A defoamer foam distribution analysis method based on image feature recognition

Abstract

The present invention discloses a defoamer foam distribution analysis method based on image feature recognition, specifically relating to the field of industrial foam behavior perception and analysis for identifying image objects characterized by random patterns, including obtaining a foam image sequence in a target area, the foam image sequence being acquired by an imaging device, and the image frames of the foam image sequence having temporal continuity; performing a disturbance feature extraction operation on the foam image sequence to obtain local disturbance velocity information of the foam edge within a preset time, membrane surface tension change trend information, and morphological boundary fluctuation information. By extracting foam structure disturbance features from the image time series and constructing a structural evolution map memory library, the trend matching mechanism is combined to identify irregular mutation image behaviors, thereby achieving dynamic perception and abnormal response to sudden foam states without prior support, thereby solving the problem of being unable to identify random pattern foam states.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial foam behavior perception and analysis for identifying image objects characterized by random patterns, and more specifically, to a defoamer foam distribution analysis method based on image feature recognition. Background Art

[0002] In high-pressure reactors, fermentation bioreactors, or thermosensitive interfacial fluid systems, foam distribution is not only influenced by conventional physical parameters such as surface tension and gas-liquid interface velocity, but can also exhibit abnormal behavior with nonlinear transition characteristics under certain extreme conditions. This behavior is often manifested as localized bubble expansion in a very short period of time, sudden fusion of adjacent bubbles, or violent collapse of the entire membrane wall. The formation process is unpredictable, extremely short-lived, and non-reproducible.

[0003] This type of foam state variation is essentially a highly uncertain and non-replicable image behavior pattern that cannot be directly learned from previous data or described through rule-driven methods. Especially under high-frame-rate imaging conditions, this behavior appears in visual images as a sudden, non-logically continuous boundary reconstruction structure, lacking the semantic edges, morphological consistency, or feature stability that traditional pattern recognition models rely on.

[0004] Current image recognition methods primarily rely on supervised learning systems, which rely on a large number of annotated images to construct a stable feature space and perform object classification or structural segmentation based on statistical distribution and spatial semantic rules. However, in the recognition of bubble bursts, almost all visible areas have no clear boundary with the normal state image in terms of significance, making it impossible to define the distinction between normal and abnormal samples. Therefore, in existing models, abnormal bubble states are treated as unlabeled noise with no recognition value, ultimately leading to serious missed detections or misjudgments.

[0005] Image recognition systems are often powerless when faced with random, anomalous images. The fundamental difficulty lies in their inability to identify image objects with unique and unpredictable patterns. Unique refers to the inability of such image objects to be assigned to any existing semantic category. Random refers to the irregularity, lack of samples, and lack of prior knowledge regarding their appearance, evolution, regional boundaries, and timing. In this mutant state, the image structure of foam does not adhere to background textures, boundary features, or shape templates, nor does it exhibit any logic of continuous evolution of topological or morphological structures. Instead, it exhibits a combination of unstructured patterns, including fragmented reconstruction, local distortion, loose boundaries, and sudden brightness changes. This directly challenges the mainstream visual system framework based on learning from stable feature spaces.

[0006] To summarize, the essence of the problem in this solution lies in the fact that the current image recognition system cannot effectively identify image object categories that have no prior support, irregular evolution, and cannot be reconstructed, that is, burst foam structures with unique random pattern characteristics, which causes the system to lose its ability to respond during the critical abnormal behavior outbreak stage; the recognition requirements of such objects not only go beyond the scope of traditional image classification, segmentation or detection, but also urgently require image recognition mechanisms with self-evolution, self-generation and self-perception capabilities. This is the core point for defoamer foam image recognition to move from static learning to dynamic uncertainty modeling. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a defoaming agent foam distribution analysis method based on image feature recognition, which extracts foam structure disturbance features from the image time series and constructs a structural evolution map memory library, and combines the trend matching mechanism to identify irregular mutation image behavior, thereby realizing dynamic perception and abnormal response to sudden foam states without prior support, so as to solve the problem that random pattern foam states cannot be identified in the above-mentioned background technology.

