:
assume a general second order polynomial as follows:
ax 2 +bxy+cy 2 +dx+ey＝1 (8)
let a= [ a, b, c, d, e ]] T ，x＝[x 2 ,xy,y 2 ,x,y] T The above equation can be expressed as ax=1, whereby the expression of the optimization problem can be further deduced:
min Da 2 (9)
sta T Ca＝1 (10)
where D represents the number of samples used. If a is used to represent the parameters of the elliptic equation, C is the following matrix constant:
the Lagrangian product factor is introduced in this calculation to obtain the following equation:
2D T Da-2λCa＝0(11)
a T Ca＝1(12)
let s=d T D, the above equation can be modified as:
Sa＝λCa(13)
a T Ca＝1(14)
because of equation a T Ca=1, so one can find μ, yielding μ 2 u i T Cu i =1, i.e.:
last hypothesis
Lambda is taken out i Feature vector u corresponding to > 0 i And can be used as an equation solution of curve fitting. By means of
The findContours function of OpenCV performs contour detection on the connected region of the measuring cylinder, finds and returns contour information of the binary image and takes the contour information as input of a curve fitting equation
Step 3: and (3) liquid level identification: the method is to binarize the scale marks and the digital areas on the measuring cylinder by adopting a local threshold value method, and morphological operation is carried out to avoid the influence of some noise.
The expansion can be used to fill some holes in the target area, so that the highlight area of the image is gradually enhanced, namely, an operation for obtaining the local maximum value. Corrosion is a local minimum operation that can be used to eliminate small, meaningless objects. The corresponding vector calculation formula is expressed as:
wherein A is a binarized image, and B is a convolution kernel.
The closed operation, i.e. the expansion and then the corrosion, is adopted, which is helpful for removing small holes or small black spots in the foreground object. The communication area of each character is basically square, and similar to the fitting method of the liquid level surface, all outlines are selected by using a minimum circumscribed rectangular frame, and the numerical area nearest to the liquid level is obtained according to the minimum height difference between the ordinate of the character frame and the liquid level surface, wherein the aspect ratio of the character is about 1.2:1. For the extracted characters, the text is identified using a modified convolutional neural network (LeNet-5).
The input sample image is replaced by an image of the original 32×32 size with an image of the input sample 28×28. The traditional LeNet-5 neural network model comprises more fully connected layers, and two fully connected layers in the original neural network are replaced by adopting an SE module through a method of introducing an attention mechanism. The attention mechanism SE module is introduced, so that on one hand, a convolution layer of the network is increased, the depth of the original neural network is expanded, and the learning of image features is more comprehensive. The LeNet-5 network uses a sigmoid activation function, but due to the computational complexity when the network input range is wide, the neuron gradient will reach zero, which will affect the back propagation, resulting in the neural network being untrained. The ReLU activation function may be better suited to fit training data than the Sigmoid activation function, which helps propagate gradients to the next network during back propagation and speeds up the convergence speed of the network model.
Step 4: measuring a sludge sedimentation rate index: and (5) after the segmented numbers are identified, calculating the sludge height and the sedimentation rate. The calculation formula of the height of the sludge layer is as follows:
where N represents the final level value (ml), N represents the number identified, H represents the height difference (ml) between the level center point and the adjacent character, and H represents the height difference (ml) between the two adjacent numbers.
From the height value and time, the sedimentation rate of the sludge can be expressed as:
wherein SV is i Represents the sedimentation rate (ml/s) of the sludge layer, SBH m And SBH (styrene butadiene rubber) n Representing the liquid level values (ml) at times m and n (t), respectively.
Step 5: extracting sludge floc image characteristics: the following binary features are extracted for each target in the floc and filamentous fungus images in turn, and morphological features and definition related parameters are extracted for each target as follows.
TABLE 1 characteristic microcosmic parameters
For the segmented flocs image, there may be many objects in each image, the above parameters (except the total area) calculated for the images of the objects are all averaged, and 17 morphological features are calculated for the segmented flocs, i.e. 17 morphological parameters are extracted for each sample. However, not all morphological parameters have the same influence factor on sludge sedimentation, so that typical correlation analysis is adopted to screen out features with higher correlation with sludge sedimentation.
The first ten features of the arrangement are the area of the flocs, the number of flocs, the form factor of the filamentous fungus, the eccentricity of the filamentous fungus, the roundness of the filamentous fungus, the radius of rotation of the filamentous fungus, the eccentricity of the flocs, the aspect ratio of the filamentous fungus, the perimeter of the filamentous fungus, and the perimeter of the outline of the flocs, respectively, and thus the feature parameters are selected as the ten parameters.
