Sewage suspended particle form evolution evaluation method based on machine vision and time sequence modeling

Abstract

The invention relates to the technical field of sewage treatment, in particular to a sewage suspended particle morphology evolution evaluation method based on machine vision and time sequence modeling, which comprises the following steps: acquiring a sewage system suspended particle image frame, extracting edge pixel coordinates of suspended particles, measuring a region area, a boundary length and a central point, screening a morphology stable frame segment and marking, and extracting the edges of the head and tail frames, comparing to form a coverage evolution graph, analyzing background gray and gradient to obtain a brightness correction graph, and marking a stable orientation propulsion structure. According to the invention, the spatial scale and time sequence linkage analysis is carried out on the edge pixels of the suspended particle area in the image sequence, so that the precise recognition and stability labeling of the evolution characteristics of the suspended particle substance structure in the dynamic reaction environment are realized; the multi-dimensional modeling capability of boundaries, morphological evolution and geometric paths of suspended particles in a sewage treatment system is effectively improved, the problems that sample collection is tedious and parameter calibration is prone to deviation are solved, and the automation level of image recognition and the continuity, stability and precision of reaction state evaluation are improved.

Description

Technical Field

[0001] This invention relates to the field of wastewater treatment technology, and in particular to a method for assessing the morphological evolution of suspended particles in wastewater based on machine vision and time-series modeling. Background Technology

[0002] Wastewater treatment technology refers to the related technological systems for purifying and treating various polluted water bodies, such as urban sewage and industrial wastewater. Its core objective is the efficient removal of pollutants such as organic matter, nitrogen and phosphorus nutrients, suspended solids, and colloids. The efficiency of this process is closely related to the physical characteristics and dynamic behavior of key suspended particulate matter in the reactor (such as suspended particles formed during flocculation and coagulation, activated sludge flocs, granular sludge, and biofilm fragments in biological treatment systems). The size, morphology, and evolutionary trends of these particles directly determine the solid-liquid separation effect, the biochemical reaction rate, and the final effluent quality. Therefore, achieving accurate monitoring and assessment of suspended particulate communities in water bodies is crucial for optimizing process control and improving treatment efficiency and stability.

[0003] Currently, in this technical field, the monitoring of the physical characteristics of suspended particles mainly relies on offline sampling laboratory analysis or traditional online image recognition methods. The former, such as microscopic observation and laser particle size analysis, while highly accurate, suffers from inherent drawbacks including strong lag, enormous workload, and inability to reflect dynamic process changes. The latter, traditional image recognition methods, refers to establishing a mapping relationship between particle images and their physical characteristics (such as particle size and morphology) through artificial intelligence algorithms, thereby achieving automatic identification. These methods typically require a large number of manually labeled sample images to build a feature database for model training. To ensure the model's generalization ability under different water quality, reagent, and reaction conditions, the sample library construction needs to cover multiple scenarios. This process is labor-intensive, costly, and highly susceptible to labeling errors and calibration deviations introduced by human intervention and fluctuations in experimental conditions, constituting a bottleneck for technology promotion.

[0004] Furthermore, traditional methods often focus on identifying static or transient features, making them ill-suited for the continuous and dynamic evolution of particle structures in real-world water treatment systems. Especially when processing consecutive frames of images, they cannot effectively track the morphological changes, movement paths, and stability transitions of individual or group particles, resulting in significant delays in state assessment, decreased identification accuracy and reliability, and an inability to meet the stringent requirements of real-time online monitoring and closed-loop control. Ultimately, this impacts the optimization and stable operation of the overall treatment process. Summary of the Invention

[0005] To address the problems of existing technologies, such as large data acquisition workload and long time cycle, limitations due to manual intervention and changes in experimental conditions, poor sample consistency and accumulated parameter calibration deviations, difficulty in adapting to the dynamic evolution of suspended particle structure during flocculation reactions, and especially in the inability to accurately track the morphological change trajectory of flocs in continuous frame image recognition, resulting in delays in state assessment and decreased recognition accuracy, thus affecting the overall treatment efficiency and stability in practical online monitoring and control applications, this invention provides a wastewater suspended particle morphology evolution assessment method based on machine vision and time-series modeling. The technical solution is as follows: On the one hand, a method for assessing the morphological evolution of suspended particles in wastewater based on machine vision and time-series modeling is provided. This method includes: S1: Acquire image frames of suspended particles in the wastewater system, extract the set of edge pixel coordinates of suspended particles in each image frame, and sequentially determine the area of ​​the region formed by the boundary of suspended particles in the image, the length of the boundary line segment and the center point of the region enclosed by the edge contour to obtain the set of morphologically stable image segment identifiers. S2: Call the morphologically stable image segment identifier set, extract the set of floating particle edge pixels of the first and last image frames of the morphologically stable image segment, perform spatial transfer comparison based on the geometric enclosing area in the image plane, and obtain the contour coverage evolution map. S3: Call the outline coverage evolution map, identify the number of pixels in the suspended particle region and locate the coordinates of the centroid of the gray-scale distribution, construct the centroid path map according to the frame sequence, record the changes in the number of pixels, and obtain continuous records of the centroid path. S4: Call the continuous recording of the center of gravity path, extract the position of the gray concentration point and the gradient direction vector of the non-target background area, analyze the spatial displacement trend, and compare the intersection with the boundary position of the suspended particle area to mark the image segment affected by the illumination shift and obtain the edge brightness correction section mapping map.

[0006] As a further aspect of the present invention, the morphologically stable image segment identifier set includes the regional morphologically stable frame sequence number, edge contour stability feature value, standard deviation of center position change, area change range evaluation result, and intra-frame temporal consistency label. The contour coverage evolution map includes the contour bounding boundary evolution curve, spatial bounding overlap heat map, suspended particle region contour expansion direction distribution map, and first and last edge overlap coefficient map. The centroid path continuous record includes the pixel centroid coordinate sequence, gray-level centroid offset path map, inter-frame centroid offset speed curve, and suspended particle pixel number change trend map. The edge brightness correction segment mapping map includes the gray-level offset concentrated region distribution map, background region illumination gradient direction map, edge recognition deviation mapping map, and image brightness error correction label set.

[0007] As a further aspect of the present invention, the step of obtaining the morphologically stable image segment identifier set specifically includes: S101: Obtain image frames of suspended particles in the wastewater system, extract the contour line pixels formed by the boundary of the suspended particle region in each frame of the image in turn, call the set of contour pixel coordinates corresponding to the boundary of the suspended particles, summarize the number of pixels in the region formed by the contour line, count the number of bus segments of the contour line, calculate the centroid coordinates of the bounding box pixels of the region formed, and obtain the image frame boundary feature data. S102: Based on the image frame boundary feature data, calculate the horizontal and vertical offset distances of the centroid coordinates between adjacent frames sequentially according to time, make a continuous judgment on the trend of offset distance change, and filter out image frame segments with continuous increasing or decreasing directions to obtain a set of boundary stable frame segment data. S103: Call the corresponding image frame sequence in the boundary stable frame segment data set, and compare the line segment direction, closure degree and line segment distribution width of the suspended particle outline boundary in each image frame. Select image frames with consistent line segment direction, similar closure level and stable width variation to form an image segment. Perform unified classification processing according to the structural characteristics of the image segment to obtain a set of morphologically stable image segment identifiers.

[0008] As a further aspect of the present invention, the step of obtaining the contour coverage evolution map specifically includes: S201: Call each image frame sequence marked by the morphologically stable image segment identifier set, extract the contour pixel coordinate set of the suspended particle edge in the first frame and the last frame, calculate the contour mapping difference value, map the contour pixel set to the corresponding image plane according to the original resolution of the image frame, construct the closed edge graphics of the suspended particle region in the two frames respectively, merge the pixel regions of the plane range occupied by the closed edge graphics, and generate the closed region dataset of the first and last contour images. S202: Call the coordinates of the edge regions of the first and last frames in the closed region dataset of the first and last contour images, perform the geometric enclosure relationship determination of the two sets of closed regions in the image plane, analyze the expansion direction and movement trend of the suspended particle region in the image plane based on the boundary expansion position of the circumscribed rectangle and the coverage ratio of the overlapping boundary, and draw the boundary expansion change form of the differentiated image segments by overlapping the line trajectory to obtain the contour coverage evolution map.

[0009] As a further aspect of the present invention, the contour mapping difference value measures the degree of spatial difference between the contour point and the global average contour position after the contour point is mapped to the image plane at the original image resolution.

[0010] As a further aspect of the present invention, the step of obtaining the continuous records of the centroid path specifically includes: S301: Call the suspended particle boundary contour region of the contour coverage evolution map, identify the pixel position in the contour region in each frame image, and count the total number of pixels in the contour region. At the same time, extract the set of gray values ​​of the corresponding pixels in the original gray matrix of the image, and perform a weighted average calculation by combining the pixel coordinates and gray values ​​to obtain the gray distribution centroid coordinates of the suspended particle region in each frame image, and generate the set of suspended particle centroid and pixel count of the image frame. S302: Based on the centroid coordinates of the image frames in the set of centroids of suspended particles and the number of pixels in the image frames, the centroids are connected sequentially in the order of the image frame sequence to form a connecting line path. The points on the connecting line path are used as reference points to record the position offset distance between adjacent frames. The number of pixels in the corresponding frame is marked as auxiliary information to obtain the centroid path image data of the suspended particle region. S303: Call the start and end positions of the path segments and the corresponding frame numbers in the image data of the center of gravity path of the suspended particle region, extract the direction of change of the center of gravity position and the magnitude of the distance change between consecutive frames, arrange the number of pixels marked on multiple points on the path in sequence, and synchronously record the number of pixels in each path segment to obtain a continuous record of the center of gravity path.

[0011] As a further aspect of the present invention, the step of obtaining the edge brightness correction segment mapping map specifically includes: S401: Call the continuous recording of the center path, exclude the set of pixel coordinates formed by the boundary of the suspended particle region in the corresponding frame image, extract the concentrated points of pixels with dense gray value changes in the remaining image region, identify the position gradient of the gray value difference direction of the pixels adjacent to the concentrated points, calculate the gray density value, determine the continuous displacement direction of the gray concentration region on the image plane according to the arrangement order of the main vector position of the gradient change direction, and generate the gray main vector trend group of non-target region. S402: Call the displacement trajectory and direction change of the gray-level principal vector of the continuous frame in the gray-level principal vector trend group of the non-target area, and combine it with the coordinate segment occupied by the boundary of the suspended particle area to perform cross mapping on the overlapping boundary position of the two on the image plane. Mark the area where the gray-level principal vector extension trajectory intersects with the contour line formed by the boundary of the suspended particle as the interference overlapping segment, record the frame number and area position index of the overlapping segment, and obtain the edge brightness correction segment mapping map.

[0012] As a further aspect of the present invention, the gray density value represents the degree of spatial concentration of pixel gray level changes within an image region.

[0013] As a further aspect of the present invention, the method further includes step S5: S5: Based on the outline of the suspended particles corresponding to the edge brightness correction section mapping map, extract the orientation angle of the main axis direction in consecutive frames, combine the continuity of the main axis direction in adjacent frames with the degree of overlap of the boundary points within the structure, determine whether the structure is maintained in a unified geometric advancement trajectory, mark the suspended particle structure with stable orientation advancement characteristics, and obtain the main axis direction stable structure label set. The main axis orientation stable structure label set includes main axis orientation stability index, main axis evolution trajectory curve, structural boundary continuity distribution value, and directional propulsion consistency identification label.

[0014] As a further aspect of the present invention, the step of obtaining the main axis direction stable structure label set specifically includes: S501: Based on the edge contour region of the suspended particles in the edge brightness correction section mapping map, extract the set of pixel coordinates of the edge of the suspended particles in each frame image in sequence, perform circumscribed ellipse fitting on the set of pixel coordinates, obtain the orientation angle value of the major axis direction of the circumscribed ellipse in the image plane, and arrange the main axis orientation angles corresponding to the image frames in sequence to generate the image frame main axis direction angle sequence. S502: Call the main axis orientation angle values ​​of consecutive frames in the image frame main axis orientation angle sequence, combine the number of overlapping pixels between the pixel sets of the suspended particle structure boundary contour in adjacent frames, synchronously compare the range of change of the number of overlapping pixels with the range of difference of the main axis angle, filter out the frame segments with abrupt changes in the main axis angle and deviations in the number of overlapping pixels, retain the frame segments with gentle angle changes and continuous distribution of boundary overlapping points, and generate a set of continuous frame segments in the structural orientation. S503: Call the boundary region of the suspended particle structure in the set of continuous frame segments in the structural direction, mark the frame segments that maintain the continuous trend of the change of the main axis orientation angle and whose contour distribution is linearly progressive in the image space, extract the frame segment number and the index number of the covered suspended particle structure position, and obtain the main axis direction stable structure label set.

[0015] As a further aspect of the present invention, the method is applied to the flocculation and sedimentation process of wastewater treatment or the biochemical treatment process of activated sludge, to monitor and evaluate the morphology, trajectory, and dynamic evolution behavior of flocculents and activated sludge, so as to optimize the treatment process parameters.

[0016] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By performing spatial-scale and temporal-series linked analysis on the edge pixels of suspended particle regions in image sequences, and combining a morphological stability screening mechanism, contour evolution trajectory construction method, and gray-level centroid path extraction technique, this method introduces background gray-level interference detection and principal axis direction consistency judgment to achieve accurate identification and stability annotation of suspended particle structural evolution characteristics in dynamic reaction environments. This effectively improves the multi-dimensional modeling capability of suspended particle boundaries, morphological evolution, and geometric paths, reduces reliance on manually labeled samples and offline feature libraries, avoids cumbersome sample collection and parameter calibration bias issues, and enhances the automation level of image recognition and the continuity, stability, and accuracy of reaction state assessment. Furthermore, this method, through precise monitoring and evaluation of the suspended particle evolution process, empowers the intelligent control of wastewater treatment processes, promotes refined process management and energy conservation, and provides technical support for water environment protection and sustainable development. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the workflow of the present invention; Figure 2 This is a detailed flowchart of S1 of the present invention; Figure 3 This is a detailed flowchart of the S2 process of the present invention; Figure 4 This is a detailed flowchart of the S3 process of the present invention; Figure 5 This is a detailed flowchart of the S4 process of the present invention; Figure 6 This is a detailed flowchart of S5 of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0019] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0020] To make the technical problems, technical solutions and advantages of this invention clearer, the following will take the morphological evolution evaluation of suspended particles in the flocculation-coagulation unit of wastewater treatment as an example, and describe it in detail with reference to the accompanying drawings.

[0021] Please see Figure 1 This invention provides a method for assessing the morphological evolution of suspended particles in wastewater based on machine vision and time-series modeling. The processing flow of this method may include the following steps: S1: Acquire image frames of suspended particles in the wastewater system, extract the set of edge pixel coordinates of the suspended particle region in each image frame, and sequentially determine the area of ​​the region formed by the boundary of suspended particles in the image, the length of the boundary line segment, and the position of the center point of the region enclosed by the edge contour. Filter continuous frame segments with consistent morphological advancement trajectory, no jumps in the center point, and stable area distribution according to the time sequence. Classify and label the continuous frame segments to obtain a set of morphologically stable image segment identifiers. S2: Call the morphologically stable image segment identifier set, extract the set of floating particle edge pixels of the first and last image frames of the morphologically stable image segment, perform spatial transfer comparison based on the geometric enclosing area in the image plane, and obtain the contour coverage evolution map. S3: Call the contour coverage evolution map, identify the number of pixels in the suspended particle region and locate the coordinates of the centroid of grayscale distribution, construct the centroid path map according to the frame sequence, record the changes in the number of pixels, and obtain continuous records of the centroid path. S4: Call the continuous recording of the center of gravity path, extract the position of the gray concentration point and the gradient direction vector of the non-target background area, analyze the spatial displacement trend, and compare the intersection with the boundary position of the suspended particle area. Mark the image segments affected by the illumination shift and obtain the edge brightness correction section mapping map. S5: Based on the outline of the suspended particles corresponding to the edge brightness correction section mapping map, extract the orientation angle of the main axis direction in consecutive frames. Combine the continuity of the main axis direction in adjacent frames with the degree of overlap of the boundary points within the structure to determine whether the structure is maintained in a unified geometric advancement trajectory. Mark the suspended particle structure with stable orientation advancement characteristics and obtain the main axis direction stable structure label set. The morphologically stable image segment identifier set includes the region morphologically stable frame sequence number, edge contour stability feature value, standard deviation of center position change, area change range evaluation result, and intra-frame temporal consistency label. The contour coverage evolution map includes the contour bounding boundary evolution curve, spatial bounding overlap heat map, suspended particle region contour expansion direction distribution map, and first and last edge overlap coefficient map. The centroid path continuous record includes the pixel centroid coordinate sequence, gray-level centroid offset path map, inter-frame centroid offset velocity curve, and suspended particle pixel number change trend map. The edge brightness correction segment mapping map includes the gray-level offset concentrated area distribution map, background region illumination gradient direction map, edge recognition deviation mapping map, and image brightness error correction label set. The main axis direction stable structure label set includes the main axis orientation stability index, main axis evolution trajectory curve, structural boundary continuity distribution value, and direction advancement consistency recognition label.

[0022] Please see Figure 2 The specific steps for obtaining the morphologically stable image segment identifier set are as follows: S101: Obtain image frames of suspended particles in the wastewater system, extract the contour line pixels formed by the boundary of the suspended particle region in each frame of the image in turn, call the set of contour pixel coordinates corresponding to the boundary of the suspended particles, summarize the number of pixels in the region formed by the contour line, count the number of bus segments of the contour line, calculate the centroid coordinates of the bounding box pixels of the region formed, and obtain the image frame boundary feature data. The pixel coordinate set of the boundary contour of the suspended particle region in each frame of the image is extracted sequentially. This requires acquiring consecutive image frames at equal time intervals using an image acquisition device. During image processing, the RGB color image is converted to grayscale to highlight edge contrast. After reducing background noise through Gaussian filtering, Canny edge detection is applied to extract edge pixels. Then, the edge contour extraction function is called to obtain continuous closed boundaries. Pixels within each closed region are checked point by point using a scanning method to determine if they are within the contour's bounding area. The number of pixels inside is counted as the representative value of the contour region. Line segments in the contour boundary are extracted, and the total number of line segments is calculated. The minimum bounding rectangle method is called to obtain the bounding box of the suspended particle region, and its centroid position is calculated. This position is calculated from the midpoint of the diagonal coordinates of the rectangle. The upper left corner of the bounding box is set to (100, 120), and the lower right corner to (200, 220), so the centroid is (150, 170). This process is repeated to obtain the above feature data of the suspended particle boundary in each frame of the image, and the data is then organized into image frame boundary feature data.

[0023] S102: Based on the image frame boundary feature data, calculate the horizontal and vertical offset distances of the centroid coordinates between adjacent frames sequentially over time, determine the continuity of the offset distance change trend, and filter out image frame segments with continuously increasing or decreasing directions to obtain a set of boundary stable frame segment data. The displacement of the center of gravity of each suspended particle region between frames is analyzed in chronological order. The horizontal and vertical offset values ​​of the center of gravity are calculated between each pair of adjacent frames, that is, the difference between the horizontal and vertical coordinates of the center of gravity of the current frame and the previous frame, which are called the horizontal offset and vertical offset, respectively. After obtaining the difference, a center of gravity displacement sequence is established, and the changing trend of the sequence between consecutive frames is observed. By traversing, it is determined whether the increasing or decreasing state is continuous and consistent. If the horizontal offset value of multiple consecutive frames is positive, it means that the center of gravity of that segment has moved to the right, and it is recorded as a "continuous positive horizontal offset" state frame segment. The start and end frame numbers of each segment are recorded and its characteristic data is saved. By judging the continuity and directional consistency of the changing trend, frame segments with stable boundary motion trends are selected. If the horizontal offset value between frame 15 and frame 24 continues to increase and the vertical change is small, it can be judged as a boundary stable segment. The set of frame segments that meet the stability conditions constitutes the boundary stable frame segment data set.

[0024] S103: Call the corresponding image frame sequence in the boundary stable frame segment data set, compare the line segment direction, closure degree and line segment distribution width of the suspended particle outline boundary in each image frame, select image frames with consistent line segment direction, similar closure level and stable width variation to form image segments, perform unified classification processing according to the structural characteristics of image segments, and obtain the morphologically stable image segment identifier set. For each frame, the direction of each line segment of the suspended particle outline is extracted. The direction angle is calculated by extracting the coordinates of the start and end points of each line segment in the outline. The direction angles of the line segments are then summarized, and the difference between the maximum and minimum values ​​is observed. If the difference is small, the line segment direction is considered to be consistent. The method of further judging whether each outline is a closed region is to compare the Euclidean distance between the start and end points of the outline. If it is less than a certain distance threshold, the degree of closure is considered to be high. This judgment is performed on the outlines in the frame, and the average level of closure of each outline is statistically analyzed. When judging the change in the width of the line segment distribution, the maximum width of the outline in the horizontal direction of the image is measured and compared with the previous and next frames. If the change is small, the width is considered to be stable. The above three dimensions are combined to judge and select image frames with basically consistent line segment directions, similar closure levels, and small width fluctuations. The image frames are grouped into a unified segment, and the segment is labeled with a classification number according to the overall outline distribution shape of the segment. Type I segment is set as directional convergence type and Type II segment as directional divergence type. The stable image segment identifier set is obtained.

[0025] Please see Figure 3 The specific steps for obtaining the contour cover evolution map are as follows: S201: Call the image frame sequence marked by the morphologically stable image segment identifier set, extract the contour pixel coordinate set of the suspended particle edge in the first frame and the last frame, calculate the contour mapping difference value, map the contour pixel set to the corresponding image plane according to the original resolution of the image frame, construct the closed edge graphics of the suspended particle region in the two frames respectively, merge the pixel regions of the plane range occupied by the closed edge graphics, and generate the closed region dataset of the first and last contour images. The contour mapping difference value is calculated using the following formula: ; in, Representative at the The first frame Contour mapping difference value of each contour point Representing the The horizontal pixel coordinates of each contour point when extracted from an image frame. Representing the The vertical pixel coordinates of each contour point when extracted from an image frame. Represents the original horizontal resolution of the image frame. Represents the original vertical resolution of the image frame. Representing the The horizontal pixel coordinates of each contour point Representing the The vertical pixel coordinates of each contour point Represents the total number of pixels in the outline; Operational logic description: Spatial mapping is performed by multiplying the horizontal and vertical pixel coordinates of each contour point by the original image resolution, and then performing a difference operation with the weighted average of the contour point coordinates to calculate the relative offset of each point on the mapping plane. The difference is then divided by the total number of points for mean normalization, and then multiplied by the magnitude of the resolution vector to unify the measurement scale. The absolute value is taken to obtain a non-negative real number, which quantifies the offset intensity of the contour point in the image space. The overall structure comprehensively utilizes multiple operations such as multiplication, summation, division, square root and absolute value to present the mapping difference between local position and global structure in a numerical way with consistent scale. The contour mapping difference value is used to measure the spatial difference between a contour point and the global average contour position after the contour point is mapped to the image plane at the original image resolution. The larger the value, the more obvious the deviation of the contour point from the overall contour structure. It can effectively depict the boundary change trend and serve as a numerical basis for subsequent region judgment and merging. Parameter assignment basis: The original resolution of the image frame is 1920×1080 pixels, therefore let... This is the resolution of common high-definition monitoring images, obtained by reading camera settings files and image metadata; Select the number of contour points (Typical extraction quantity), the coordinates of the five points are obtained from the actual selection: ; The coordinate values ​​are obtained through pixel-level segmentation software or contour extraction algorithms. Calculate the reference sum: ; single point Difference values: ; First calculate the sub-items: ; Reference for reduction: ; Divide by 5: ; Resolution norm: ; Multiplication: ; Take the absolute value: ; Threshold / Benchmark Setting: Introducing a benchmark value The noise impact range can be reasonably referenced. The benchmark value is calculated by averaging five identical n=5 points extracted from the non-target area. Worth selling, typical range is to In this example, we assume the median value is 5e8. Therefore, If the value is greater than the baseline, it indicates a significant difference in point mapping, which can serve as the basis for subsequent regional fusion decisions. The results show that there is a direct correspondence between the actual example and the step results, and the difference value is used as the contour mapping difference quantity for the generation of the closed region dataset of the first and last contour images; The advantage of this formula is that by scaling the coordinates of each contour point, comparing it with the overall average weighted value, multiplying it by the resolution norm, and taking the absolute value, it comprehensively reflects the magnitude of the contour mapping offset, providing a quantitative basis for region fusion. Table 1: Contour Point Monitoring Coordinates and Intermediate References and Data Table

[0026] As shown in Table 1, the table lists the point number, horizontal and vertical coordinates, and weighted average value, and calculates the reference sum as 2,632,500, which verifies the aforementioned summation process.

[0027] S202: Call the coordinates of the edge regions of the first and last frames in the closed region dataset of the first and last contour images, perform the geometric enclosing relationship determination of the two sets of closed regions in the image plane, analyze the expansion direction and movement trend of the suspended particle region in the image plane based on the boundary expansion position of the circumscribed rectangle and the coverage ratio of the overlapping boundary, and draw the boundary expansion change form of the differentiated image segments by overlapping the line trajectory to obtain the contour coverage evolution map. The pixel coordinate sets of enclosed regions in the first and last frames are extracted separately, and geometric relationship determination operations are performed on them to generate the minimum bounding rectangle of each enclosed region. The rectangle boundary is obtained by statistically analyzing the minimum and maximum values ​​of the x and y coordinates in the contour. Then, the positional relationship between the two rectangles is determined. The criteria are set to determine whether the rectangle in the last frame encloses the rectangle in the first frame, or whether the two intersect. The coverage is calculated based on the pixel percentage of the intersecting area. The movement direction of the suspended particle region on the image plane is determined based on the change in the center position of the bounding rectangle. If the rectangle in the last frame deviates from the position in the first frame, a significant change in the suspended particle region is detected. If the difference in area between two regions is large, then there is an expansion behavior. By drawing line trajectories of the first and last contour boundaries in the same image coordinate system and overlapping them to show the change process of the display area, the upper left corner of the first frame contour boundary is set to (120, 100), the lower right corner is set to (300, 280), and the last frame is set to (150, 130) - (330, 310). Then it can be determined that the region has drifted 30 pixels to the lower right, and the area has expanded to 1.21 times the original size. The first and last contour boundaries are represented by red and blue lines, which intuitively show the evolution relationship on the image and obtain the contour coverage evolution map.

[0028] Please see Figure 4 The specific steps for obtaining continuous records of the centroid path are as follows: S301: Call the suspended particle boundary contour region of the contour coverage evolution map, identify the pixel position in the contour region in each frame image, and count the total number of pixels in the contour region. At the same time, extract the set of gray values ​​of the corresponding pixels in the original gray matrix of the image, and perform weighted average calculation by combining the pixel coordinates and gray values ​​to obtain the gray distribution centroid coordinates of the suspended particle region in each frame image, and generate the set of suspended particle centroid and pixel count of the image frame. In each frame, the closed graphic region enclosed by the contour is first identified. By traversing the image coordinate points inside the contour, the position of the pixels inside the boundary is confirmed. The effective pixel set within the contour is quickly filtered out using a mask. The effective pixel set is counted to obtain the total number of pixels in the contour region of the current frame. At the same time, the original grayscale matrix of the image is called back to extract the grayscale values ​​corresponding to the above pixel positions, forming a grayscale value set within the region. The grayscale values ​​are distributed between 0 and 255, representing the distribution of different brightness in the region. Combining the pixel coordinate information and grayscale values, the set is subjected to grayscale weighted average processing to obtain the grayscale distribution centroid position of the suspended particle region in the current frame. By traversing the horizontal and vertical coordinates of each pixel and its grayscale value, the data is combined and calculated to generate a single centroid coordinate point for subsequent path tracking analysis. Each frame needs to perform a complete pixel localization within the contour, grayscale value extraction, and grayscale centroid coordinate calculation to generate a set of image frame suspended particle centroid and pixel count.

[0029] S302: Based on the centroid coordinates of the image frames in the set of centroids of suspended particles and the number of pixels, the centroids are connected sequentially in the order of the image frame sequence to form a connecting line path. The points on the connecting line path are used as reference points to record the position offset distance between adjacent frames. The number of pixels in the corresponding frame is marked as auxiliary information to obtain the centroid path image data of the suspended particle region. By connecting the centroid coordinates of each frame in the chronological order of the image sequence, a continuous centroid path image is constructed, forming a trajectory composed of multiple centroid points. This path truly reflects the movement trend of the suspended particle region over time. Each connection point represents the gray-scale centroid coordinate position in the image frame, and the positional difference between adjacent connection points represents the actual movement distance of the centroid in the image. At the same time, the pixel count information corresponding to each point can be bound to the connection line to observe the relationship between the centroid shift and the change in the density of suspended particles. In the process of constructing the path, the gray-scale centroid coordinates of suspended particles need to be extracted frame by frame, and the points are plotted on a unified image coordinate system. The points are connected by line segments to form a visualized time path diagram. Each node is labeled with the corresponding frame number and the number of pixels in the region. This path diagram is not only used for morphological trend analysis, but also facilitates the identification of whether the region is concentrated or dispersed. It is one of the important bases for analyzing the movement pattern of suspended particles in the flocculation area of ​​the sedimentation tank, and obtaining centroid path image data of the suspended particle region.

[0030] S303: Call the start and end positions of the path segments and the corresponding frame numbers in the image data of the centroid path of the suspended particle region, extract the direction of change of the centroid position and the magnitude of the distance change between consecutive frames, arrange the number of pixels marked on multiple points on the path in sequence, and synchronously record the number of pixels in each path segment to obtain a continuous record of the centroid path. The start and end positions of each path segment and its corresponding image frame number are extracted one by one. The line connecting each two points in the path can be regarded as a trajectory of the center of gravity movement. By comparing the horizontal and vertical changes in the center of gravity coordinates of the preceding and following frames, the direction of regional offset is determined. The coordinate difference between the start and end points of each line segment is used to identify whether the suspended particle region has undergone directional expansion or irregular displacement. Auxiliary information is extracted from the points on the path, mainly including the number of pixels in the suspended particle region of the corresponding frame. The number of pixels is arranged in chronological order to form a continuously changing pixel sequence. The correlation between the change in regional density and the direction of movement is analyzed in combination with the direction of change of the center of gravity. If the number of pixels in a certain path segment continues to increase and the direction remains consistent, it indicates that the suspended particles in that region are concentrating in a specific direction. Conversely, if the number of pixels decreases rapidly and is accompanied by multiple changes in the direction of the center of gravity, there is regional diffusion or boundary fragmentation. The start and end coordinates, frame number, displacement direction, change amplitude and pixel number sequence of each path segment are formed into a one-to-one corresponding data record to obtain a continuous record of the center of gravity path.

[0031] Please see Figure 5 The specific steps for obtaining the edge brightness correction section mapping map are as follows: S401: Call the continuous recording of the center of gravity path, exclude the set of pixel coordinates formed by the boundary of the suspended particle region in the corresponding frame image, extract the concentrated points of pixels with dense gray value changes in the remaining image region, identify the position gradient of the gray value difference direction of the pixels adjacent to the concentrated points, calculate the gray density value, determine the continuous displacement direction of the gray concentration region on the image plane according to the arrangement order of the principal vector position of the gradient change direction, and generate the gray principal vector trend group of non-target region. The grayscale density value is calculated using the following formula: ; in, Representing the The gray density value of each candidate gray-level concentration region Representing the The first region grayscale value of each pixel. Representing the The average of pixel grayscale values ​​within a region. , Representing the first In the region, the first The x and y coordinates of each pixel. , Representing the first The average of the x and y coordinates of pixels in each region, Represents the total number of pixels. This represents the average grayscale value of pixels in the non-suspended particle region of the image. The formula calculation logic is as follows: By weighting the gray-level difference of each pixel in the candidate region with its relative position distance, the spatial distribution density of gray-level changes within the region is evaluated. The difference between the gray-level value of each pixel and the average gray-level value of the region is calculated, and combined with its Euclidean distance relative to the center of the region, a coupling model is constructed. The sum of the coupling results of the pixels is used as the numerator, reflecting a comprehensive index of the overall gray-level and spatial dispersion. The denominator is normalized using the number of pixels and the absolute difference between the average gray-level value of the candidate region and the background gray-level value, balancing the influence of region size and the gray-level difference between the entire image and the background on the result. The higher the gray-level density value obtained, the more concentrated and spatially compact the gray-level changes in the region are, and the higher the target recognition potential. Conversely, it indicates that the gray-level distribution is sparse or there is no significant concentrated change, which is the background region. This calculation logic takes into account both local feature intensity and spatial structure, and is suitable for the rapid identification and sorting of target regions in complex images. Gray density value represents the spatial concentration of pixel gray level changes within a certain area of ​​an image. The higher the value, the more significant and concentrated the gray level differences are, indicating that the area contains target features. Conversely, a lower value indicates that the gray level changes in the area are scattered or have no significant differences, and is a background area. Parameter meaning: : indicates the first The average grayscale value of pixels within the region is calculated as follows: ; : Represents the average gray level of pixels in the non-target area of ​​the image, which is calculated by traversing and averaging the regions in the image that clearly do not contain the target. The innovation of this formula lies in the fact that by introducing a coupling mechanism between pixel gray-level difference and spatial coordinate offset, it effectively evaluates the density distribution of gray-level changes in local areas of the image. At the same time, by normalizing the overall average gray-level difference, it improves the feature discrimination ability between different image regions, which is beneficial for the extraction of target regions and anomaly detection. Actual parameter acquisition and quantization process: The parameters are obtained as follows, and the calculation process is illustrated with examples: Region selection: Image preprocessing software (such as OpenCV) is used to segment candidate regions and non-target regions; grayscale value : Obtained by traversing the image pixel matrix, such as the first pixel in the image. The first in the region The grayscale value of each pixel is 134; coordinates The index position of a pixel in the image is determined by its position. For example, if a pixel is in row 35 and column 50 of the image, then... ; Average gray value Obtained by iterating through grayscale values ​​and averaging the results; Average gray level in non-gravel areas : Select a background area that does not contain gravel targets (such as sky background, open ground, etc.) to collect data and then calculate the average; Data sampling and calculation examples: Set candidate region number The region contains pixels The specific data collected is as follows: Table 2: Pixel Data Table for Candidate Regions

[0032] As shown in Table 2, the first Each candidate region contains 4 pixels, which record the grayscale value and coordinate position respectively; Calculate according to Table 2: Calculate the average gray value of the region : ; Calculate the average of the x and y coordinates : ; ; Average gray level in non-gravel areas (Calculated by averaging the grayscale values ​​of 1000 pixels sampled from non-target areas). Substitute into the numerator of the formula (for each) (Calculations to be performed) : ; : ; : ; : ; Summation of numerators: ; Denominator calculation: ; Denominator: ; Substitute into the formula to calculate: ; The result shows that the grayscale density value is 0.1864, indicating that the first... The spatially dispersed grayscale variations in the region indicate that it lacks significant abrasive characteristics. If this value is close to or below an empirical threshold (e.g., 0.2), the region can be preliminarily identified as a non-target region. In subsequent steps, multiple regions can be considered together. Sorting by value, the target area of ​​the gravel with concentrated gray-scale changes is selected; Set threshold It is obtained after training based on a batch of image samples. The specific setup process is as follows: Extracting regions containing and without gravel from 30 images value; The statistical mean of the gravel area is The mean value of the non-gravel area is ; Taking into account both standard deviation and discriminative power, the threshold is set as follows: ; Due to the results of the current example Therefore, it is classified as a non-target area.

[0033] S402: Call the displacement trajectory and direction change of the gray-level principal vector in the trend group of gray-level principal vector of non-target area, combine it with the coordinate segment occupied by the boundary of the suspended particle area, perform cross mapping on the overlapping boundary position of the two on the image plane, mark the area where the gray-level principal vector extension trajectory intersects with the contour line formed by the suspended particle boundary as the interference overlapping segment, record the frame number and area position index of the overlapping segment, and obtain the edge brightness correction segment mapping map; The displacement trajectory and direction change data of each principal vector in consecutive frames are extracted and compared with the boundary coordinate segments of the identified suspended particle regions in the same frame. By establishing a cross-mapping relationship in the image plane coordinate system, it is determined whether the grayscale principal vector extension path coincides with the boundary contour of the suspended particles. The principal vector path is projected onto the image plane in a pixel-level step manner, and it is checked whether the coordinates of each step fall within the area surrounded by the suspended particle contour. If overlap occurs, the frame number and intersection pixel position of the path segment are recorded and marked as an interference overlap segment. The intersection points of the path and the contour region are marked, and their position indices (such as the pixel coordinates of the upper left and lower right corners of the rectangle) and the corresponding frame numbers are saved to form a set of interference overlap segment data. The overlap segments in this set are mapped onto the image plane, highlighting the area where the main vector coincides with the target contour. The starting point of the grayscale main vector in frame 22 is set to (240, 180), and the direction is (260, 190). The path passes through the boundary point of the suspended particles (250, 185). This segment is defined as interference overlap and marked as the rectangular coverage area in the image. The edge brightness correction segment mapping map is obtained.

[0034] Please see Figure 6 The specific steps for obtaining the label set of the stable structure along the main axis are as follows: S501: Based on the edge contour region of the suspended particles in the edge brightness correction section mapping map, extract the set of pixel coordinates of the edge of the suspended particles in each frame image in sequence, perform circumscribed ellipse fitting on the set of pixel coordinates, obtain the orientation angle value of the major axis direction of the circumscribed ellipse in the image plane in the image frame, and arrange the principal axis orientation angles corresponding to the image frames in sequence to generate the image frame principal axis direction angle sequence. After calling the edge contour region of the suspended particles determined in the edge brightness correction segment mapping map, each frame of the image is processed sequentially. By accurately extracting the set of boundary pixel coordinates of the contour region, the contour point set of the corresponding frame is constructed, and its original positional relationship is preserved in the image plane coordinate system. The circumscribed ellipse fitting operation is performed on each point set, and the geometric shape approximation technique is used to fit the shape model of the boundary so that the shape of the contour can be described by a unique closed ellipse. After the fitting is completed, the principal axis direction of the ellipse in the image coordinate space is obtained. This direction can be obtained by analyzing the angle formed between the major axis of the fitted ellipse and the x and y coordinates. This angle is defined as the principal axis direction angle, which represents the spatial arrangement trend of the suspended particle structure in the image frame. After all image frames are processed in the above process, each image frame obtains an angle value representing the orientation of the principal axis of its suspended particle structure. The angle values ​​are arranged in the order of the image frames to generate the image frame principal axis direction angle sequence.

[0035] S502: Call the main axis orientation angle values ​​of consecutive frames in the image frame main axis direction angle sequence, combine the number of overlapping pixels between the pixel sets of the suspended particle structure boundary contour in adjacent frames, synchronously compare the range of change of the number of overlapping pixels with the range of difference of the main axis angle, filter out the frame segments with abrupt changes in the main axis angle and deviations in the number of overlapping pixels, retain the frame segments with gentle angle changes and continuous distribution of boundary overlap points, and generate a set of continuous frame segments in the structural direction. To determine whether there are abrupt changes in the comparison process, the angle change amplitude between adjacent image frames is calculated frame by frame to determine whether there are jumps that do not conform to the characteristics of morphological stability. At the same time, the pixel sets of the boundary contours of suspended particles in adjacent image frames are cross-compared to count the number of overlapping pixels between them and analyze the continuity of the boundary. By comparing the trend of the change in the number of intersection pixels of the contours with the trend of the change in the principal axis angle, the two changes are compared in a linked manner. When there is a sudden increase in angle change accompanied by a decrease in the number of boundary overlap points, it is marked as a structurally discontinuous frame segment and is removed. Conversely, when the angle change process is gradual and the intersection of the boundary contours remains stable, the image frame segment is retained. The entire image sequence is screened frame by frame to form a set of frame segments containing only stable structural changes, coherent boundary contours, and traceable spatial morphology, thus generating a set of continuous frame segments in the structural direction.

[0036] S503: Call the boundary region of the suspended particle structure in the set of continuous frame segments in the structural direction, mark the frame segments that maintain the continuous trend of the change of the main axis orientation angle and whose contour distribution is linearly progressive in the image space, extract the frame segment number and the index number of the covered suspended particle structure position, and obtain the main axis direction stable structure label set; For each frame sequence, spatial distribution directionality is identified. For the outline boundaries of suspended particles in continuous images within the frame segment, the overall extension direction of the circumscribed ellipse principal axis is analyzed to determine whether it exhibits a linear progression relationship within the image space. By observing the extension trend of the major axis of the fitted ellipse in each frame, its principal axis extension is plotted on the image plane. The principal axis extension of the frame is then fitted into a spatial trajectory. If a linear trend is observed, it indicates that the suspended particle structure in that segment is continuously advancing in a certain direction. Simultaneously, the frame number of the linear trend segment is extracted, and the image pixel position index corresponding to the outline structure in that frame segment is obtained. These are then classified as structural labels with stable progression relationships. In the output set of stable structural labels in the principal axis direction, each item corresponds to an image frame segment that is visually continuous, morphologically coherent, and has a clearly defined principal axis extension. This set is used for subsequent identification processing and data modeling analysis of high-stability regions, thus obtaining the set of stable structural labels in the principal axis direction.

[0037] The method is applied to the flocculation and sedimentation process or the activated sludge biochemical treatment process in wastewater treatment to monitor and evaluate the morphology, trajectory, and dynamic evolution of flocs and activated sludge, so as to optimize the treatment process parameters.

[0038] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for assessing the morphological evolution of suspended particles in wastewater based on machine vision and temporal modeling, characterized in that, Includes the following steps: S1: Acquire image frames of suspended particles in the wastewater system, extract the set of edge pixel coordinates of suspended particles in each image frame, and sequentially determine the area of ​​the region formed by the boundary of suspended particles in the image, the length of the boundary line segment and the center point of the region enclosed by the edge contour to obtain the set of morphologically stable image segment identifiers. S2: Call the morphologically stable image segment identifier set, extract the set of floating particle edge pixels of the first and last image frames of the morphologically stable image segment, perform spatial transfer comparison based on the geometric enclosing area in the image plane, and obtain the contour coverage evolution map. S3: Call the outline coverage evolution map, identify the number of pixels in the suspended particle region and locate the coordinates of the centroid of the gray-scale distribution, construct the centroid path map according to the frame sequence, record the changes in the number of pixels, and obtain continuous records of the centroid path. S4: Call the continuous recording of the center of gravity path, extract the position of the gray concentration point and the gradient direction vector of the non-target background area, analyze the spatial displacement trend, and compare the intersection with the boundary position of the suspended particle area to mark the image segment affected by the illumination shift and obtain the edge brightness correction section mapping map.

2. The method for identifying and analyzing the physical characteristics of suspended particles in wastewater based on image recognition according to claim 1, characterized in that, The morphologically stable image segment identifier set includes the region morphologically stable frame sequence number, edge contour stability feature value, standard deviation of center position change, area change range evaluation result, and intra-frame temporal consistency label. The contour coverage evolution map includes the contour bounding boundary evolution curve, spatial bounding overlap heat map, suspended particle region contour expansion direction distribution map, and first and last edge overlap coefficient map. The centroid path continuous record includes the pixel centroid coordinate sequence, gray-level centroid offset path map, inter-frame centroid offset speed curve, and suspended particle pixel number change trend map. The edge brightness correction segment mapping map includes the gray-level offset concentrated region distribution map, background region illumination gradient direction map, edge recognition deviation mapping map, and image brightness error correction label set.

3. The method for identifying and analyzing the physical characteristics of suspended particles in wastewater based on image recognition according to claim 1, characterized in that, The specific steps for obtaining the morphologically stable image segment identifier set are as follows: S101: Obtain image frames of suspended particles in the wastewater system, extract the contour line pixels formed by the boundary of the suspended particle region in each frame of the image in turn, call the set of contour pixel coordinates corresponding to the boundary of the suspended particles, summarize the number of pixels in the region formed by the contour line, count the number of bus segments of the contour line, calculate the centroid coordinates of the bounding box pixels of the region formed, and obtain the image frame boundary feature data. S102: Based on the image frame boundary feature data, calculate the horizontal and vertical offset distances of the centroid coordinates between adjacent frames sequentially according to time, make a continuous judgment on the trend of offset distance change, and filter out image frame segments with continuous increasing or decreasing directions to obtain a set of boundary stable frame segment data. S103: Call the corresponding image frame sequence in the boundary stable frame segment data set, and compare the line segment direction, closure degree and line segment distribution width of the suspended particle outline boundary in each image frame. Select image frames with consistent line segment direction, similar closure level and stable width variation to form an image segment. Perform unified classification processing according to the structural characteristics of the image segment to obtain a set of morphologically stable image segment identifiers.

4. The method for identifying and analyzing the physical characteristics of suspended particles in wastewater based on image recognition according to claim 3, characterized in that, The specific steps for obtaining the contour coverage evolution map are as follows: S201: Call each image frame sequence marked by the morphologically stable image segment identifier set, extract the contour pixel coordinate set of the suspended particle edge in the first frame and the last frame, calculate the contour mapping difference value, map the contour pixel set to the corresponding image plane according to the original resolution of the image frame, construct the closed edge graphics of the suspended particle region in the two frames respectively, merge the pixel regions of the plane range occupied by the closed edge graphics, and generate the closed region dataset of the first and last contour images. S202: Call the coordinates of the edge regions of the first and last frames in the closed region dataset of the first and last contour images, perform the geometric enclosure relationship determination of the two sets of closed regions in the image plane, analyze the expansion direction and movement trend of the suspended particle region in the image plane based on the boundary expansion position of the circumscribed rectangle and the coverage ratio of the overlapping boundary, and draw the boundary expansion change form of the differentiated image segments by overlapping the line trajectory to obtain the contour coverage evolution map.

5. The method for identifying and analyzing the physical characteristics of suspended particles in wastewater based on image recognition according to claim 4, characterized in that, The contour mapping difference value measures the spatial difference between the position of a contour point after it is mapped to the image plane at the original image resolution and the position of the global average contour.

6. The method for identifying and analyzing the physical characteristics of suspended particles in wastewater based on image recognition according to claim 4, characterized in that, The specific steps for obtaining the continuous records of the centroid path are as follows: S301: Call the suspended particle boundary contour region of the contour coverage evolution map, identify the pixel position in the contour region in each frame image, and count the total number of pixels in the contour region. At the same time, extract the set of gray values ​​of the corresponding pixels in the original gray matrix of the image, and perform a weighted average calculation by combining the pixel coordinates and gray values ​​to obtain the gray distribution centroid coordinates of the suspended particle region in each frame image, and generate the set of suspended particle centroid and pixel count of the image frame. S302: Based on the centroid coordinates of the image frames in the set of centroids of suspended particles and the number of pixels in the image frames, the centroids are connected sequentially in the order of the image frame sequence to form a connecting line path. The points on the connecting line path are used as reference points to record the position offset distance between adjacent frames. The number of pixels in the corresponding frame is marked as auxiliary information to obtain the centroid path image data of the suspended particle region. S303: Call the start and end positions of the path segments and the corresponding frame numbers in the image data of the center of gravity path of the suspended particle region, extract the direction of change of the center of gravity position and the magnitude of the distance change between consecutive frames, arrange the number of pixels marked on multiple points on the path in sequence, and synchronously record the number of pixels in each path segment to obtain a continuous record of the center of gravity path.

7. The method for identifying and analyzing the physical characteristics of suspended particles in wastewater based on image recognition according to claim 6, characterized in that, The specific steps for obtaining the edge brightness correction section mapping map are as follows: S401: Call the continuous recording of the center path, exclude the set of pixel coordinates formed by the boundary of the suspended particle region in the corresponding frame image, extract the concentrated points of pixels with dense gray value changes in the remaining image region, identify the position gradient of the gray value difference direction of the pixels adjacent to the concentrated points, calculate the gray density value, determine the continuous displacement direction of the gray concentration region on the image plane according to the arrangement order of the main vector position of the gradient change direction, and generate the gray main vector trend group of non-target region. S402: Call the displacement trajectory and direction change of the gray-level principal vector of the continuous frames in the gray-level principal vector trend group of the non-target area, and combine it with the coordinate segment occupied by the boundary of the suspended particle area to perform cross mapping on the overlapping boundary position of the two on the image plane. Mark the area where the gray-level principal vector extension trajectory intersects with the contour line formed by the boundary of the suspended particle as the interference overlapping segment, record the frame number and area position index of the overlapping segment, and obtain the edge brightness correction segment mapping map; the gray-level density value represents the degree of spatial concentration of pixel gray-level changes in the image area.

8. The method for identifying and analyzing the physical characteristics of suspended particles in wastewater based on image recognition according to claim 1, characterized in that, The method further includes step S5: S5: Based on the outline of the suspended particles corresponding to the edge brightness correction section mapping map, extract the orientation angle of the main axis direction in consecutive frames, combine the continuity of the main axis direction in adjacent frames with the degree of overlap of the boundary points within the structure, determine whether the structure is maintained in a unified geometric advancement trajectory, mark the suspended particle structure with stable orientation advancement characteristics, and obtain the main axis direction stable structure label set. The main axis orientation stable structure label set includes main axis orientation stability index, main axis evolution trajectory curve, structural boundary continuity distribution value, and directional propulsion consistency identification label.

9. The method for identifying and analyzing the physical characteristics of suspended particles in wastewater based on image recognition according to claim 8, characterized in that, The specific steps for obtaining the label set of the axial direction stabilizing structure are as follows: S501: Based on the edge contour region of the suspended particles in the edge brightness correction section mapping map, extract the set of pixel coordinates of the edge of the suspended particles in each frame image in sequence, perform circumscribed ellipse fitting on the set of pixel coordinates, obtain the orientation angle value of the major axis direction of the circumscribed ellipse in the image plane, and arrange the main axis orientation angles corresponding to the image frames in sequence to generate the image frame main axis direction angle sequence. S502: Call the main axis orientation angle values ​​of consecutive frames in the image frame main axis orientation angle sequence, combine the number of overlapping pixels between the pixel sets of the suspended particle structure boundary contour in adjacent frames, synchronously compare the range of change of the number of overlapping pixels with the range of difference of the main axis angle, filter out the frame segments with abrupt changes in the main axis angle and deviations in the number of overlapping pixels, retain the frame segments with gentle angle changes and continuous distribution of boundary overlapping points, and generate a set of continuous frame segments in the structural orientation. S503: Call the boundary region of the suspended particle structure in the set of continuous frame segments in the structural direction, mark the frame segments that maintain the continuous trend of the change of the main axis orientation angle and whose contour distribution is linearly progressive in the image space, extract the frame segment number and the index number of the covered suspended particle structure position, and obtain the main axis direction stable structure label set.

10. The wastewater suspended particle morphology evolution assessment method based on machine vision and temporal modeling according to claim 1, characterized in that, The method is applied to the flocculation and sedimentation process or the activated sludge biochemical treatment process in wastewater treatment to monitor and evaluate the morphology, trajectory, and dynamic evolution of flocs and activated sludge, so as to optimize the treatment process parameters.

Images