[0008] To achieve the above object, the present invention provides the following technical solution: a defoamer foam distribution analysis method based on image feature recognition, comprising:

[0009] S1. Acquire a foam image sequence in a target area, wherein the foam image sequence is acquired by an imaging device, and image frames of the foam image sequence have temporal continuity;

[0010] S2. Performing a disturbance feature extraction operation on the foam image sequence to obtain local disturbance velocity information, membrane surface tension change trend information, and morphological boundary fluctuation information of the foam edge within a preset time, and using these as foam structure data to characterize the evolution dynamics of the foam structure;

[0011] S3. Constructing a structural evolution atlas memory library, encoding the foam structure data extracted from multiple image frames in time series and annotating their spatial positions, and storing them by evolution stage to form a reference atlas set for characterizing the foam evolution trend;

[0012] S4. Performing structural instability detection in the current image frame, identifying potential abnormal evolution areas by analyzing the morphological continuity interruption, texture connection breakage, and boundary focus change of the foam structure data in the local area;

[0013] S5. Based on the structural evolution atlas memory library, performing trend matching comparison on the foam structure data extracted from the current image frame to determine whether there is a characteristic pattern consistent with the recorded abnormal mutation evolution trend;

[0014] S6. When the comparison result shows that the current foam structure data is consistent with the abnormal mutation evolution trend recorded in the structure evolution map memory library, a mutation recognition signal is generated, and the image recognition system is triggered to execute an abnormal state response accordingly.

[0015] In a preferred embodiment, in S2, adjacent image frames in the foam image sequence are temporally registered, and the temporal displacement of corresponding foam edge pixels between the image frames is calculated to obtain the local position change rate of the foam edge in continuous time, thereby obtaining local disturbance velocity information;

[0016] When obtaining information on the variation trend of membrane surface tension, the grayscale gradient amplitude of the foam edge region is extracted from each image frame of the foam image sequence through a gradient operation operator. Combined with the variation in the edge contour expansion and contraction rate of the foam edge region in multiple consecutive image frames, the gradient variation rate and contour deformation rate of the unit length boundary on the time axis are calculated, thereby constructing a membrane surface tension variation curve to obtain information used to characterize the variation trend of membrane surface tension.

[0017] In multiple continuous image frames in the foam image sequence, the contour morphology of the foam edge is extracted by an edge detection operator, and the contour matching offset and curvature change rate of the foam edge between adjacent image frames are calculated. The local bending amplitude and contour deformation intensity of the foam edge in the time dimension are statistically analyzed, and a fluctuation change curve reflecting the dynamic change characteristics of the edge is constructed. By performing feature point extraction and frequency domain analysis operations on the fluctuation change curve, the boundary displacement amplitude and curvature disturbance frequency are extracted from it and stored and encoded as morphological boundary fluctuation information.

[0018] In a preferred embodiment, in S3, the foam structure data extracted from each image frame in the foam image sequence are numbered in chronological order, and the foam structure data in each frame is labeled with its spatial position index in the image coordinate system, thereby completing the time series encoding and spatial position labeling of the foam structure data;

[0019] Based on the local disturbance velocity information, membrane surface tension change trend information, and morphological boundary fluctuation information contained in the foam structure data, the numerical change range and change rate of the three types of information in the time dimension are extracted respectively, and the evolution state index is constructed according to the corresponding numerical change range and change rate;

[0020] According to the numerical variation range and variation rate corresponding to the local disturbance velocity information, the membrane surface tension variation trend information and the morphological boundary fluctuation information in the evolution state index, the foam structure data is divided into three evolution stages: when the numerical variation range and variation rate of the three types of information are all lower than the corresponding preset evolution judgment threshold, it is defined as a steady-state stage; when the numerical variation range or variation rate of any one type of information exceeds the corresponding preset evolution judgment threshold, it is defined as a transition state stage; when the numerical variation range and variation rate of any two types of information among the three types of information exceed the corresponding preset evolution judgment threshold, it is defined as a mutation state stage; otherwise, the foam structure data is marked as a state to be updated, and the evolution stage classification of the foam structure data is completed accordingly;

[0021] The foam structure data that have completed time series encoding and spatial position annotation are stored in the structural evolution map memory according to the evolution stage they belong to, so as to construct a reference map set for characterizing the foam evolution trend.

