CN-121982341-A - Method, device, equipment and medium for monitoring sludge state of oxidation ditch

CN121982341ACN 121982341 ACN121982341 ACN 121982341ACN-121982341-A

Abstract

The application discloses a method, a device, equipment and a medium for monitoring the sludge state of an oxidation ditch, which relate to the technical field of computer vision and comprise the steps of collecting a sludge surface image, carrying out histogram equalization treatment on a sludge region in the image by utilizing a self-adaptive edge detection algorithm to obtain an image to be processed, utilizing a feature encoder to determine a visual feature vector, acquiring a process parameter to generate a parameter coding vector, carrying out weighted fusion on the parameter coding vector and the visual feature vector to obtain a target feature vector, constructing a historical mode set to determine a deviation degree index between the target feature vector and the feature vector of a historical stable mode, determining a process parameter abnormal score between the process parameter and a process parameter baseline, determining a modal deviation degree based on the deviation degree index and the abnormal score, constructing a time sequence trend graph, determining a deviation rate change speed at adjacent moments, determining a state monitoring result based on the modal deviation degree, and improving the efficiency of monitoring the sludge state of the oxidation ditch.

Inventors

LIU QING
LU GUILIN
WANG MENG
YAN JIANBANG

Assignees

昕彤赋能(长沙)人工智能行业应用系统有限公司

Dates

Publication Date: 20260505
Application Date: 20260327

Claims (10)

1. The method for monitoring the sludge state of the oxidation ditch is characterized by comprising the following steps of: Acquiring corresponding sludge surface images by using imaging equipment deployed in an oxidation ditch, and then performing histogram equalization treatment on sludge areas in each sludge surface image by using a self-adaptive edge detection algorithm to obtain an image to be treated; Determining a visual feature vector corresponding to the image to be processed by using a feature encoder, determining a technological parameter corresponding to the image to be processed, generating a corresponding parameter code vector by using a technological parameter code function and based on the technological parameter, and carrying out weighted fusion on the parameter code vector and the visual feature vector to obtain a target feature vector; Constructing a history mode set for describing the system steady state, determining a deviation index between the target feature vector and a feature vector of a history steady mode based on the history mode set, determining a process parameter anomaly score between a current process parameter and a preset process parameter baseline, and determining a modal deviation by utilizing a multi-modal fusion mechanism based on the deviation index and the process parameter anomaly score; and constructing a time sequence trend graph based on each modal deviation degree, determining the deviation rate change speed between adjacent moments in the time sequence trend graph, and then determining a state monitoring result based on the deviation rate change speed and the modal deviation degree.
2. The method for monitoring sludge status in an oxidation ditch according to claim 1, wherein the steps of acquiring corresponding sludge surface images by using an imaging device disposed in the oxidation ditch, and performing histogram equalization processing on sludge areas in each sludge surface image by using an adaptive edge detection algorithm to obtain an image to be processed comprise: Respectively arranging digital microscopic cameras at an inlet of an anaerobic zone, a middle part of an aerobic zone and a backflow channel of an oxidation ditch, then utilizing the digital microscopic cameras to acquire each sludge surface image of a corresponding process section based on a preset acquisition frequency, and determining metadata which respectively correspond to each sludge surface image and comprise acquisition time, sampling position, optical amplification factor, sludge concentration, dissolved oxygen, temperature and pH value; generating corresponding numbers for the sludge surface images, carrying out consistency check on the metadata and the numbers, and storing the metadata and the numbers into a database after the verification result characterization check is passed; Performing size normalization processing on each sludge surface image to obtain a corresponding normalized image, then performing sludge region identification on the normalized image by using a self-adaptive edge detection algorithm, and cutting off black edge parts which do not contain sludge information in the image according to an identification result to obtain an interested region only comprising sludge; and optimizing brightness and contrast of the sludge floc structure and the visual presentation of the filamentous fungus morphology in the region of interest by using a self-adaptive histogram equalization algorithm to obtain an image to be processed.
3. The method for monitoring sludge status in oxidation ditch according to claim 1, wherein the determining, by using a feature encoder, a visual feature vector corresponding to the image to be processed, determining a process parameter corresponding to the image to be processed, then generating a corresponding parameter code vector by using a process parameter code function based on the process parameter, and performing weighted fusion on the parameter code vector and the visual feature vector to obtain a target feature vector, comprises: Inputting the image to be processed into a feature encoder, and extracting deep visual features of the image to be processed by using the feature encoder to obtain visual feature vectors with preset dimensions, wherein the feature encoder is an encoder adopting self-supervision learning or contrast learning training; Acquiring technological parameters corresponding to the image to be processed, and converting the technological parameters into parameter coding vectors with the same dimensionality as the visual feature vectors by utilizing a technological parameter coding function; and carrying out weighted summation on the parameter coding vector and the visual feature vector based on a preset fusion weight coefficient to generate a target feature vector for representing the visual property and the technological condition of the sludge.
4. The method for monitoring the sludge state of an oxidation ditch according to claim 1, wherein the constructing a history pattern set for describing the steady state of the system comprises: storing target feature vectors corresponding to each historical collection time as historical feature vectors, and constructing a historical mode set for describing the visual mode and process condition association characteristics of the oxidation ditch sludge system in a stable running state based on each historical feature vector; And obtaining a target feature vector corresponding to the current acquisition time, and iteratively updating the history mode set based on a preset updating coefficient to obtain a new history mode set, wherein the preset updating coefficient is a coefficient set based on the succession rate of the microbial community in the oxidation ditch.
5. The method of claim 1, wherein determining a deviation index between the target feature vector and a feature vector of a historical steady mode based on the set of historical modes, determining a process parameter anomaly score between a current process parameter and a preset process parameter baseline, and then determining a modal deviation using a multi-modal fusion mechanism based on the deviation index and the process parameter anomaly score, comprises: Distributing corresponding weights based on the difference degree between each historical feature vector and the current target feature vector in the historical pattern set and based on the time interval between the acquisition time of each historical feature vector and the current acquisition time, wherein the numerical value corresponding to the weights and the numerical value corresponding to the time interval are in a negative correlation relationship; carrying out weighted calculation on the difference between each historical characteristic vector and the current target characteristic vector and the weight to obtain a deviation degree index for describing the deviation degree of the current sludge visual property relative to the historical stable mode; Determining the deviation amplitude of the current process parameter relative to a preset process parameter baseline, determining a process parameter anomaly score based on the deviation amplitude, and then carrying out weighted summation on the deviation index and the process parameter anomaly score based on a preset multi-mode fusion coefficient to obtain the mode deviation degree used for representing the visual property and the process condition anomaly degree of the sludge.
6. The method according to claim 1, wherein the constructing a time-series trend graph based on each of the modal deviation degrees, and determining a deviation rate change speed between adjacent time points in the time-series trend graph, and then determining a state monitoring result based on the deviation rate change speed and the modal deviation degree, comprises: Arranging the modal deviation degrees according to the sequence from front to back of the generation time to obtain time sequence data used for representing the time evolution trend of the modal deviation degrees; determining a difference value between the modal deviation degrees of two adjacent acquisition moments based on the time series data, and determining a deviation rate change speed used for representing the change speed of the deviation degrees based on the difference value; Comparing the modal deviation degree at the current acquisition time with a first preset threshold value to obtain a first comparison result, and comparing the deviation rate change speed corresponding to the current acquisition time with a second preset threshold value to obtain a second comparison result; and if the first comparison result indicates that the modal deviation degree is larger than the first preset threshold value and the deviation rate change speed is larger than the second preset threshold value, judging that the current running state of the oxidation ditch is an abnormal state, obtaining a state monitoring result and triggering abnormal early warning.
7. The method according to any one of claims 1 to 6, wherein after constructing a time-series trend graph based on each of the modal deviation degrees and determining a deviation rate change speed between adjacent time points in the time-series trend graph, and then determining a status monitoring result based on the deviation rate change speed and the modal deviation degree, further comprising: If the current running state of the oxidation ditch is characterized as an abnormal state, activating an acousto-optic warning device arranged in the oxidation ditch so as to mark the abnormal occurrence position in a graphical mode on a display interface of a central control system by utilizing the acousto-optic warning device; Or the working condition operation record in the historical database is called, the process parameters at the current collection time and the preset reasoning rules are combined to generate an operation suggestion comprising an aeration quantity adjusting instruction and a sludge reflux ratio adjusting instruction, and the operation suggestion is issued to an operation interface; Or judging whether the duration of the abnormal state is longer than a preset duration, if so, suspending the execution of the automatic regulation and control program, and pushing the time sequence trend data comprising the modal deviation degree, the modal fusion score and the structural report of the potential abnormality cause analysis generated based on the knowledge base to a designated operator terminal.
8. An oxidation ditch sludge state monitoring device, characterized by comprising: The image acquisition module is used for acquiring corresponding sludge surface images by using imaging equipment deployed in the oxidation ditch, and then performing histogram equalization processing on sludge areas in the sludge surface images by using a self-adaptive edge detection algorithm to obtain an image to be processed; The target feature vector generation module is used for determining a visual feature vector corresponding to the image to be processed by utilizing a feature encoder, determining technological parameters corresponding to the image to be processed, generating a corresponding parameter code vector by utilizing a technological parameter code function and based on the technological parameters, and carrying out weighted fusion on the parameter code vector and the visual feature vector to obtain a target feature vector; The system comprises a modal deviation determining module, a multi-modal fusion mechanism and a model deviation determining module, wherein the modal deviation determining module is used for constructing a historical mode set for describing a system steady state, then determining a deviation index between the target feature vector and a feature vector of a historical steady mode based on the historical mode set, determining a process parameter anomaly score between a current process parameter and a preset process parameter baseline, and then determining modal deviation based on the deviation index and the process parameter anomaly score by utilizing the multi-modal fusion mechanism; and the state monitoring result determining module is used for constructing a time sequence trend graph based on the modal deviation degree, determining the deviation rate change speed between adjacent moments in the time sequence trend graph and then determining a state monitoring result based on the deviation rate change speed and the modal deviation degree.
9. An electronic device, comprising: A memory for storing a computer program; A processor for executing the computer program to implement the oxidation ditch sludge state monitoring method as claimed in any one of claims 1 to 7.
10. A computer readable medium for storing a computer program, wherein the computer program when executed by a processor implements the oxidation ditch sludge condition monitoring method according to any one of claims 1 to 7.

Description

Method, device, equipment and medium for monitoring sludge state of oxidation ditch Technical Field The invention relates to the technical field of computer vision, in particular to a method, a device, equipment and a medium for monitoring the sludge state of an oxidation ditch. Background At present, an oxidation ditch sewage treatment system is widely applied to urban and industrial sewage treatment, and water quality purification is realized by degrading organic matters in water through microorganisms, but the existing monitoring method mainly depends on manual experience or predefined state categories, and potential changes of sludge and microbial communities cannot be captured in real time, particularly a tiny but key abnormal mode, so that abnormality is often found in a delayed manner, and the running risk of the system and the manual intervention cost are increased. Traditional oxidation ditch monitoring methods rely on manual inspection or fixed threshold sensors to react with microscopic changes in microbial communities and sludge structure evolution. When the system is subjected to early abnormal change, such as abnormal microorganism distribution or fluctuation of sludge sedimentation performance, the existing method cannot be timely found, and water quality is easy to drop and even the system is easy to disturb. Sludge conditions are highly dynamic and complex, and microbial community composition, floc structure and sedimentation characteristics can change over time and process parameters, which are difficult to capture by traditional fixed-class-based monitoring methods. For example, abnormal proliferation of filamentous bacteria or the process of floc disintegration often first manifest as minor visual changes, which if not perceived in time, may be discovered after the problem is exacerbated, increasing operational risk and leading to frequent manual intervention. From the above, how to improve the efficiency of monitoring the status of the oxidation ditch sludge in the oxidation ditch sludge status monitoring process is a problem to be solved urgently. Disclosure of Invention In view of the above, the present invention aims to provide a method, a device, equipment and a medium for monitoring the status of oxidation ditch sludge, which can improve the efficiency of monitoring the status of oxidation ditch sludge in the process of monitoring the status of oxidation ditch sludge. The specific scheme is as follows: in a first aspect, the application provides a method for monitoring the sludge state of an oxidation ditch, which comprises the following steps: Acquiring corresponding sludge surface images by using imaging equipment deployed in an oxidation ditch, and then performing histogram equalization treatment on sludge areas in each sludge surface image by using a self-adaptive edge detection algorithm to obtain an image to be treated; Determining a visual feature vector corresponding to the image to be processed by using a feature encoder, determining a technological parameter corresponding to the image to be processed, generating a corresponding parameter code vector by using a technological parameter code function and based on the technological parameter, and carrying out weighted fusion on the parameter code vector and the visual feature vector to obtain a target feature vector; Constructing a history mode set for describing the system steady state, determining a deviation index between the target feature vector and a feature vector of a history steady mode based on the history mode set, determining a process parameter anomaly score between a current process parameter and a preset process parameter baseline, and determining a modal deviation by utilizing a multi-modal fusion mechanism based on the deviation index and the process parameter anomaly score; and constructing a time sequence trend graph based on each modal deviation degree, determining the deviation rate change speed between adjacent moments in the time sequence trend graph, and then determining a state monitoring result based on the deviation rate change speed and the modal deviation degree. Optionally, the acquiring the corresponding sludge surface image by using an imaging device deployed in the oxidation ditch, and then performing histogram equalization processing on the sludge area in each sludge surface image by using a self-adaptive edge detection algorithm to obtain an image to be processed, including: Respectively arranging digital microscopic cameras at an inlet of an anaerobic zone, a middle part of an aerobic zone and a backflow channel of an oxidation ditch, then utilizing the digital microscopic cameras to acquire each sludge surface image of a corresponding process section based on a preset acquisition frequency, and determining metadata which respectively correspond to each sludge surface image and comprise acquisition time, sampling position, optical amplification factor, sludge concentration, dissolved oxygen