And (3) predicting the sludge sedimentation rate: a depth random configuration network (deep scn) is introduced to predict the sedimentation velocity of a sludge region, a prediction model is trained by 180 activated sludge phase-contrast microscopic images, and each image is the area of ten flocs, the number of flocs, the shape factor of the filiform bacteria, the eccentricity of the filiform bacteria, the roundness of the filiform bacteria, the rotation radius of the filiform bacteria, the eccentricity of the flocs, the aspect ratio of the filiform bacteria, the perimeter parameter characteristics of the contour line of the flocs and corresponds to a ZSV value. Then, 10 activated sludge phase contrast microscopic images are used as a group for testing, and the output results of all sample pictures are averaged to obtain a ZSV value of 18 days. Because the artificial sampling error of the activated sludge is large, the model output and the test target cannot be completely consistent, so that the ZSV value (yk-1) of the previous day is tried to be input as an input characteristic, and eleven characteristics are added.
The activation function of the model is set as a hyperbolic tangent Tanh function, tanh is a deformation of the SIGMOD, but compared with the SIGMOD function, tanh is 0 mean, the extremum of the derivative function is 1, which is more beneficial when facing the gradient disappearance, so it is often used as the activation function of the neural network hidden layer. Depth random configuration network (deep scn) predicts sedimentation velocity of sludge region and compares with conventional Vesilind model and average value of true values is shown in fig. 2.
Claims (10)
1. A method for predicting sedimentation velocity of an image analysis sewage treatment sludge region, the method comprising the steps of:
step 1: and (3) image acquisition: collecting an activated sludge sample, and placing the activated sludge sample into a measuring cylinder for video shooting in a sedimentation process;
step 2: image preprocessing: graying treatment, sobel operator edge detection, global threshold segmentation, vertical projection and horizontal projection are carried out on the collected sludge video to determine a measuring cylinder area;
step 3: and (3) liquid level identification: performing morphological operation, selecting scale marks and numbers by using a minimum external rectangular frame, and sending the scale marks and numbers into a neural network (LeNet-5) for recognition;
step 4: measuring a sludge sedimentation rate index: recognizing the liquid level to measure the height value of a sludge layer (SBH), and calculating the Zone Sedimentation Velocity (ZSV);
step 5: extracting sludge floc image characteristics: extracting morphological characteristic parameters of flocs and filamentous fungi;
step 6: and (3) predicting the sludge sedimentation rate: the depth random configuration network is adopted, the extracted floc characteristic is used as the input of a neural network, and the output of the neural network is the predicted regional sedimentation velocity (ZSV).
2. The method for predicting sedimentation velocity in an image analysis sewage treatment sludge area according to claim 1, wherein the image acquisition means that sludge samples are acquired in an aerobic tank, an anaerobic tank and an anoxic tank at fixed points, instantaneous water samples are acquired at a water outlet of a reaction tank, the acquired sludge samples are put into an air bath constant temperature oscillator in groups for uniform oscillation, and the sludge samples which are uniformly stirred are poured into an experiment measuring cylinder every other hour for sedimentation experiments; and taking white shading paper as a shooting background of the sedimentation measuring cylinder, and sequentially carrying out video shooting of the sludge sedimentation process with the duration of one hour.
3. The method for predicting sedimentation velocity in an image analysis sewage treatment sludge area according to claim 1, wherein the image preprocessing means loading a photographed sedimentation video into a computer, capturing RGB images at time intervals, performing graying and binarizing operations on the original images, and converting the captured color images into gray images by a floating point algorithm.
4. The method for predicting sedimentation velocity of an image analysis sewage treatment sludge region according to claim 3, wherein the operation of graying and binarizing the original image is performed by binarizing the gray image by a global threshold method; determining the position of the ROI (measuring cylinder) by adopting a projection method and extracting the ROI; detecting a scale mark and a digital area on the measuring cylinder by Sobel edge; because the direct binarization can not divide the real liquid level surface, the processing is carried out by using the local binarization.
5. The method for predicting sedimentation velocity in an image analysis sewage treatment sludge area according to claim 4, wherein the image acquisition, the camera and the liquid level surface are not always kept in a vertical state, as the muddy water mixed liquid descends, the height difference of the upper shadow and the lower shadow of the acquired image liquid level surface can accurately position the liquid level surface, the size of the final liquid level value error is affected, and a method for fitting the shape of the liquid level surface is adopted in combination with an adaptive threshold method.
6. The method for predicting sedimentation velocity in an image analysis sewage treatment sludge region according to claim 1, wherein the liquid level identification means that a local threshold value method is adopted to binarize scale marks and a digital region on a measuring cylinder, and morphological operation is performed to avoid influence of some noise.
7. The method for predicting the sedimentation velocity of an image analysis sewage treatment sludge region according to claim 1, wherein the measurement of the sedimentation velocity index is to perform the calculation of the sludge height and the sedimentation velocity after the recognition of the divided numbers.