[0022] In a preferred embodiment, in S4, the step of performing structural instability detection on the current image frame includes:

[0023] S401, in the current image frame, dividing the image area corresponding to the foam structure data into image grid units, and using each grid unit in the image grid units as an analysis area;

[0024] S402. Performing a continuity analysis operation on each analysis region, the continuity analysis operation comprising: detecting the structural integrity of the foam structure in the spatial distribution based on the contour morphological image extracted from the foam structure boundary in the region, identifying whether there are morphological destruction features such as contour fracture, edge segment interruption, or closed structure loss, so as to determine whether there is a morphological continuity interruption in the analysis region; and analyzing the pixel grayscale directional consistency based on the texture gradient distribution map in the analysis region, identifying whether there are abnormal texture features such as texture structure rupture, pattern continuity discontinuity, or gradient direction abrupt change, so as to determine whether there is a texture connection break in the analysis region;

[0025] S403. In each analysis area, based on the boundary position corresponding to the foam structure data, the contour line of the foam structure is extracted by an edge detection algorithm, and the focus response value of each boundary pixel on the contour line is calculated using the Laplace operator; the focus response difference between adjacent pixels along the pixel sequence of the contour line is calculated, and the average value, range and standard deviation of the focus response difference sequence are counted to form a parameter group for describing the boundary focus change; the parameter group of the boundary focus change is input into the structural stability calculation function together with the results of the morphological continuity interruption and the texture connection break. If the output score value of the structural stability calculation function exceeds the preset threshold, the analysis area is marked as an abnormal evolution area.

[0026] In a preferred embodiment, the structural stability calculation function inputs the three statistical indicators in the parameter group of boundary focus change as continuous variables, and the three statistical indicators are the mean, range and standard deviation of the focus response difference sequence; at the same time, the results of the morphological continuity interruption and the results of the texture connection break are converted into Boolean variable forms, wherein when the result is "yes", the value corresponding to the Boolean variable form is 1, and when the result is "no", the value corresponding to the Boolean variable form is 0, and the values ​​of the two Boolean variables and the three statistical indicators together constitute the input vector of the structural stability calculation function; the structural stability calculation function performs a weighted summation on the input vector based on a set of preset weights, and outputs a score value, which is used to quantify the comprehensive structural disturbance degree of the analysis area.

[0027] In a preferred embodiment, in S5, based on the local disturbance velocity information, membrane surface tension change trend information, and morphological boundary fluctuation information contained in the foam structure data extracted from the current image frame, the numerical change sequences within the continuous time window are extracted respectively, and the three types of sequences are combined to construct a set of three-dimensional time series vector groups;

[0028] Normalization processing is performed on the three-dimensional time series vector group, and a sliding window mechanism is used to extract local feature subsequences. Each local feature subsequence is matched and compared with a reference sequence marked as an abnormal mutation evolution trend phase in the structural evolution map memory library. The matching comparison includes calculating the dynamic time warping distance and the directional gradient cosine similarity.

[0029] In a preferred embodiment, in S5, if the local feature subsequence meets the corresponding preset matching thresholds in both the dynamic time warping distance and the directional gradient cosine similarity, a trend consistency mark is output according to the judgment result, and the current foam structure data is determined to present an abnormal mutation evolution trend phase; otherwise, it is determined that the current foam structure data does not have the characteristics of an abnormal mutation evolution trend phase and is marked as an inconsistent trend.

[0030] Technical effects and advantages of the present invention:

[0031] By constructing a structural evolution graph memory library and introducing a three-dimensional time series trend matching mechanism, the present invention achieves dynamic recognition and response to sudden foam structural changes, solving the failure problem of existing image recognition systems when faced with random mutation foam states that cannot be reconfigured and have no prior support.

[0032] By modeling the foam disturbance velocity, membrane surface tension variation trend, and boundary fluctuation information as three types of time-series variables of structural dynamics, the overall characterization of unstable behaviors during foam evolution is improved, thereby identifying local disturbance behaviors that do not rely on image semantic edges for sudden changes.

[0033] The structural instability detection module implements a linked analysis of morphological continuity interruptions, texture connection breaks, and boundary focus changes, quantifying the structural stability level of local areas. Combined with a scoring mechanism, it locates potential abnormal areas, improving the ability to identify latent mutation bubbles in images at an early stage.

[0034] With the help of the trend comparison method of dynamic time warping and directional gradient cosine similarity, a multi-dimensional matching judgment is made between the current bubble evolution path and the historical abnormal path, which reduces the generalization failure caused by traditional static feature learning and ensures the robustness of the system recognition of non-template and non-repeatable bubble behaviors. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The figure is a flow chart of the method steps of the present invention. DETAILED DESCRIPTION

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0037] Refer to the instruction manual Figure 1 A defoamer foam distribution analysis method based on image feature recognition according to an embodiment of the present invention includes:

[0038] S1. Acquire a foam image sequence in a target area, wherein the foam image sequence is acquired by an imaging device, and image frames of the foam image sequence have temporal continuity;

[0039] S2. Performing a disturbance feature extraction operation on the foam image sequence to obtain local disturbance velocity information, membrane surface tension change trend information, and morphological boundary fluctuation information of the foam edge within a preset time, and using these as foam structure data to characterize the evolution dynamics of the foam structure;

[0040] S3. Constructing a structural evolution atlas memory library, encoding the foam structure data extracted from multiple image frames in time series and annotating their spatial positions, and storing them by evolution stage to form a reference atlas set for characterizing the foam evolution trend;

[0041] S4. Performing structural instability detection in the current image frame, identifying potential abnormal evolution areas by analyzing the morphological continuity interruption, texture connection breakage, and boundary focus change of the foam structure data in the local area;

[0042] S5. Based on the structural evolution atlas memory library, performing trend matching comparison on the foam structure data extracted from the current image frame to determine whether there is a characteristic pattern consistent with the recorded abnormal mutation evolution trend;

[0043] S6. When the comparison result shows that the current foam structure data is consistent with the abnormal mutation evolution trend recorded in the structure evolution map memory library, a mutation recognition signal is generated, and the image recognition system is triggered to execute an abnormal state response accordingly.

[0044] In S2, adjacent image frames in the foam image sequence are temporally registered, and the temporal displacement of corresponding foam edge pixels between image frames is calculated to obtain the local position change rate of the foam edge in continuous time, thereby obtaining local disturbance velocity information;

[0045] When obtaining information on the variation trend of membrane surface tension, the grayscale gradient amplitude of the foam edge region is extracted from each image frame of the foam image sequence through a gradient operation operator. Combined with the variation in the edge contour expansion and contraction rate of the foam edge region in multiple consecutive image frames, the gradient variation rate and contour deformation rate of the unit length boundary on the time axis are calculated, thereby constructing a membrane surface tension variation curve to obtain information used to characterize the variation trend of membrane surface tension.

[0046] In multiple continuous image frames in the foam image sequence, the contour morphology of the foam edge is extracted by an edge detection operator, and the contour matching offset and curvature change rate of the foam edge between adjacent image frames are calculated. The local bending amplitude and contour deformation intensity of the foam edge in the time dimension are statistically analyzed, and a fluctuation change curve reflecting the dynamic change characteristics of the edge is constructed. By performing feature point extraction and frequency domain analysis operations on the fluctuation change curve, the boundary displacement amplitude and curvature disturbance frequency are extracted from it and stored and encoded as morphological boundary fluctuation information.

[0047] In S3, the foam structure data extracted from each image frame in the foam image sequence are numbered in chronological order, and the foam structure data in each frame is labeled with its spatial position index in the image coordinate system, thereby completing the time series encoding and spatial position labeling of the foam structure data;

[0048] Based on the local disturbance velocity information, membrane surface tension change trend information, and morphological boundary fluctuation information contained in the foam structure data, the numerical change range and change rate of the three types of information in the time dimension are extracted respectively, and the evolution state index is constructed according to the corresponding numerical change range and change rate;

