CN113688751A - Method and device for analyzing alum blossom characteristics by using image recognition technology - Google Patents
Method and device for analyzing alum blossom characteristics by using image recognition technology Download PDFInfo
- Publication number
- CN113688751A CN113688751A CN202111001467.4A CN202111001467A CN113688751A CN 113688751 A CN113688751 A CN 113688751A CN 202111001467 A CN202111001467 A CN 202111001467A CN 113688751 A CN113688751 A CN 113688751A
- Authority
- CN
- China
- Prior art keywords
- image
- alum blossom
- alum
- model
- blossom
- 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
- 229940037003 alum Drugs 0.000 title claims abstract description 242
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000005516 engineering process Methods 0.000 title claims abstract description 18
- 238000013139 quantization Methods 0.000 claims abstract description 59
- 238000012549 training Methods 0.000 claims abstract description 40
- 238000011156 evaluation Methods 0.000 claims abstract description 4
- 238000001514 detection method Methods 0.000 claims description 27
- 238000011158 quantitative evaluation Methods 0.000 claims description 16
- 238000013528 artificial neural network Methods 0.000 claims description 15
- 230000011218 segmentation Effects 0.000 claims description 14
- 230000004913 activation Effects 0.000 claims description 9
- 238000002372 labelling Methods 0.000 claims description 9
- 239000000284 extract Substances 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 229910052720 vanadium Inorganic materials 0.000 claims description 4
- LEONUFNNVUYDNQ-UHFFFAOYSA-N vanadium atom Chemical compound [V] LEONUFNNVUYDNQ-UHFFFAOYSA-N 0.000 claims description 4
- 229940088594 vitamin Drugs 0.000 claims description 4
- 229930003231 vitamin Natural products 0.000 claims description 4
- 235000013343 vitamin Nutrition 0.000 claims description 4
- 239000011782 vitamin Substances 0.000 claims description 4
- 150000003722 vitamin derivatives Chemical class 0.000 claims description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 43
- 238000004364 calculation method Methods 0.000 description 15
- 238000004062 sedimentation Methods 0.000 description 15
- 239000000701 coagulant Substances 0.000 description 14
- 230000008569 process Effects 0.000 description 14
- 238000010586 diagram Methods 0.000 description 11
- 230000015271 coagulation Effects 0.000 description 9
- 238000005345 coagulation Methods 0.000 description 9
- 230000003287 optical effect Effects 0.000 description 8
- 238000012360 testing method Methods 0.000 description 8
- 238000004458 analytical method Methods 0.000 description 7
- 238000002474 experimental method Methods 0.000 description 7
- 238000005189 flocculation Methods 0.000 description 7
- 230000016615 flocculation Effects 0.000 description 7
- 230000006870 function Effects 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 210000002569 neuron Anatomy 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 238000010801 machine learning Methods 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 238000000746 purification Methods 0.000 description 4
- 239000010419 fine particle Substances 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000011002 quantification Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000019771 cognition Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 239000013078 crystal Substances 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000005484 gravity Effects 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 239000012535 impurity Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 235000011837 pasties Nutrition 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 241000282326 Felis catus Species 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001647 drug administration Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 230000004899 motility Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06N3/045—Combinations of networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a device for analyzing alum blossom characteristics by utilizing an image recognition technology, wherein the method comprises the following steps: step S1, collecting an alum blossom image by using an underwater camera, marking the collected image, generating training data, establishing an alum blossom recognition AI model, and performing autonomous training on the alum blossom recognition AI model by using the marking data as a supervision signal to obtain a trained alum blossom recognition AI model; step S2, collecting an alum blossom image by using an underwater camera, and identifying by using the trained alum blossom identification AI model; and step S3, dividing the single-frame image into M multiplied by M areas, respectively calculating alum blossom quantization indexes for the single areas, and combining various alum blossom quantization indexes for the single areas to obtain the quantization evaluation index of the single-frame image.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method and a device for analyzing alum blossom characteristics by using an image recognition technology.
Background
The coagulation effect directly influences the control of the subsequent process, and the effective coagulation dosage control method can control the addition of the coagulant in the sense of the actual optimal dosage, thereby achieving the effect of obtaining the optimal effluent quality with the least medicament consumption. Coagulation chemical dosing is an important link of water treatment and has the characteristics of large lag, nonlinearity and the like. How to realize the automatic control of coagulation dosing is always a concern in the water production industry. The conventional chemical dosing control methods such as a mathematical model method, a flow current method, a simulated filter method and the like have no scale popularization and application because of the limitations and unreliability.
At present, each water plant still combines the manual observation of alum blossom effect and an outlet turbidimeter of a sedimentation tank to check the dosing effect according to the water flow entering the plant, and the dosing amount of a coagulant is mainly controlled by the flow ratio, so that the accurate control and detection of the water quality cannot be realized. The fixed programmed automatic control system of the flow ratio is adopted, the test result is only representative of the water quality at the sampling moment, the determined coagulant dosage has the problems of discontinuity and hysteresis, the optimal control of the coagulant dosage is difficult to realize in the running process of a water plant, and the accurate control of the coagulant dosage cannot be realized.
In order to effectively form a more objective solution which integrates various factors and systematizes the subjective motility and experience of people. In recent years, intelligent control of coagulant addition is always the focus of water service workers, and the comprehensive application of an intelligent control method and an advanced coagulation control technology is more and more inclined so as to realize the optimal control of coagulant addition. An underwater particle shooting technology and a computer artificial intelligent control method are a better research direction for realizing intelligent control of coagulation drug administration. With the rise of artificial intelligence, especially deep learning, deep convolutional neural network models can be effectively applied to the fields of image recognition, object monitoring and the like. Not only can extract the texture characteristics of the alum blossom image, but also can realize the purpose of identifying the alum blossom image. And the collected alum blossom images can be subjected to threshold segmentation and morphological processing, main image features are screened out from a plurality of extracted image features to be subjected to standardization processing, the standardized features are trained by combining an SVM algorithm, a BP neural network and a GRNN neural network, and whether the alum feeding amount is appropriate or not is judged. The computer vision realized through deep learning can help people to supervise the production environment, identify product flaws and identify fault hidden dangers, so the technology basically has the condition for identifying the alum blossom in the water making process of a water plant and helps to realize more precise and intelligent analysis of the alum blossom image. The active, automatic and self-learning system is used in the process link of the water production process to improve the efficiency, reduce the loss of electricity and medicine, achieve the accurate control of coagulation effect and the real-time monitoring of water quality, and have important significance for improving the management level of a water plant.
The traditional alum blossom detection algorithm based on machine learning and image processing mainly comprises a digital image processing module, a feature engineering module and a machine learning module, as shown in fig. 1. In the algorithm complexity degree, the algorithm is in a logic serial state and needs to repeat feature engineering and machine learning until corresponding scenes are met, so that the algorithm is low in operation speed and large in repeated workload; in the aspect of robustness, when a scene changes, an algorithm needs to be redesigned, and the algorithm robustness is poor.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention aims to provide a method and a device for analyzing alum blossom characteristics by using an image recognition technology, so as to realize intelligent control of coagulant addition and realize accurate coagulant addition.
In order to achieve the above object, the present invention provides a method for analyzing alum blossom characteristics by using image recognition technology, comprising the following steps:
step S1, collecting an alum blossom image by using an underwater camera, marking the collected image, generating training data, establishing an alum blossom recognition AI model, and performing autonomous training on the alum blossom recognition AI model by using the marking data as a supervision signal to obtain a trained alum blossom recognition AI model;
step S2, collecting an alum blossom image by using an underwater camera, and identifying by using the trained alum blossom identification AI model;
and step S3, dividing the single-frame image into M multiplied by M areas, respectively calculating alum blossom quantization indexes for the single areas, and combining various alum blossom quantization indexes for the single areas to obtain the quantization evaluation index of the single-frame image.
Preferably, the step S1 further includes:
s100, collecting alum blossom images within the field depth range by using an underwater camera;
step S101, determining alum blossom labeling standards of various categories, marking collected images, and generating training data;
step S102, an alumen ustum recognition AI model of the multi-view alumen ustum detection segmentation network is established, marking data are used as supervision signals, the multi-view alumen ustum detection segmentation network is trained, and the marking data are fitted.
Preferably, in step S101, marking criteria of three categories of alumen ustum are determined according to the related business of the alumen ustum and the shot image, marking the collected image according to the determined marking criteria, and adding remark data to the image in which fluffy alumen ustum, sheet alumen ustum, alumen ustum and fuzzy fluffy image cannot be determined.
Preferably, in step S102, the AI model extracts high-level features of the image through multilayer convolution and activation layers by using a multilayer neural network, and obtains a detection frame of the alum blossom through regression task regression by using the extracted high-level features of the image, and obtains the type of the alum blossom through classification task classification by using the high-level features of the image.
Preferably, in step S102, the AI model is trained by using backpropagation optimization model parameters through a large amount of alum blossom data.
Preferably, in step S2, the trained AI model is used to identify the captured image, and attributes such as category, probability, and location of each alumen ustum are detected.
Preferably, the quantitative index of each alumen ustum class in the single region comprises the number, confidence average, confidence median, area average and area median of the class in the region.
Preferably, in step S3, the three kinds of alum blossom quantization indexes of all regions of M × M are combined to obtain 15M2And the vanadium flower quantitative index of the vitamin is the quantitative evaluation index of the single-frame image.
Preferably, in step S3, F frames are selected at equal intervals from the image collected within 1 minute, the alum blossom quantization index of the single frame image is calculated, and then the same alum blossom quantization indexes of the F frames are averaged to obtain the alum blossom quantization index after time quantization.
In order to achieve the above object, the present invention further provides an apparatus for analyzing alum blossom characteristics by using an image recognition technique, comprising:
the device comprises an alum blossom identification AI model building and training unit, a parameter setting unit and a parameter setting unit, wherein the alum blossom identification AI model building and training unit is used for collecting an alum blossom image by using an underwater camera, marking the collected image, generating training data, building an alum blossom identification AI model, and performing autonomous training on the alum blossom identification AI model by using the marking data as a supervision signal to obtain a trained alum blossom identification AI model;
the image acquisition and identification unit is used for acquiring an alum blossom image by using an underwater camera and identifying by using the trained alum blossom identification AI model;
and the quantitative evaluation index detection unit is used for dividing the single-frame image into M multiplied by M areas, respectively calculating the alum blossom quantitative indexes of the single area, and combining the three types of alum blossom quantitative indexes of the single area to obtain the quantitative evaluation index of the single-frame image.
Preferably, the quantitative evaluation index detection unit selects F frames at equal intervals from the image acquired within 1 minute, calculates an alum blossom quantitative index of a single frame image, and then averages the same alum blossom quantitative indexes of the F frames to obtain an alum blossom quantitative index after time quantization.
Compared with the prior art, the method and the device for analyzing the alum blossom characteristics by using the image recognition technology realize the recognition of the alum blossom image by constructing the alum blossom recognition AI model and detect the alum blossom quantitative index, so that whether the alum throwing amount is proper or not can be judged, the intelligent control of coagulant adding is realized, and the accurate adding of the coagulant is realized.
Drawings
FIG. 1 is a diagram of a conventional alum blossom detection algorithm based on machine learning and image processing;
FIG. 2a is a block diagram of a sensor (neuron) according to the present invention;
FIG. 2b is a block diagram of the multi-layered sensor of the present invention;
FIG. 2c is a schematic diagram of the learning (training) process of the multi-layered sensor of the present invention;
FIG. 3 is a flowchart illustrating the steps of a method for analyzing alum blossom features using image recognition technology according to the present invention;
FIG. 4 is a schematic view of a raw water purification process;
FIG. 5 is a schematic view of flake alum, fluffy alum and fuzzy fluffy in the embodiment of the present invention;
FIG. 6 is a schematic view illustrating depth of field of a simple optical system according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of three types of alum blossom labeling standards in the embodiment of the present invention;
FIG. 8 is a schematic structural diagram of a multi-layer neural network used in the AI model for alum blossom identification according to an embodiment of the present invention;
FIG. 9 is a schematic diagram of a multi-view alumen ustum detection segmentation network according to an embodiment of the present disclosure;
FIG. 10 is a schematic diagram of dividing a single frame image into 2 × 2 divided images according to an embodiment of the present invention;
FIG. 11 is a diagram illustrating a calculation of a single-region alumen ustum quantization index according to an embodiment of the present disclosure;
FIG. 12 is a system architecture diagram of an apparatus for analyzing alum blossom features using image recognition technology according to the present invention;
FIG. 13 is a schematic view showing the importance of each indicator of alum blossom quantification to turbidity of settled water in the embodiment of the present invention;
FIG. 14 is a schematic view showing the importance of different types of alum blossom in all regions according to the present invention.
Detailed Description
Other advantages and capabilities of the present invention will be readily apparent to those skilled in the art from the present disclosure by describing the embodiments of the present invention with specific embodiments thereof in conjunction with the accompanying drawings. The invention is capable of other and different embodiments and its several details are capable of modification in various other respects, all without departing from the spirit and scope of the present invention.
Before the present invention is introduced, the basic principle of the convolutional neural network is introduced:
the structure of the perceptron (neuron) is shown in fig. 2a, and specifically as follows:
the perceptron contains a set of parameters, including linear and non-linear calculations;
the perceptron is a neural network with a single artificial neuron, having an input layer, and a set of connections connecting the input unit and the output unit;
the goal of the perceptron is to classify the pattern provided to the input unit.
The basic operation performed by the output unit is to multiply each input (xn) by its connection strength or weight (wn) and pass the sum of the products to the output unit;
weighted sum of inputs
The result of the comparison with the threshold value θ is passed to a step function (activation function). If the sum exceeds a threshold, the step function outputs a "1", otherwise a "0" is output.
For example, the input may be pixels in an image, or more commonly, features extracted from the original image, such as contours of objects in the image. Each time an image is input, the sensor will determine whether the image is a member of a certain category, such as cats. The output can only be one of two states, on if the image is in the category, and off otherwise. "on" and "off" correspond to 1 and 0, respectively, in the binary value.
The structure of the multilayer sensor is shown in fig. 2b, and specifically comprises the following steps:
a feedforward network comprising a plurality of levels of sensors, the feedforward network calculated layer-by-layer:
the input value is transmitted forward from the neuron of the input layer by the weighted connection layer by layer, passes through the hidden layer and finally reaches the output layer to obtain output. In the forward propagation process of the signal, the weight of the network is fixed and unchanged, and the state of each layer of neurons only affects the state of the next layer of neurons. The process is as follows:
the learning (training) process of the multi-layer perceptron is shown in fig. 2c, which is as follows:
forward propagation as computing output by the perceptron;
backpropagation is a process of training the sensor parameters, the input to the backpropagation being a feedforward that is propagated forward;
the output result is compared with the value given by the trainer, and the difference is used to update the weight of the connected output unit to reduce the error;
the weights between the input unit and the hidden layer are updated by back-propagating the error according to how much each weight contributes to the error.
Training with a large number of samples, the concealment units generate selective features that can distinguish between different input modes, so that they can distinguish between different classes in the output layer.
FIG. 3 is a flowchart illustrating the steps of a method for analyzing alum blossom features by using image recognition technology according to the present invention. As shown in fig. 3, the method for analyzing alum blossom characteristics by using image recognition technology of the present invention comprises the following steps:
and step S1, acquiring an alum blossom image by using the underwater camera, marking the acquired image, generating training data, establishing an alum blossom recognition AI model, and performing autonomous training on the alum blossom recognition AI model by using the marking data as a supervision signal so as to obtain the trained alum blossom recognition AI model.
Generally, raw water is purified mainly in a flocculation tank, a sedimentation tank and a filter tank, as shown in fig. 4, a coagulant is added into the flocculation tank to generate granular crystal alum flocs through chemical reaction, water flow enters the sedimentation tank through the flocculation tank, alum flocs are basically generated and completed at a water inlet of the sedimentation tank, the granular alum flocs are continuously combined with each other, move and adsorb raw water impurities in the coagulation tank and the sedimentation tank, the shape of the granular alum flocs is from granular to flaky, the flaky alum flocs are mutually moved and adsorbed to form fluffy alum flocs, the density of the alum flocs is greater than that of water, and finally the fluffy alum flocs are precipitated to the bottom of the sedimentation tank due to the action of gravity to complete primary water purification and reduce the turbidity of the raw water. The flakes are classified into flaky flakes and fluffy flakes according to their shapes by alum blossom business analysis. However, as the parameters such as the focal length of the camera are fixed, the alum blossom close to the camera appears pasty on the image, so that the specific form of the alum blossom can not be distinguished by human eyes, and the alum blossom cannot be simply classified into sheet-shaped and fluffy alum blossom, but needs to be separately classified into fuzzy fluffy alum blossom to be used as a supplementary category of the alum blossom. Therefore, considering the business of alum blossom and the influence of a camera, in the alum blossom recognition AI model, alum blossom is divided into: flake-like alumen ustum, fluffy alumen ustum, and fuzzy fluffy as shown in fig. 5.
Specifically, step S1 further includes:
and S100, collecting alum blossom images within the field depth range by using an underwater camera.
Taking the simple optical system of fig. 6 as an example to introduce the stereo range of the underwater camera, theoretically, an optical system with a positive focal length can realize infinite imaging, but cannot realize infinite imaging of the camera due to optical path energy loss and CCD (optical sensor) phase element size. The minimum diameter of the dispersed spot supported by the camera is delta, and the depth of field of the optical system of the camera can be calculated through the focal length of the camera, the size of the phase element and the related parameters of the optical system. The depth of field has its physical meaning: the object can obtain a relatively clear image in an image plane (CCD) within the depth of field of the object plane. Therefore, the underwater camera shoots alum blossom in a cuboid with the range near the depth of field, and in the specific embodiment of the invention, the underwater camera adopts two cameras, is placed at the water inlet of the sedimentation tank, is installed on a three-pipe supporting plate, shoots at an angle of 80cm away from the water surface and inclines downwards at 45 degrees.
And S101, determining various categories of alum blossom labeling standards, marking the collected images, and generating training data.
Before the AI model is established, a certain amount of marking data is needed to be used as a supervision signal, the AI model is trained autonomously, and the marking data is fitted, so that the marking data is an important link of the AI model. The alum blossom is distinguished with finer granularity by combining alum blossom business and shooting conditions. On the label, a problem is introduced: how to distinguish various categories of alum blossom according to shot images.
In actual conditions, both flake-shaped and fluffy alumen ustum are artificially defined, the classification of alumen ustum is not classified into fine particle sizes in the previous research, the two alumen ustums are not strictly and clearly defined in the industry, and a clear boundary line is not clear in image vision, and meanwhile, in the prior work, the alumen ustum is only observed on the side of a sedimentation tank through human eyes. Therefore, in the invention, a set of relatively complete labeling standards of 3 types of alum flowers are determined according to the related business of the alum flowers and the shot images, as shown in fig. 7, so that the collected images are marked according to the determined labeling standards to generate training data.
In fig. 7, it can be seen that in the labeling standard, there is a remark note indicating that the fluffy flakes, the flaky flakes, the fluffy flakes, and the fuzzy fluffy flakes cannot be identified on the image. For remark data, before data preprocessing, according to the label records of the remark data, probability values of corresponding labels are given, and category uncertainty of the remark data is represented. The purpose is as follows: (1) the method accords with human cognition, and guides an alum blossom recognition AI model to learn human universality cognition on alum blossom; (2) and introducing alumen ustum category uncertainty in a data layer, and enhancing the robustness of the model.
Step S102, an alumen ustum recognition AI model of the multi-view alumen ustum detection segmentation network is established, marking data are used as supervision signals, the multi-view alumen ustum detection segmentation network is trained, and the marking data are fitted.
In the specific embodiment of the present invention, the alumen ustum recognition AI model employs a multilayer neural network, as shown in fig. 8, in the multilayer neural network, a convolutional layer + activation function can be regarded as a perceptron, the entire network includes a multilayer perceptron, and features are continuously extracted and compressed through a plurality of convolutional layer + activation functions (perceptrons), so as to finally obtain higher-level features, that is, original features are concentrated one by one through the multilayer neural network, so that the finally obtained features are more reliable, and finally, various tasks can be performed by using the last-level features: such as classification (e.g., determining a class), regression (e.g., regression out of a detection box), etc.
In the invention, the operating principle of the alum blossom identification AI model is mainly as follows:
1. extracting high-level features of the image through a plurality of layers of convolution and activation layers;
2. regression is carried out by utilizing the high-grade characteristics of the image through a regression task to obtain a detection frame (position and size) of the alum blossom;
3. by utilizing the high-level characteristics of the images and through a classification task, the types of the alum flocs (fluffy, flaky and fuzzy alum flocs) are obtained by classification.
The training process is as follows:
collecting a large number of alumen ustum images and marking manually (marking position, size and type of alumen ustum)
Training a alumen ustum model (multilayer neural network) by using back propagation optimization model parameters through a large amount of alumen ustum data.
Compared with the traditional machine learning algorithm, the deep learning-based alum blossom detection segmentation algorithm does not need manual adjustment of the algorithm, and can automatically extract and combine features. The invention deposits a detection segmentation algorithm and an industrial visual platform in actual combat, and introduces actual combat experience and academia into the algorithm; in the aspect of video information, the invention can improve the extraction and integration of the video information and improve the precision and robustness of the detection segmentation algorithm by adopting an optical flow method and a video spatiotemporal feature extraction technology. As shown in fig. 9, the present invention can detect alum flocs well under different underwater shooting conditions by a multi-view alum floc detection and segmentation algorithm based on deep learning.
And step S2, acquiring an alum blossom image by using an underwater camera, and recognizing by using the trained alum blossom recognition AI model.
In the invention, the trained alumen ustum recognition AI model can be used for detecting attributes such as each detected alumen ustum type, probability, position and the like in the shot image. The list of detected attributes is as follows:
TABLE 1 Alum blossom correlation attribute table for testing and segmenting network
And step S3, dividing the single-frame image into M multiplied by M areas, respectively calculating alum blossom quantization indexes for the single areas, and combining the 3 types of alum blossom quantization indexes for the single areas to obtain the quantization evaluation indexes of the single-frame image, wherein the quantization indexes of each alum blossom category of the single areas comprise the number, the confidence average, the confidence median, the area average and the area median of the category in the area.
Since the alumen ustum distribution is not uniform in the shot image area and the alumen ustum distribution information is retained, the single frame image is divided into M × M areas before the alumen ustum characteristics are quantized. Taking 2 × 2 as an example, as shown in fig. 10, the entire image is divided into 4 regions as shown in the figure in a 2 × 2 format, and then alum blossom quantization indexes are calculated for the respective regions.
Calculating the alumen ustum quantization index of the single frame and the single area: the calculation method of the alumen ustum quantization index in a single frame and single region will be described by taking the upper left region of fig. 10 as an example. As shown in FIG. 11, for a single region, the number, the confidence average, the confidence median, the area average and the area median of the region of each alumen ustum category are calculated.
Specifically, the confidence average and the confidence median are calculated as follows:
wherein,
the mean confidence of the type c alumen ustum is shown,
representing the confidence median of the c-type alum blossom, NcThe number of c-type alumen ustum in this region is shown.
Area average and area median calculation mode:
wherein,
the average area of the c-type alumen ustum,
the median of the area of the c-type alumen ustum is shown, and A is the area of the region.
Therefore, the quantization index of each alum blossom class in a single region is as follows:
wherein c represents alum blossom category, and i represents division area.
In the invention, the physical significance represented by each index of alum blossom is as follows:
(1) the number is as follows: detecting the number of alum blossom in a certain type of current area;
(2) confidence mean, median confidence: the confidence, namely the probability of the detected alum flowers is predicted by the alum flower recognition AI model, and the confidence coefficient that the detected alum flowers are in the current category is represented;
(3) area average, area median: the ratio of the sum of the areas of the detected alumen ustum in a certain category of the current region relative to the area of the region represents the density of the alumen ustum in the certain category of the region.
As described above, the single region has 5 quantitative evaluation indexes per type of alumen ustum, and the total of 5 × 3 — 15 alumen ustum indexes in the single region can be obtained by classifying the alumen ustum type.
Single-frame multi-region alumen ustum quantization index combination: combining the 3-class alumen ustum quantization indexes of all the M multiplied by M areas to obtain 15M2And the vanadium flower quantitative index of the vitamin is the quantitative evaluation index of the single-frame image.
In the example of division by 2 × 2 in fig. 10, the number of extracted alum blossom quantization indexes is 15 × 2 × 2 — 60.
And (3) quantifying alum blossom quantifying index time: because the raw water data is updated once in 1 minute, and because of the FPS characteristics of the camera, there are about 360 frames of images in 1 minute, if all the images participate in the calculation of the alum blossom quantization index, it will result in: (1) each frame of image needs to be subjected to calculation of an alum blossom identification AI model and single-frame multi-region alum blossom quantitative evaluation, and the calculation complexity is overlarge; (2) adjacent frames have similar alum blossom quantization indexes and redundant features.
Therefore, in order to actually calculate the 1-minute alum blossom quantification index, time must be quantified by the following method: selecting F frames from the images collected within 1 minute at equal intervals, calculating the alum blossom quantization index of a single frame image, and finally averaging the same alum blossom quantization indexes of the F frames to obtain the alum blossom quantization index after time quantization.
FIG. 12 is a system architecture diagram of an apparatus for analyzing alum blossom features by using image recognition technology according to the present invention. As shown in fig. 12, the present invention provides an apparatus for analyzing alum blossom characteristics by using image recognition technology, comprising:
the alum blossom identification AI model building and training unit 10 is used for collecting alum blossom images by using an underwater camera, marking the collected images, generating training data, building an alum blossom identification AI model, and performing autonomous training on the alum blossom identification AI model by using the marking data as a supervision signal to obtain the trained alum blossom identification AI model.
Generally, raw water is purified mainly in a flocculation tank, a sedimentation tank and a filter tank, coagulant is added into the flocculation tank to generate granular crystal alum floc through chemical reaction, water flow enters the sedimentation tank through the flocculation tank, alum floc is basically generated and completed at a water inlet of the sedimentation tank, the granular alum floc is continuously combined with each other in the flocculation tank and the sedimentation tank, moves and adsorbs raw water impurities, the form of the granular alum floc is from granular to flaky, the flaky alum floc is internally and mutually moved and adsorbed to form fluffy alum floc, the density of the alum floc is greater than that of water, and finally the fluffy alum floc is precipitated to the bottom of the sedimentation tank due to the action of gravity to complete primary water purification and reduce the turbidity of raw water. The flakes are classified into flaky flakes and fluffy flakes according to their shapes by alum blossom business analysis. However, as the parameters such as the focal length of the camera are fixed, the alum blossom close to the camera appears pasty on the image, so that the specific form of the alum blossom can not be distinguished by human eyes, and the alum blossom cannot be simply classified into sheet-shaped and fluffy alum blossom, but needs to be separately classified into fuzzy fluffy alum blossom to be used as a supplementary category of the alum blossom. Therefore, considering the business of alum blossom and the influence of a camera, in the alum blossom recognition AI model, alum blossom is divided into: flake-like alumen ustum, fluffy alumen ustum, and fuzzy fluffy as shown in fig. 5.
Specifically, the alum blossom identification AI model building and training unit 10 further includes:
and the image acquisition module is used for acquiring the alum blossom image within the field depth range by utilizing the underwater camera.
In the invention, the shooting range of the underwater camera is alum blossom in a cuboid near the depth of field.
And the marking module is used for determining various categories of alum blossom marking standards, marking the acquired images and generating training data.
Before the AI model is established, a certain amount of marking data is needed to be used as a supervision signal, the AI model is trained autonomously, and the marking data is fitted, so that the marking data is an important link of the AI model. The alum blossom is distinguished with finer granularity by combining alum blossom business and shooting conditions. On the label, a problem is introduced: how to distinguish various categories of alum blossom according to shot images
In actual conditions, both flake-shaped and fluffy alumen ustum are artificially defined, the classification of alumen ustum is not classified into fine particle sizes in the previous research, the two alumen ustums are not strictly and clearly defined in the industry, and a clear boundary line is not defined in image vision, and meanwhile, teachers of water plant professionals only observe the alumen ustum on the side of a sedimentation tank through human eyes in the past work. Therefore, in the invention, a set of relatively complete labeling standards of 3 types of alum flocs is finally determined according to the related business of the alum flocs and the shot images, so that the collected images are labeled according to the determined labeling standards to generate training data.
And the training module is used for establishing an alumen ustum recognition AI model of the multi-view alumen ustum detection segmentation network, training the multi-view alumen ustum detection segmentation network by using the marking data as a supervision signal, and fitting the marking data.
The training module utilizes a multilayer neural network to establish an alum blossom recognition AI model, the multilayer neural network is shown in figure 8, in the multilayer neural network, a convolutional layer + activation function can be regarded as a sensor, the whole network comprises a multilayer sensor, and the characteristics are continuously extracted and compressed through a plurality of convolutional layer + activation functions (sensors), so that higher-level characteristics can be finally obtained, namely, the original characteristics are concentrated one step by one step through the multilayer neural network, the finally obtained characteristics are more reliable, and finally, various tasks can be performed by utilizing the last-level characteristics: such as classification (e.g., determining a class), regression (e.g., regression out of a detection box), etc.
In the invention, the operating principle of the alum blossom identification AI model is mainly as follows:
1. extracting high-level features of the image through a plurality of layers of convolution and activation layers;
2. regression is carried out by utilizing the high-grade characteristics of the image through a regression task to obtain a detection frame (position and size) of the alum blossom;
3. by utilizing the high-level characteristics of the images and through a classification task, the types of the alum flocs (fluffy, flaky and fuzzy alum flocs) are obtained by classification.
The training process is as follows:
collecting a large number of alumen ustum images and marking manually (marking position, size and type of alumen ustum)
Training a alumen ustum model (multilayer neural network) by using back propagation optimization model parameters through a large amount of alumen ustum data.
And the image acquisition and identification unit 11 is used for acquiring an alum blossom image by using an underwater camera and identifying by using the trained alum blossom identification AI model.
In the invention, the trained alumen ustum recognition AI model can be used for detecting attributes such as each detected alumen ustum type, probability, position and the like in the shot image. The list of detected attributes is as follows:
TABLE 1 Alum blossom correlation attribute table for testing and segmenting network
The quantitative evaluation index detection unit 12 is configured to divide a single frame image into M × M regions, calculate alum blossom quantitative indexes for the single regions, and combine 3 types of alum blossom quantitative indexes for the single regions to obtain a quantitative evaluation index of the single frame image, where the quantitative index of each alum blossom category of the single region includes the number, the confidence average, the confidence median, the area average, and the area median of the category in the region.
Since the alumen ustum distribution is not uniform in the shot image area and the alumen ustum distribution information is retained, the single frame image is divided into M × M areas before the alumen ustum characteristics are quantized. Taking 2 × 2 as an example, as shown in fig. 10, the entire image is divided into 4 regions in a 2 × 2 format, and the alumen ustum quantization index is calculated for each of the 4 regions
Calculating the alumen ustum quantization index of the single frame and the single area: the calculation method of the alumen ustum quantization index in a single frame and single region will be described by taking the upper left region of fig. 10 as an example. As shown in FIG. 11, for a single region, the number, the confidence average, the confidence median, the area average and the area median of the region of each alumen ustum category are calculated.
Specifically, the confidence average and the confidence median are calculated as follows:
wherein,
the mean confidence of the type c alumen ustum is shown,
representing the confidence median of the c-type alum blossom, NcThe number of c-type alumen ustum in this region is shown.
Area average and area median calculation mode:
wherein,
the average area of the c-type alumen ustum,
the median of the area of the c-type alumen ustum is shown, and A is the area of the region.
Therefore, the quantization index of each alum blossom class in a single region is as follows:
wherein c represents alum blossom category, and i represents division area.
In the invention, the physical significance represented by each index of alum blossom is as follows:
(1) the number is as follows: detecting the number of alum blossom in a certain type of current area;
(2) confidence mean, median confidence: the confidence, namely the probability of the detected alum flowers is predicted by the alum flower recognition AI model, and the confidence coefficient that the detected alum flowers are in the current category is represented;
(3) area average, area median: the ratio of the sum of the areas of the detected alumen ustum in a certain category of the current region relative to the area of the region represents the density of the alumen ustum in the certain category of the region.
As described above, the single region has 5 quantitative evaluation indexes per type of alumen ustum, and the total of 5 × 3 — 15 alumen ustum indexes in the single region can be obtained by classifying the alumen ustum type.
Single-frame multi-region alumen ustum quantization index combination: combining the 3-class alumen ustum quantization indexes of all the M multiplied by M areas to obtain 15M2And the vanadium flower quantitative index of the vitamin is the quantitative evaluation index of the single-frame image.
In the example of division by 2 × 2 in fig. 10, the number of extracted alum blossom quantization indexes is 15 × 2 × 2 — 60.
And (3) quantifying alum blossom quantifying index time: because the raw water data is updated once in 1 minute, and because of the FPS characteristics of the camera, there are about 360 frames of images in 1 minute, if all the images participate in the calculation of the alum blossom quantization index, it will result in: (1) each frame of image needs to be subjected to calculation of an alum blossom identification AI model and single-frame multi-region alum blossom quantitative evaluation, and the calculation complexity is overlarge; (2) adjacent frames have similar alum blossom quantization indexes and redundant features.
Therefore, in order to actually calculate the 1-minute alum blossom quantification index, time must be quantified by the following method: selecting F frames from the images collected within 1 minute at equal intervals, calculating the alum blossom quantization index of a single frame image, and finally averaging the same alum blossom quantization indexes of the F frames to obtain the alum blossom quantization index after time quantization.
Examples
Experimental data: the system is communicated with water plant service experts, and the water purification system is subjected to service transformation in the aspect of a water plant, so that alum flocs, mostly in granular form, are caused at the water inlet of the sedimentation tank due to excessive addition of the coagulant. Therefore, in the experiments for confirming the initial correlation parameters, the data from No. 4/month 24 to No. 4/month 25 are used as compared with the conventional data. In this embodiment, data of 38 hours is used as a training set, data of 10 hours is used as a test set, and in the experimental process, the training set: 2280, test set: 600 strips.
Single frame image area division M × M experiment: in the experimental process, a single frame image is divided into 1 × 1, 2 × 2, 3 × 3, 4 × 4, 5 × 5, 6 × 6, 7 × 7 and 10 × 10 regions respectively, a support vector regression model (only using alum blossom quantization index for modeling) is established, and in the same test set, the scores of the models are as follows:
TABLE 2 Single frame image region partitioning MxM experiment
Alum blossom quantization index time quantization parameter F frame experiment: in the experimental process, 10, 25, 30, 40, 50, 60 and 70 frames of images are respectively sampled at equal intervals every minute to carry out alum blossom quantitative index calculation, a support vector regression model (only using alum blossom quantitative index for modeling) is established, and in the same test set, the scores of the models are as follows in the following table 3:
TABLE 3 Alum quantization index time quantization parameter F frame experiment
| |
10 | 25 | 30 | 40 | 50 | 60 | 70 |
| Test set | 600 | 600 | 600 | 600 | 600 | 600 | 600 |
| Model scoring | 0.4851 | 0.5928 | 0.6056 | 0.6031 | 0.6059 | 0.6138 | 0.5975 |
And (3) analysis: (1) in the single-frame image region division experiment, although the scores of the division regions of 2 × 2 and 5 × 5 are both about 0.4, in the project, the single-frame image is divided into 2 × 2 to be subjected to alum blossom quantization index calculation relative to the balance between the calculation amount and the precision; (2) for the time quantization parameter experiment of the alum blossom quantization index, the real-time performance of the algorithm is considered, 25 frames of images are selected to be collected every minute, and the alum blossom quantization index after time quantization is calculated.
By the analysis, the multi-modal data source hysteresis regression precipitated water turbidity model is established, and besides the precipitated water turbidity after about 2 hours is predicted, whether the alum blossom quantitative index is reasonable or not can be evaluated according to the trained model. The model established according to the invention can calculate the normalized importance degree of each index of the alum blossom quantization for the turbidity of the settled water and provide the quantization index, as shown in figure 13.
As can be seen from FIG. 13, the quantitative indicators of the fluffy in each divided region have high importance (for example, the average confidence of the region 2_ fluffy _ confidence is 0.17, the number of the region 2_ fluffy _ number is 0.13, etc.), and the sheet-shaped fluffy and fuzzy fluffy have low importance for the prediction of the turbidity of the settled water, which is consistent with the qualitative conclusion of the fluffy business analysis. Further, the importance of different types of alum blossom in all regions was analyzed, as shown in FIG. 14. With reference to fig. 13 and 14, from the quantitative analysis, the quantitative index of fluffy alum blossom is important for the prediction of the turbidity of the settled water, and is consistent with the qualitative conclusion of the business analysis, so that the rationality of the alum blossom fine particle classification and quantitative index is verified.
In summary, the quantitative indicators (the number of alum flowers, the confidence average, the confidence median, the area average and the area median) of all the areas per minute for each category are averaged and normalized to be displayed on the bar graph of the Web interface, and it is known that the coagulation effect of alum flowers is better when the area is larger as the number of fluffy alum flowers is larger.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Modifications and variations can be made to the above-described embodiments by those skilled in the art without departing from the spirit and scope of the present invention. Therefore, the scope of the invention should be determined from the following claims.
Claims (10)
1. A method for analyzing alum blossom characteristics by using an image recognition technology is characterized by comprising the following steps:
step S1, collecting an alum blossom image by using an underwater camera, marking the collected image, generating training data, establishing an alum blossom recognition AI model, and performing autonomous training on the alum blossom recognition AI model by using the marking data as a supervision signal to obtain a trained alum blossom recognition AI model;
step S2, collecting an alum blossom image by using an underwater camera, and identifying by using the trained alum blossom identification AI model;
and step S3, dividing the single-frame image into M multiplied by M areas, respectively calculating alum blossom quantization indexes for the single areas, and combining various alum blossom quantization indexes for the single areas to obtain the quantization evaluation index of the single-frame image.
2. The method of claim 1, wherein the step S1 further comprises:
s100, collecting alum blossom images within the field depth range by using an underwater camera;
step S101, determining alum blossom labeling standards of various categories, marking collected images, and generating training data;
step S102, an alumen ustum recognition AI model of the multi-view alumen ustum detection segmentation network is established, marking data are used as supervision signals, the multi-view alumen ustum detection segmentation network is trained, and the marking data are fitted.
3. The method of claim 2, wherein the method comprises the steps of: in step S101, marking criteria of three categories of alumen ustum are determined according to related services of the alumen ustum and the shot image, the collected image is marked according to the determined marking criteria, and remark data is added to the image in which fluffy alumen ustum, flaky alumen ustum, alumen ustum and fuzzy fluffy image cannot be determined.
4. The method of claim 3, wherein the method comprises the steps of: in step S102, the AI model extracts high-level features of the image through multilayer convolution and activation layers by using a multilayer neural network, obtains a detection frame of the alum blossom through regression task regression by using the extracted high-level features of the image, and obtains the type of the alum blossom through classification task classification by using the high-level features of the image.
5. The method of claim 4, wherein the method comprises the steps of: in step S102, training the alumen ustum recognition AI model by using back propagation optimization model parameters through a large amount of alumen ustum data; in step S2, the trained alumen ustum recognition AI model is used to recognize the collected image, and each alumen ustum type, probability and position attribute are detected.
6. The method of claim 5, wherein the method comprises the steps of: the quantitative indexes of each alum blossom category of the single region comprise the number, the confidence average, the confidence median, the area average and the area median of the category in the region.
7. The method of claim 6, wherein the method comprises the steps of: in step S3, all the M × M three-class alumen ustum quantization indexes of all the regions are combined to obtain 15M2And the vanadium flower quantitative index of the vitamin is the quantitative evaluation index of the single-frame image.
8. The method of claim 7, wherein the method comprises the steps of: in step S3, F frames are selected at equal intervals from the image collected within 1 minute, the metric of the F frame image is calculated, and then the metric of the F frame image is averaged to obtain the metric of the time-quantized alum.
9. An apparatus for analyzing alum blossom characteristics by using image recognition technology, comprising:
the device comprises an alum blossom identification AI model building and training unit, a parameter setting unit and a parameter setting unit, wherein the alum blossom identification AI model building and training unit is used for collecting an alum blossom image by using an underwater camera, marking the collected image, generating training data, building an alum blossom identification AI model, and performing autonomous training on the alum blossom identification AI model by using the marking data as a supervision signal to obtain a trained alum blossom identification AI model;
the image acquisition and identification unit is used for acquiring an alum blossom image by using an underwater camera and identifying by using the trained alum blossom identification AI model;
and the quantitative evaluation index detection unit is used for dividing the single-frame image into M multiplied by M areas, respectively calculating the alum blossom quantitative indexes of the single area, and combining the three types of alum blossom quantitative indexes of the single area to obtain the quantitative evaluation index of the single-frame image.
10. The apparatus of claim 9, wherein the quantitative evaluation index detection unit selects F frames from the images collected within 1 minute at equal intervals, calculates the quantitative index of alum floc in a single frame of image, and then averages the same alum floc quantitative indexes of the F frames to obtain the time-quantized alum floc quantitative index.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111001467.4A CN113688751A (en) | 2021-08-30 | 2021-08-30 | Method and device for analyzing alum blossom characteristics by using image recognition technology |
| PCT/CN2021/119244 WO2023029117A1 (en) | 2021-08-30 | 2021-09-18 | Method and apparatus for analyzing alum floc feature by using image recognition technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111001467.4A CN113688751A (en) | 2021-08-30 | 2021-08-30 | Method and device for analyzing alum blossom characteristics by using image recognition technology |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113688751A true CN113688751A (en) | 2021-11-23 |
Family
ID=78583803
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111001467.4A Pending CN113688751A (en) | 2021-08-30 | 2021-08-30 | Method and device for analyzing alum blossom characteristics by using image recognition technology |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN113688751A (en) |
| WO (1) | WO2023029117A1 (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115375009A (en) * | 2022-07-28 | 2022-11-22 | 上海城市水资源开发利用国家工程中心有限公司 | Method for establishing coagulation intelligent monitoring linkage system |
| CN116229518A (en) * | 2023-03-17 | 2023-06-06 | 百鸟数据科技(北京)有限责任公司 | Bird species observation method and system based on machine learning |
| CN117315454A (en) * | 2023-11-29 | 2023-12-29 | 河北中瀚水务有限公司 | Evaluation method, device and system for flocculation reaction process |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117929375B (en) * | 2024-03-21 | 2024-06-04 | 武汉奥恒胜科技有限公司 | Water quality detection method and water quality detector based on image processing |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120033865A1 (en) * | 2009-04-15 | 2012-02-09 | Koninklijke Philips Electronics N.V. | Quantification of medical image data |
| JP2018032372A (en) * | 2017-03-24 | 2018-03-01 | 株式会社サイバーブレーン | Program, device, system and method for generating image information of image subject |
| US20190251402A1 (en) * | 2018-02-13 | 2019-08-15 | Slingshot Aerospace, Inc. | Adaptive neural network selection to extract particular results |
| CN111242918A (en) * | 2020-01-10 | 2020-06-05 | 深圳信息职业技术学院 | Image segmentation method and system based on Kalman filtering and Markov random field |
| CN111833369A (en) * | 2020-07-21 | 2020-10-27 | 中冶赛迪重庆信息技术有限公司 | Alum image processing method, system, medium and electronic device |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112101352A (en) * | 2020-09-10 | 2020-12-18 | 广州深视未来智能科技有限责任公司 | Underwater alumen ustum state identification method and monitoring device, computer equipment and storage medium |
| CN112441654A (en) * | 2020-11-02 | 2021-03-05 | 广州晋合水处理设备有限公司 | Control system and method suitable for coagulating sedimentation |
-
2021
- 2021-08-30 CN CN202111001467.4A patent/CN113688751A/en active Pending
- 2021-09-18 WO PCT/CN2021/119244 patent/WO2023029117A1/en active Application Filing
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120033865A1 (en) * | 2009-04-15 | 2012-02-09 | Koninklijke Philips Electronics N.V. | Quantification of medical image data |
| JP2018032372A (en) * | 2017-03-24 | 2018-03-01 | 株式会社サイバーブレーン | Program, device, system and method for generating image information of image subject |
| US20190251402A1 (en) * | 2018-02-13 | 2019-08-15 | Slingshot Aerospace, Inc. | Adaptive neural network selection to extract particular results |
| CN111242918A (en) * | 2020-01-10 | 2020-06-05 | 深圳信息职业技术学院 | Image segmentation method and system based on Kalman filtering and Markov random field |
| CN111833369A (en) * | 2020-07-21 | 2020-10-27 | 中冶赛迪重庆信息技术有限公司 | Alum image processing method, system, medium and electronic device |
Non-Patent Citations (2)
| Title |
|---|
| AHMAD, A. ET AL.: "Study on gray water treatment using cement kiln dust and chemical coagulants", 《JOURNAL OF APPLIED WATER ENGINEERING AND RESEARCH》, vol. 8, no. 2, 14 May 2020 (2020-05-14), pages 161 * |
| 王冬敏等: "网格絮凝池的发展及研究展望", 《广东化工》, vol. 43, no. 17, 15 September 2016 (2016-09-15), pages 79 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115375009A (en) * | 2022-07-28 | 2022-11-22 | 上海城市水资源开发利用国家工程中心有限公司 | Method for establishing coagulation intelligent monitoring linkage system |
| CN115375009B (en) * | 2022-07-28 | 2023-09-22 | 上海城市水资源开发利用国家工程中心有限公司 | Method for establishing intelligent monitoring linkage system for coagulation |
| CN116229518A (en) * | 2023-03-17 | 2023-06-06 | 百鸟数据科技(北京)有限责任公司 | Bird species observation method and system based on machine learning |
| CN116229518B (en) * | 2023-03-17 | 2024-01-16 | 百鸟数据科技(北京)有限责任公司 | Bird species observation method and system based on machine learning |
| CN117315454A (en) * | 2023-11-29 | 2023-12-29 | 河北中瀚水务有限公司 | Evaluation method, device and system for flocculation reaction process |
| CN117315454B (en) * | 2023-11-29 | 2024-03-12 | 河北中瀚水务有限公司 | Evaluation method, device and system for flocculation reaction process |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2023029117A1 (en) | 2023-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113688751A (en) | Method and device for analyzing alum blossom characteristics by using image recognition technology | |
| Zarie et al. | Flotation froth image classification using convolutional neural networks | |
| CN105975941B (en) | A kind of multi-direction vehicle detection identifying system based on deep learning | |
| Wang et al. | Process working condition recognition based on the fusion of morphological and pixel set features of froth for froth flotation | |
| CN116205919B (en) | Hardware part production quality detection method and system based on artificial intelligence | |
| Myrans et al. | Automated detection of fault types in CCTV sewer surveys | |
| CN112598713A (en) | Offshore submarine fish detection and tracking statistical method based on deep learning | |
| CN104992223A (en) | Intensive population estimation method based on deep learning | |
| CN111966053A (en) | Intelligent flocculant decision making system | |
| CN108363953B (en) | Pedestrian detection method and binocular monitoring equipment | |
| CN113065578A (en) | Image visual semantic segmentation method based on double-path region attention coding and decoding | |
| CN113642474A (en) | Hazardous area personnel monitoring method based on YOLOV5 | |
| Liu et al. | A classification method of glass defect based on multiresolution and information fusion | |
| CN108647722B (en) | Zinc ore grade soft measurement method based on process size characteristics | |
| CN113012195A (en) | Method for estimating river surface flow velocity based on optical flow calculation and readable storage medium | |
| CN114170549A (en) | Water surface floater detection method based on deep learning | |
| CN114255357A (en) | Group-breeding pig identity identification and health monitoring method based on computer vision | |
| CN111833369A (en) | Alum image processing method, system, medium and electronic device | |
| CN113450573A (en) | Traffic monitoring method and traffic monitoring system based on unmanned aerial vehicle image recognition | |
| CN106650593A (en) | Passenger flow statistical method and device | |
| CN108931621B (en) | Zinc ore grade soft measurement method based on process texture characteristics | |
| CN111145145A (en) | Image surface defect detection method based on MobileNet | |
| Chen et al. | Implementation of a Fruit Quality Classification Application Using an Artificial Intelligence Algorithm. | |
| Sun et al. | Exploiting deeply supervised inception networks for automatically detecting traffic congestion on freeway in China using ultra-low frame rate videos | |
| CN115526852A (en) | Molten pool and splash monitoring method in selective laser melting process based on target detection and application |
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 |