8. The method for predicting sedimentation velocity in an image analysis sewage treatment sludge region according to claim 1, wherein the sludge floc image feature extraction means binary features extracted sequentially for each target in the floc and filamentous fungus images, including morphological feature extraction and related parameter definition for each target; and screening out the characteristics with higher sludge sedimentation correlation by adopting typical correlation analysis.
9. The method for predicting sedimentation velocity in an image analysis sewage treatment sludge region according to claim 8, wherein the first ten features of the arrangement are respectively an area of flocs, a number of flocs, a shape factor of filamentous fungi, an eccentricity of filamentous fungi, a roundness of filamentous fungi, a radius of rotation of filamentous fungi, an eccentricity of flocs, an aspect ratio of filamentous fungi, a perimeter of filamentous fungi, and a perimeter of outline of flocs, and thus the feature parameters are selected as the ten parameters.
10. The method for predicting sedimentation velocity in an image analysis sewage treatment sludge region according to claim 1, wherein the prediction of sedimentation velocity of sludge refers to predicting sedimentation velocity in a sludge region by a depth random configuration network (deep scn), wherein a prediction model is trained by 180 active sludge phase contrast microscopic images, each image is an area of ten flocs, the number of flocs, a shape factor of a filamentous fungus, an eccentricity of the filamentous fungus, roundness of the filamentous fungus, a radius of rotation of the filamentous fungus, eccentricity of the flocs, aspect ratio of the filamentous fungus, perimeter parameter characteristics of contour lines of the flocs, and corresponds to one ZSV value; then using 10 activated sludge phase contrast microscopic images as a group for testing, and taking all sample pictures
Taking the average value as the ZSV value of 18 days; the ZSV value of the previous day (yk-1) is also input as an input feature,
there are eleven features in total.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310127946.3A CN116343002A (en) | 2023-02-17 | 2023-02-17 | Image analysis sewage treatment sludge area sedimentation velocity prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310127946.3A CN116343002A (en) | 2023-02-17 | 2023-02-17 | Image analysis sewage treatment sludge area sedimentation velocity prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116343002A true CN116343002A (en) | 2023-06-27 |
Family
ID=86886685
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310127946.3A Pending CN116343002A (en) | 2023-02-17 | 2023-02-17 | Image analysis sewage treatment sludge area sedimentation velocity prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116343002A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117218586A (en) * | 2023-09-21 | 2023-12-12 | 北京市自来水集团有限责任公司技术研究院 | Image recognition method, device, equipment and medium for measuring sedimentation velocity of suspended matters |
-
2023
- 2023-02-17 CN CN202310127946.3A patent/CN116343002A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117218586A (en) * | 2023-09-21 | 2023-12-12 | 北京市自来水集团有限责任公司技术研究院 | Image recognition method, device, equipment and medium for measuring sedimentation velocity of suspended matters |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11836976B2 (en) | Method for recognizing seawater polluted area based on high-resolution remote sensing image and device | |
| CN111626190B (en) | Water level monitoring method for scale recognition based on clustering partition | |
| CN106650812B (en) | A kind of urban water-body extracting method of satellite remote-sensing image | |
| CN115601374A (en) | Chromosome image segmentation method | |
| JP2017521779A (en) | Detection of nuclear edges using image analysis | |
| CN114067114B (en) | Area nuclear segmentation counting method based on algae cell morphology | |
| CN112785560A (en) | Air tightness detection water body updating method and system based on artificial intelligence | |
| CN107730499A (en) | A kind of leucocyte classification method based on nu SVMs | |
| CN111369526B (en) | Multi-type old bridge crack identification method based on semi-supervised deep learning | |
| CN116343002A (en) | Image analysis sewage treatment sludge area sedimentation velocity prediction method | |
| CN114882040B (en) | Sewage treatment detection method based on template matching | |
| CN116310845B (en) | Intelligent monitoring system for sewage treatment | |
| CN115249246A (en) | Optical glass surface defect detection method | |
| CN116309599B (en) | Water quality visual monitoring method based on sewage pretreatment | |
| CN112991271A (en) | Aluminum profile surface defect visual detection method based on improved yolov3 | |
| CN110298410A (en) | Weak target detection method and device in soft image based on deep learning | |
| CN114549446A (en) | Cylinder sleeve defect mark detection method based on deep learning | |
| CN115131356A (en) | Steel plate defect classification method based on richness | |
| CN116883408A (en) | Integrating instrument shell defect detection method based on artificial intelligence | |
| CN110751150A (en) | FPGA-based binary neural network license plate recognition method and system | |
| CN115330792A (en) | Sewage detection method and system based on artificial intelligence | |
| CN115187788A (en) | Crop seed automatic counting method based on machine vision | |
| CN115359003A (en) | Two-step tunnel gray image crack identification method, system, medium and equipment | |
| CN115496716A (en) | Single and double micronucleus cell image detection method based on deep learning and related equipment | |
| CN114897823A (en) | Cytology sample image quality control method, system and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |