Concrete mixing host visual monitoring system and method thereof

Abstract

The application discloses a concrete mixing main machine visual monitoring system and method, and belongs to the technical field of concrete production automation and intelligent quality control, which comprises the following steps: S1, collecting a video image sequence of concrete movement in a mixing main machine in real time; S2, extracting a dynamic visual feature vector and a static feature vector based on the video image sequence; S3, calculating an internal rheological state index based on the dynamic visual feature vector and a preset dynamic rheological-visual feature coupling model; S4, calculating a basic slump prediction value based on the static feature vector and a preset static correlation model; S5, determining a final corrected slump by a preset correction function in combination with the internal rheological state index and the basic slump prediction value; and S6, calculating a comprehensive risk index according to the final corrected slump, the basic slump prediction value and the internal rheological state index. The application can deeply capture the internal rheological characteristics caused by chemical admixtures, which cannot be perceived by traditional visual methods.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of concrete production automation and intelligent quality control, in particular to a concrete mixing main machine visual monitoring system and method thereof. BACKGROUND

[0002] In the quality control of concrete production, the traditional mixing process monitoring method mainly relies on manual experience or conventional visual systems, which indirectly judge the working performance of concrete by observing its macroscopic apparent fluidity. However, these methods face fundamental problems when dealing with modern high-performance concrete. The widespread use of chemical admixtures causes concrete to exhibit complex non-Newtonian fluid characteristics, which seriously decouples its visual representation from its true internal working performance.

[0003] This situation leads to the fact that producers cannot accurately and timely grasp the true state of concrete, making it difficult to discover and diagnose internal rheological performance abnormalities caused by admixtures, affecting the precise regulation of the production process. In addition, the traditional method lacks the ability to quantitatively evaluate this risk of discrepancy between appearance and reality, and cannot provide scientific decision-making basis for preventing construction problems.

[0004] The above situation and deficiencies are mainly due to the limitations of monitoring and analysis technology. The analysis method that simply relies on macroscopic visual features is not sensitive enough to the microscopic structural changes and rheological property changes caused by chemical components, resulting in a single dimension of information acquisition. As a result, when concrete has potential quality risks, managers cannot quickly obtain comprehensive and accurate quantitative information, delaying the adjustment opportunity, and ultimately may lead to serious engineering accidents such as pumping pipe blockage and segregation.

[0005] The above information disclosed in the background section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0006] The purpose of the present application is to provide a concrete mixing main machine visual monitoring system and method thereof to solve the problems raised in the background art.

[0007] The technical solution of the present application is as follows:

[0008] S1, real-time acquisition of video image sequences of concrete moving in the mixing main machine;

[0009] S2, based on the video image sequence, extracting dynamic visual feature vectors and static feature vectors;

[0010] S3, based on the dynamic visual feature vector and the preset dynamic rheological-visual feature coupling model, calculating the internal rheological state index;

[0011] S4, based on the static feature vector and a preset static correlation model, a basic slump prediction value is calculated;

[0012] S5, in combination with the intrinsic rheological state index and the basic slump prediction value, a final corrected slump is determined through a preset correction function;

[0013] S6, according to the final corrected slump, the basic slump prediction value and the intrinsic rheological state index, a comprehensive risk index is calculated;

[0014] S7, based on the comprehensive risk index and a preset risk threshold, a control instruction is output.

[0015] Preferably, the dynamic visual feature vector comprises: global shear intensity and structural entropy rate.

[0016] Preferably, the determination of the global shear intensity comprises:

[0017] calculating the optical flow field between consecutive image frames in the video image sequence;

[0018] based on the optical flow field, calculating the mean value of the spatial gradient norm of the optical flow velocity field as the global shear intensity.

[0019] Preferably, the determination of the structural entropy rate comprises:

[0020] calculating the gray level co-occurrence matrix of a single frame image in the video image sequence to obtain the Shannon entropy;

[0021] based on the Shannon entropy, calculating the first derivative in the time dimension as the structural entropy rate.

[0022] Preferably, the static feature vector comprises: surface flatness index and average optical flow velocity amplitude.

[0023] Preferably, the intrinsic rheological state index is calculated by linear combination of the global shear intensity and the structural entropy rate.

[0024] Preferably, the output control instruction comprises:

[0025] if the comprehensive risk index is higher than the risk threshold, triggering a high-risk alarm and generating a preventive control instruction;

[0026] if the comprehensive risk index is not higher than the risk threshold, no preventive control instruction is generated.

[0027] A concrete mixing main machine visual monitoring system comprises:

[0028] an image sequence acquisition unit for acquiring a video image sequence of concrete moving in the mixing main machine in real time;

[0029] The feature extraction unit is configured to extract a dynamic visual feature vector and a static feature vector based on the video image sequence;

[0030] The state recognition unit is configured to calculate an intrinsic rheological state index based on the dynamic visual feature vector and a preset dynamic flow-rheological visual feature coupling model;

[0031] The basic prediction unit is configured to calculate a basic slump prediction value based on the static feature vector and a preset static correlation model;

[0032] The correction calculation unit is configured to determine a final corrected slump by a preset correction function in combination with the intrinsic rheological state index and the basic slump prediction value;

[0033] The risk assessment unit is configured to calculate a comprehensive risk index according to the final corrected slump, the basic slump prediction value and the intrinsic rheological state index;

[0034] The instruction generation unit is configured to output a control instruction based on the comprehensive risk index and a preset risk threshold.

[0035] The present application provides a concrete mixing host visual monitoring system and method, which has the following improvements and advantages compared with the prior art.

[0036] 1. The present application innovatively extracts two types of visual feature vectors in parallel, namely dynamic and static visual feature vectors. The static feature vector includes a surface flatness index and an average optical flow velocity amplitude, which are used to quantify the traditional macroscopic flow form that can be perceived by the naked eye, and a basic slump prediction value is calculated based on the static feature vector. The prediction value constitutes an initial baseline evaluation of the workability of the concrete. The present application introduces a dynamic visual feature vector, which includes a global shear strength and a structural entropy rate. The global shear strength quantifies the shear deformation rate of the material as a whole from a macroscopic perspective by analyzing the spatial gradient of the optical flow velocity field. The structural entropy rate quantifies the degree of reorganization of the internal component structure of the material from a microscopic perspective by calculating the time derivative of the image Shannon entropy. The synergistic combination of these two features can deeply capture the intrinsic rheological characteristics triggered by chemical admixtures, which cannot be perceived by traditional visual methods.

[0037] 2. The present application can calculate an intrinsic rheological state index. The index quantifies the complex rheological behavior as a physical quantity that can be indirectly measured. The essence of the present application is to use the intrinsic rheological state index to nonlinearly correct the aforementioned basic slump prediction value through a preset correction function. This correction mechanism only applies effective adjustment when the intrinsic rheological state of the concrete significantly deviates from the normal baseline, thereby achieving accurate compensation of the basic prediction deviation and obtaining a highly accurate corrected slump.

[0038] 4.The application is not simply state prediction, a set of quantitative risk assessment and control mechanism is established;Through an algorithm that integrates the degree of abnormality of rheological state and the deviation of visual prediction, a comprehensive risk index is calculated;The index converts the vague construction risk into an objective and quantifiable evaluation index;When the index exceeds the preset risk threshold, the system can automatically trigger a high-risk alarm and generate preventive control instructions;This active intervention capability turns the monitoring system from a passive state observer to an active risk manager, which can effectively prevent serious engineering accidents such as pumping pipe blockage and segregation bleeding caused by abnormal concrete performance, greatly improving the intelligent level of concrete production and the reliability of quality control. BRIEF DESCRIPTION OF DRAWINGS

[0039] The application will be further explained in conjunction with the accompanying drawings and embodiments:

[0040] Figure 1 is a flow chart of a concrete mixing main machine visual monitoring method of the application. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical scheme and advantages of the application more clear and explicit, the application will be further described in detail in conjunction with specific embodiments. Embodiment 1

[0042] Please refer to Figure 1 , the application provides a concrete mixing main machine visual monitoring method, comprising:

[0043] S1, real-time acquisition of video image sequence of concrete movement in the mixing main machine;

[0044] S2, based on the video image sequence, extracting dynamic visual feature vector and static feature vector;

[0045] S3, based on the dynamic visual feature vector and the preset dynamic rheological-visual feature coupling model, calculating the internal rheological state index;

[0046] S4, based on the static feature vector and the preset static correlation model, calculating the basic slump prediction value;

[0047] S5, combining the internal rheological state index and the basic slump prediction value, determining the final corrected slump through the preset correction function;

[0048] S6, according to the final corrected slump, the basic slump prediction value and the internal rheological state index, calculating the comprehensive risk index;

[0049] S7, based on the comprehensive risk index and the preset risk threshold, outputting the control instruction;

[0050] The embodiment provides a complete method closed loop of concrete mixing host visual monitoring;

[0051] In step S1, an image sequence acquisition unit is arranged in the interior or opening of the concrete mixing host, so as to continuously capture the dynamic behavior of the concrete in the mixing process; in the embodiment, the unit continuously records a video by an industrial high-speed camera at a preset frame rate, to generate a real-time video image sequence; the sequence provides original data input for all subsequent analyses;

[0052] In step S2, a feature extraction unit is configured to process the video image sequence acquired in the previous step; the unit aims to extract low-dimensional and quantized features representing different physical properties of the concrete from high-dimensional pixel data; in the embodiment, the unit extracts two types of feature vectors in parallel:

[0053] Dynamic visual feature vector: the vector aims to quantify the microscopic structure changes and macroscopic shear behavior of the concrete paste, especially the strong non-Newtonian fluid effects introduced by chemical admixtures, which are difficult to distinguish by conventional visual methods;

[0054] Static feature vector: the vector aims to quantify the macroscopic fluidity and surface morphology of the concrete, which are usually the main basis for judging its workability in human experience;

[0055] In step S3, a state recognition unit receives the dynamic visual feature vector and performs calculation according to a preset dynamic rheology-visual feature coupling model; the model aims to deconstruct the correlation between visual appearance and internal physical state, establish a mapping relationship between dynamic visual features and the internal rheological state of the concrete, and thus diagnose the real physical state; in the embodiment, the model is used to solve an internal rheological state index; the internal rheological state index refers to a scalar for quantifying the activity or rate of the internal rheological behavior of the concrete, and functions to provide an indirectly measurable index for the missing rheological state dimension of the traditional visual model, which needs to be highly sensitive to the thixotropy and shear thinning effects caused by chemical admixtures; the preset dynamic rheology-visual feature coupling model is established based on a series of offline calibration experiments: by simultaneously performing laboratory rheometer tests and video acquisition in the target mixer on multiple groups of concrete samples containing different admixture contents, the specific correlation between the visual feature calculation value and the rheometer measured value, such as the shear thinning index, is maximized as the optimization goal, and the model parameters are solved and fixed by regression analysis method;

[0056] In step S4, the basic prediction unit receives the static feature vector and performs calculation according to a preset static correlation model; the purpose of the model is to simulate the traditional artificial experience to establish the basic correlation between the visual apparent fluidity and the physical workability of the concrete, which is the slump in this embodiment; in this embodiment, the model is used to calculate the basic slump prediction value; the basic slump prediction value refers to the slump value predicted only according to the visual macroscopic fluidity without considering the complex rheological effect, and the function is to provide the initial and uncorrected benchmark prediction; the preset static correlation model is calibrated by regression analysis of the static feature values calculated from the video and the slump values measured in the physical experiments on a large number of benchmark concrete samples without high-efficiency admixtures;

[0057] In step S5, the correction calculation unit collects the intrinsic rheological state index and the basic slump prediction value obtained in the previous steps; the unit performs operation through a preset correction function, and the purpose is to correct the deviation of the basic slump prediction value by using the non-Newtonian fluid effect quantified by the intrinsic rheological state index and ignored by the basic prediction model; in this embodiment, the function calculates the final corrected slump; the corrected slump refers to the final prediction value of the real workability of the concrete considering the visual apparent fluidity and the intrinsic rheological state; the structure and parameters of the preset correction function are carefully designed to ensure that the correction effect is significant only when the intrinsic rheological state deviates significantly from the normal benchmark, so as to realize accurate compensation;

[0058] In step S6, the risk assessment unit receives the final corrected slump, the basic slump prediction value and the intrinsic rheological state index; the purpose of the unit is to quantify the potential construction risk caused by the disconnection between the visual representation and the real state; in this embodiment, a comprehensive risk index is calculated by a specific algorithm; the comprehensive risk index refers to a dimensionless index, and the function is to comprehensively evaluate the abnormal degree of the rheological state and the visual prediction deviation degree caused thereby, and the higher the value is, the greater the potential risk is;

[0059] In step S7, the instruction generation unit performs decision judgment on the comprehensive risk index calculated in the previous steps; the purpose of the unit is to convert the quantitative risk assessment into specific and executable production intervention instructions; in this embodiment, the index is compared with a preset risk threshold value; the source of the preset risk threshold value is the distribution of the comprehensive risk index corresponding to the concrete batches that have appeared problems such as pump pipe blockage and segregation bleeding in the statistical analysis of history, so as to determine the critical value with statistical significance; based on the comparison result, the system outputs the corresponding control instructions;

[0060] The method provided by the present application extracts dynamic and static visual features in parallel, and is respectively used for diagnosing internal rheological state and predicting basic workability, and the former is used for correcting the latter, so that a complete diagnosis-prediction-correction-risk control technical closed loop is formed; the method solves the fundamental problem that the visual monitoring in the prior art is seriously decoupled from the real workability due to the introduction of chemical admixtures, realizes quantitative diagnosis of the internal rheological state of concrete and accurate prediction of the workability index, and can automatically generate a preventive control instruction based on the quantitative risk assessment, thereby significantly improving the intelligent level of concrete production and the reliability of quality control.

[0061] The dynamic visual feature vector comprises: global shear strength and structural entropy rate.

[0062] In the embodiment, the composition of the dynamic visual feature vector is limited; the vector at least contains two core components: global shear strength and structural entropy rate.

[0063] The global shear strength is defined as a scalar obtained by calculating the optical flow field between continuous image frames and calculating the spatial gradient norm of the optical flow velocity field; the index aims to quantize the average shear deformation rate of the concrete material under the driving of the stirring blade from the macroscopic level; higher global shear strength usually means that the overall macroscopic fluidity of the material is stronger, or is experiencing a severe shear thinning process.

[0064] The structural entropy rate is defined as a scalar obtained by first calculating the gray level co-occurrence matrix of a single image to obtain the Shannon entropy representing the texture complexity, and then calculating the first-order derivative of the Shannon entropy of continuous frames in the time dimension; the index aims to quantize the degree of reorganization of the relative position relationship between the aggregates, mortar and other components inside the concrete from the microscopic level; higher structural entropy rate usually means that the internal structure of the material is accelerating deconstruction and reorganization, which is highly related to the rheological behavior such as thixotropy.

[0065] By combining the macroscopic global shear strength and the microscopic structural entropy rate, the dynamic visual feature vector of the embodiment can more comprehensively and profoundly capture the complex rheological phenomena caused by chemical admixtures; the global shear strength focuses on the macroscopic result of the shear rate, while the structural entropy rate focuses on the microscopic process of structure reorganization, and the two complement each other, providing more rich-dimensional and more reliable information input for subsequent accurate identification of the internal rheological state, thereby improving the accuracy of the index diagnosis of the internal rheological state.

[0066] The determination of the global shear strength comprises:

[0067] The optical flow field between continuous image frames in the video image sequence is calculated.

[0068] Based on the optical flow field, the mean value of the spatial gradient norm of the optical flow velocity field is obtained as the global shear strength.

[0069] This embodiment details the steps for calculating the global shear strength;

[0070] The calculation process includes: calculating the optical flow field between consecutive image frames in a video image sequence; the optical flow field refers to a two-dimensional vector field, where each vector represents a point on the image between two consecutive frames, i.e., an extremely short time interval. The motion displacement within the image area; in this embodiment, a mature optical flow estimation algorithm is used to process every two consecutive image frames in the acquired video sequence to obtain a velocity vector field covering the entire image area. ;

[0071] Based on the aforementioned optical flow field, the mean value of the spatial gradient norm of the optical flow velocity field is calculated; the spatial gradient of the optical flow velocity field refers to the velocity vector field. In spatial coordinates The rate of change of the velocity field reflects the velocity difference between adjacent material points and is directly related to the intensity of shear deformation; in this embodiment, the obtained velocity field... Calculate its Jacobian matrix, and further calculate the norm of this matrix; the norm value represents the shear strength at each point. Then, spatially average the shear strength values ​​of all pixels in the image, and define this average as the global shear strength. ;

[0072] This method for determining global shear strength directly maps the concept of shear rate in the physical world onto calculable image features through optical flow analysis. Instead of simply measuring average velocity, it accurately captures the relative deformation intensity caused by velocity differences within the material by calculating the spatial gradient of the velocity field. This method is more sensitive and direct to rheological effects such as shear thinning, thus providing macroscopic features with clearer physical meaning and more precise quantification for identifying the intrinsic rheological state.

[0073] Determining the rate of change of structural entropy includes:

[0074] Calculate the gray-level co-occurrence matrix of a single frame in a video image sequence to obtain the Shannon entropy;

[0075] Based on Shannon entropy, the first derivative is obtained in the time dimension to serve as the rate of change of structural entropy;

[0076] This embodiment details the steps for calculating the structural entropy change rate;

[0077] The calculation process includes: calculating the gray-level co-occurrence matrix (GLCM) of a single frame in the video image sequence to obtain the Shannon entropy; the GLCM is a matrix that describes the distribution of pixel pairs with specific gray levels in an image and is a classic method for analyzing image texture; in this embodiment, the GLCM is calculated for each frame in the video sequence; subsequently, the Shannon entropy is calculated based on this matrix. Shannon entropy Here, it refers to a scalar used to quantify the complexity and disorder of image textures, which is used to evaluate the randomness of the visual texture of concrete surfaces formed by aggregates and mortar; high entropy values ​​usually correspond to a rough, complex textured surface.

[0078] Based on the Shannon entropy mentioned above, the first derivative is calculated in the time dimension; since the instantaneous Shannon entropy has already been calculated for each frame of the video sequence... This yields a time series of entropy values ​​with respect to time. In this embodiment, the first derivative of this time series with respect to time is approximated using numerical methods such as the finite difference method. The result is defined as the structural entropy rate of change. ;

[0079] This method for determining the rate of change of structural entropy cleverly utilizes entropy from information theory to quantify the visual representation of concrete microstructures. By further calculating the rate of change of entropy over time, this method can capture the dynamic process of a structure moving from order to disorder, or vice versa, rather than a static snapshot. This is particularly effective for characterizing rheological properties such as thixotropy, which are directly related to the rate of structural reorganization, and provides a micro-dynamic perspective that traditional visual methods cannot reach for identifying the intrinsic rheological state. Example 2

[0080] Static feature vectors include: surface smoothness index and average optical flow velocity amplitude;

[0081] In this embodiment, the static feature vector The composition of the vector is defined; the vector contains at least two core components: a surface smoothness index and an average optical flow velocity amplitude.

[0082] Surface flatness index It is defined as a dimensionless scalar quantity used to quantify the degree to which the surface of concrete tends to flatten during the mixing process; its purpose is to capture the ability of concrete to resist plastic deformation under gravity, which is closely related to yield stress; in this embodiment, It can be calculated by analyzing the edge gradient distribution or frequency domain energy distribution of the image;

[0083] Calculation method based on image edge gradient statistics

[0084] The physical intuition of this method is that the more uneven and rough the concrete surface is, the richer and sharper the texture edges it presents in the image are, and thus the higher the overall gradient strength is.