[0049] According to the numerical variation range and variation rate corresponding to the local disturbance velocity information, the membrane surface tension variation trend information and the morphological boundary fluctuation information in the evolution state index, the foam structure data is divided into three evolution stages: when the numerical variation range and variation rate of the three types of information are all lower than the corresponding preset evolution judgment threshold, it is defined as a steady-state stage; when the numerical variation range or variation rate of any one type of information exceeds the corresponding preset evolution judgment threshold, it is defined as a transition state stage; when the numerical variation range and variation rate of any two types of information among the three types of information exceed the corresponding preset evolution judgment threshold, it is defined as a mutation state stage; otherwise, the foam structure data is marked as a state to be updated, and the evolution stage classification of the foam structure data is completed accordingly;

[0050] The foam structure data that have completed time series encoding and spatial position annotation are stored in the structural evolution map memory according to the evolution stage they belong to, so as to construct a reference map set for characterizing the foam evolution trend.

[0051] In S4, the step of performing structural instability detection on the current image frame includes:

[0052] S401, in the current image frame, dividing the image area corresponding to the foam structure data into image grid units, and using each grid unit in the image grid units as an analysis area;

[0053] S402. Performing a continuity analysis operation on each analysis region, the continuity analysis operation comprising: detecting the structural integrity of the foam structure in the spatial distribution based on the contour morphological image extracted from the foam structure boundary in the region, identifying whether there are morphological destruction features such as contour fracture, edge segment interruption, or closed structure loss, so as to determine whether there is a morphological continuity interruption in the analysis region; and analyzing the pixel grayscale directional consistency based on the texture gradient distribution map in the analysis region, identifying whether there are abnormal texture features such as texture structure rupture, pattern continuity discontinuity, or gradient direction abrupt change, so as to determine whether there is a texture connection break in the analysis region;

[0054] S403. In each analysis area, based on the boundary position corresponding to the foam structure data, the contour line of the foam structure is extracted by an edge detection algorithm, and the focus response value of each boundary pixel on the contour line is calculated using the Laplace operator; the focus response difference between adjacent pixels along the pixel sequence of the contour line is calculated, and the average value, range and standard deviation of the focus response difference sequence are counted to form a parameter group for describing the boundary focus change; the parameter group of the boundary focus change is input into the structural stability calculation function together with the results of the morphological continuity interruption and the texture connection break. If the output score value of the structural stability calculation function exceeds the preset threshold, the analysis area is marked as an abnormal evolution area.

[0055] The structural stability calculation function inputs the three statistical indicators in the parameter group of boundary focus change as continuous variables, and the three statistical indicators are the average value, range and standard deviation of the focus response difference sequence; at the same time, the results of the morphological continuity interruption and the results of the texture connection break are converted into Boolean variable forms, where when the result is "yes", the value corresponding to the Boolean variable form is 1, and when the result is "no", the value corresponding to the Boolean variable form is 0, and the values ​​of the two Boolean variables and the three statistical indicators together constitute the input vector of the structural stability calculation function; the structural stability calculation function performs a weighted summation on the input vector based on a set of preset weights and outputs a score value, which is used to quantify the comprehensive structural disturbance degree of the analysis area.

[0056] In S5, based on the local disturbance velocity information, membrane surface tension change trend information, and morphological boundary fluctuation information contained in the foam structure data extracted from the current image frame, the numerical change sequences within the continuous time window are extracted respectively, and the three types of sequences are combined to construct a set of three-dimensional time series vector groups;

[0057] Normalization processing is performed on the three-dimensional time series vector group, and a sliding window mechanism is used to extract local feature subsequences. Each local feature subsequence is matched and compared with a reference sequence marked as an abnormal mutation evolution trend phase in the structural evolution map memory library. The matching comparison includes calculating the dynamic time warping distance and the directional gradient cosine similarity.

[0058] In S5, if the local feature subsequence satisfies the corresponding preset matching thresholds in both the dynamic time warping distance and the directional gradient cosine similarity, a trend consistency mark is output according to the determination result, and the current foam structure data is determined to present an abnormal mutation evolution trend phase; otherwise, the current foam structure data is determined to not have the characteristics of an abnormal mutation evolution trend phase and is marked as an inconsistent trend;

[0059] It can be understood that the abnormal mutation evolution trend refers to the characteristic evolution sequence pre-annotated and stored in the structural evolution map memory library, which can reflect the drastic changes in the foam structure in a short period of time. This characteristic evolution sequence is composed of multiple representative foam structure data. Each data includes the numerical evolution process of three types of information in the time dimension, such as local perturbation velocity, membrane surface tension change trend and morphological boundary fluctuation. It is composed of characteristic subsequences extracted by the sliding window mechanism. The whole is used to describe the typical evolution path of the foam under abnormal mutation state and serve as the target reference for current structural trend matching.

[0060] It should also be noted that after the mutation recognition signal is generated, the image recognition system first uses the image frame number and spatial position of the foam structure data corresponding to the signal as input to locate the abnormal area in the current image frame that needs to be analyzed in detail; then the image recognition system re-extracts the morphological features of the foam structure in the abnormal area, including but not limited to the boundary contour image, texture distribution map and focus gradient map, and calculates the structural integrity parameters, texture consistency parameters and focus response difference statistics in the area in turn; then the image recognition system calls the abnormal feature judgment model based on the state vector composed of the above three subclass parameters, performs a parameter threshold comparison operation, and determines whether the evolution trend characteristics of continuous mutation are met; if met, the image recognition system outputs an abnormal state confirmation signal, and updates the image state label of the current frame to "abnormal frame", and at the same time caches the abnormal frame and the corresponding parameters into the abnormal frame cache queue for subsequent trend tracking and state accumulation analysis.

[0061] It should be noted that this approach is based on an understanding of the dynamic evolution of the foam dissipation process. To address the limitations of traditional image feature analysis methods, a systematic distribution analysis method integrating foam disturbance extraction, trend memory matching, and image recognition response was constructed. In the initial implementation phase, a temporally continuous foam image sequence was acquired using imaging equipment to ensure that multiple dynamic behavior characteristics, such as foam edge changes, membrane surface tension fluctuations, and boundary disturbances, could be extracted along the time axis. This design aims to obtain information on the evolution of the foam structure, providing a continuous and reliable data foundation for subsequent structural modeling and state assessment.

[0062] In the disturbance feature extraction module performed on this image sequence, local disturbance velocity is extracted through temporal registration and pixel-level displacement calculation. The membrane surface tension variation curve is constructed by combining grayscale gradient and contour change rate. The fluctuation process of the morphological boundary is further quantified through curvature and offset analysis. The three types of information reflect different physical change mechanisms in the time dimension, complementing each other to form the foam structure data. After encoding, this data is sequentially stored in the structural evolution map memory bank. Based on dimensions such as disturbance rate and trend slope, steady state, transition state, and mutation state are divided into a set of clearly classified reference trend maps. This structural design not only enables the accumulation of knowledge of historical evolution trajectories, but also provides a target reference for subsequent judgment.

[0063] During the analysis, the current frame image is first spatially divided into analysis regions. Continuity analysis is then performed on the integrity of the foam structure's boundaries and consistency of its texture distribution, region by region, identifying morphological continuity interruptions and texture connection breaks. Furthermore, the focus response differences between pixels along the foam boundary are calculated, and the mean, range, and standard deviation are used to form a parameter set for boundary focus variation. The two discrete results (i.e., whether there are interruptions or breaks) and the three focus statistical indices form the inputs to a structural stability scoring function, which outputs a score as a weighted sum. This score is then used to determine whether the current region exhibits an abnormal evolutionary trend of significant structural disturbances.

[0064] At the same time, to identify short-term mutation trends, the scheme also introduces a trend matching mechanism. It extracts a continuous change sequence of three types of dynamic information from the current image frame and constructs it into a three-dimensional time series vector. Through normalization and a sliding window mechanism, it extracts local feature subsequences and performs dual-index matching with reference sequences labeled "abnormal mutation evolution trend phase" in the atlas memory library. Specifically, this involves determining the dynamic time warping distance and the directional gradient cosine similarity. Only when both indicators meet the preset threshold is the trend determined to be consistent, and the current structural evolution is considered to be a mutation process.

[0065] Once trend consistency is confirmed, the system automatically triggers the image recognition module to relocate the abnormal area based on the image frame number and spatial position associated with the mutation. It then conducts a more in-depth structural status review of the area, extracts new feature vectors, and executes anomaly judgment logic to confirm whether the area is indeed in a state of continuous mutation. If confirmed, the frame is marked as an abnormal frame and written to the abnormality cache queue for subsequent status tracking or strategy optimization.

[0066] This scheme uses data-driven time series modeling and graph memory matching strategies to achieve hierarchical recognition and response triggering of complex foam evolution states, enhancing the intelligent perception and dynamic monitoring capabilities of foam dissipation behavior.

[0067] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A defoamer foam distribution analysis method based on image feature recognition, characterized in that: include: S1. Acquire a foam image sequence in a target area, wherein the foam image sequence is acquired by an imaging device, and image frames of the foam image sequence have temporal continuity; S2. Performing a disturbance feature extraction operation on the foam image sequence to obtain local disturbance velocity information, membrane surface tension change trend information, and morphological boundary fluctuation information of the foam edge within a preset time, and using these as foam structure data to characterize the evolution dynamics of the foam structure; S3. Constructing a structural evolution atlas memory library, encoding the foam structure data extracted from multiple image frames in time series and annotating their spatial positions, and storing them by evolution stage to form a reference atlas set for characterizing the foam evolution trend; S4. Performing structural instability detection in the current image frame, identifying potential abnormal evolution areas by analyzing the morphological continuity interruption, texture connection breakage, and boundary focus change of the foam structure data in the local area; S5. Based on the structural evolution atlas memory library, performing trend matching comparison on the foam structure data extracted from the current image frame to determine whether there is a characteristic pattern consistent with the recorded abnormal mutation evolution trend; S6. When the comparison result indicates that the current foam structure data is consistent with the abnormal mutation evolution trend recorded in the structure evolution map memory library, a mutation recognition signal is generated, and accordingly, the image recognition system is triggered to execute an abnormal state response; In S4, the step of detecting structural instability in the current image frame includes: S401, in the current image frame, dividing the image area corresponding to the foam structure data into image grid units, and using each grid unit in the image grid units as an analysis area; S402. Performing a continuity analysis operation on each analysis region, the continuity analysis operation comprising: detecting the structural integrity of the foam structure in the spatial distribution based on the contour morphological image extracted from the foam structure boundary in the region, identifying whether there are morphological destruction features such as contour fracture, edge segment interruption, or closed structure loss, so as to determine whether there is a morphological continuity interruption in the analysis region; and analyzing the pixel grayscale directional consistency based on the texture gradient distribution map in the analysis region, identifying whether there are abnormal texture features such as texture structure rupture, pattern continuity discontinuity, or gradient direction abrupt change, so as to determine whether there is a texture connection break in the analysis region; S403. In each analysis area, based on the boundary position corresponding to the foam structure data, the contour line of the foam structure is extracted by an edge detection algorithm, and the focus response value of each boundary pixel on the contour line is calculated using the Laplace operator; the focus response difference between adjacent pixels along the pixel sequence of the contour line is calculated, and the average value, range and standard deviation of the focus response difference sequence are counted to form a parameter group for describing the boundary focus change; the parameter group of the boundary focus change is input into the structural stability calculation function together with the results of the morphological continuity interruption and the texture connection break. If the output score value of the structural stability calculation function exceeds the preset threshold, the analysis area is marked as an abnormal evolution area.

2. The defoamer foam distribution analysis method based on image feature recognition according to claim 1, characterized in that: In S2, adjacent image frames in the foam image sequence are temporally registered, and the temporal displacement of corresponding foam edge pixels between image frames is calculated to obtain the local position change rate of the foam edge in continuous time, thereby obtaining local disturbance velocity information; When obtaining information on the variation trend of membrane surface tension, the grayscale gradient amplitude of the foam edge region is extracted from each image frame of the foam image sequence through a gradient operation operator. Combined with the variation in the edge contour expansion and contraction rate of the foam edge region in multiple consecutive image frames, the gradient variation rate and contour deformation rate of the unit length boundary on the time axis are calculated, thereby constructing a membrane surface tension variation curve to obtain information used to characterize the variation trend of membrane surface tension. In multiple continuous image frames in the foam image sequence, the contour morphology of the foam edge is extracted by an edge detection operator, and the contour matching offset and curvature change rate of the foam edge between adjacent image frames are calculated. The local bending amplitude and contour deformation intensity of the foam edge in the time dimension are statistically analyzed, and a fluctuation change curve reflecting the dynamic change characteristics of the edge is constructed. By performing feature point extraction and frequency domain analysis operations on the fluctuation change curve, the boundary displacement amplitude and curvature disturbance frequency are extracted from it and stored and encoded as morphological boundary fluctuation information.

3. The defoamer foam distribution analysis method based on image feature recognition according to claim 2, characterized in that: In S3, the foam structure data extracted from each image frame in the foam image sequence are numbered in chronological order, and the foam structure data in each frame is labeled with its spatial position index in the image coordinate system, thereby completing the time series encoding and spatial position labeling of the foam structure data; Based on the local disturbance velocity information, membrane surface tension change trend information, and morphological boundary fluctuation information contained in the foam structure data, the numerical change range and change rate of the three types of information in the time dimension are extracted respectively, and the evolution state index is constructed according to the corresponding numerical change range and change rate; According to the numerical variation range and variation rate corresponding to the local disturbance velocity information, the membrane surface tension variation trend information and the morphological boundary fluctuation information in the evolution state index, the foam structure data is divided into three evolution stages: when the numerical variation range and variation rate of the three types of information are all lower than the corresponding preset evolution judgment threshold, it is defined as a steady-state stage; when the numerical variation range or variation rate of any one type of information exceeds the corresponding preset evolution judgment threshold, it is defined as a transition state stage; when the numerical variation range and variation rate of any two types of information among the three types of information exceed the corresponding preset evolution judgment threshold, it is defined as a mutation state stage; Otherwise, the foam structure data is marked as being in a state to be updated, and the evolution stage classification of the foam structure data is completed accordingly; The foam structure data that have completed time series encoding and spatial position annotation are stored in the structural evolution map memory according to the evolution stage they belong to, so as to construct a reference map set for characterizing the foam evolution trend.

4. The defoamer foam distribution analysis method based on image feature recognition according to claim 3, characterized in that: The structural stability calculation function inputs the three statistical indicators in the parameter group of boundary focus change as continuous variables, and the three statistical indicators are the mean, range and standard deviation of the focus response difference sequence; at the same time, the results of morphological continuity interruption and texture connection break are converted into Boolean variable forms, where when the result is "yes", the corresponding value of the Boolean variable form is 1, and when the result is "no", the corresponding value of the Boolean variable form is 0, and the values ​​of the two Boolean variables and the three statistical indicators together constitute the input vector of the structural stability calculation function; the structural stability calculation function performs a weighted summation on the input vector based on a set of preset weights and outputs a score value, which is used to quantify the comprehensive structural disturbance degree of the analysis area.

5. The defoamer foam distribution analysis method based on image feature recognition according to claim 4, characterized in that: In S5, based on the local disturbance velocity information, membrane surface tension change trend information, and morphological boundary fluctuation information contained in the foam structure data extracted from the current image frame, the numerical change sequences within the continuous time window are extracted respectively, and the three types of sequences are combined to construct a set of three-dimensional time series vector groups; Normalization processing is performed on the three-dimensional time series vector group, and a sliding window mechanism is used to extract local feature subsequences. Each local feature subsequence is matched and compared with a reference sequence marked as an abnormal mutation evolution trend phase in the structural evolution map memory library. The matching comparison includes calculating the dynamic time warping distance and the directional gradient cosine similarity.

6. The defoamer foam distribution analysis method based on image feature recognition according to claim 5, characterized in that: In S5, if the local feature subsequence meets the corresponding preset matching thresholds in both the dynamic time warping distance and the directional gradient cosine similarity, a trend consistency mark is output according to the judgment result, and the current foam structure data is determined to present an abnormal mutation evolution trend phase; otherwise, it is determined that the current foam structure data does not have the characteristics of an abnormal mutation evolution trend phase and is marked as an inconsistent trend.

Images