[0085] For a single gray-scale image Apply a standard edge detection operator, such as Sobel or Canny operator, to calculate the gradient magnitude of each pixel in the image, and obtain the gradient map ;

[0086] Calculate the spatial average of the gradient magnitudes of all pixels in the entire image, or in a pre-defined region of interest (ROI) The calculation formula is:

[0087]

[0088] wherein, is the total number of pixels involved in the calculation; is the gradient magnitude of the i-th pixel in the image; is the spatial coordinate of the pixel;

[0089] Since the surface flatness index should be positively correlated with the flatness, while the average gradient should be negatively correlated with the flatness, a reverse function relationship can be used to construct ; For example, the following formula can be used for calculation:

[0090]

[0091] wherein, is a normal number calibrated through experiments, and its dimension is the inverse of the dimension of the average gradient , so as to ensure that is a dimensionless quantity; this formula ensures that when the surface tends to be absolutely flat , ; when the surface is rougher , the value is lower;

[0092] The surface of concrete with poor fluidity and high plasticity will be rougher and more uneven, corresponding to a lower value; before calculation, the index will be normalized;

[0093] The average optical flow velocity magnitude is a scalar obtained by spatially averaging the magnitudes of all velocity vectors in the optical flow field calculated above; the purpose is to macroscopically evaluate the overall movement speed of the concrete material, which is the most intuitive visual representation of fluidity; a higher value usually corresponds to concrete with better fluidity;

[0094] By combining the surface flatness index reflecting static state with the average optical flow velocity amplitude reflecting dynamic trend, the static feature vector of the embodiment can more robustly evaluate the macroscopic apparent fluidity of concrete; the surface flatness focuses on the plastic behavior of the material after it is left for a short period of time, while the average optical flow velocity focuses on the overall rate of the material in motion; the combination of the two can provide more comprehensive visual information for the basic slump flow prediction model, thereby improving the accuracy and robustness of the basic prediction value A more accurate benchmark value is a prerequisite for achieving final high-precision prediction.

[0095] The intrinsic rheological state index is calculated by linear combination of global shear strength and structural entropy rate;

[0096] In the embodiment, the calculation paradigm of the intrinsic rheological state index is clarified, that is, a linear combination model is used; the model is built on the physical hypothesis: strong shear thinning or thixotropic effect must be manifested in the macroscopic as an increase in shear rate, characterized by , while in the microcosmic, it is accompanied by accelerated deconstruction and recombination of internal structure, characterized by ;

[0097] The calculation formula of the intrinsic rheological state index is as follows:

[0098]

[0099] Wherein, : intrinsic rheological state index, physical dimension of reciprocal time , used to represent the activity or rate of rheological behavior;

[0100] : global shear strength, dimension of ;

[0101] : structural entropy rate, dimension of ;

[0102] : dimensionless weight coefficient; the value is not arbitrarily set, but obtained through a series of offline calibration experiments; in the calibration process, the optimization goal is to maximize the Pearson correlation coefficient between the calculated value and the true value of the rheological performance measured by the laboratory rheometer, and the optimal coefficient pair is solved by least squares regression analysis;

[0103] Relatedly, the basic slump flow prediction value and the final corrected slump The calculation of the final corrected slump also involves the specific model;

[0104] The base slump The calculation formula of the final corrected slump is as follows:

[0105]

[0106] Wherein, : base slump, physical dimension is length ;

[0107] : surface flatness index, dimensionless quantity;

[0108] : average optical flow velocity amplitude, dimension is ;

[0109] : weight coefficient and bias constant, both dimensions are length ;

[0110] : weight coefficient, dimension is calibrated as time T to ensure the dimension of the term is length ;

[0111] The source of the above parameters is to calibrate the characteristic values calculated from the video of a large number of reference concrete samples without high-performance admixtures and the physically measured slump values through multiple linear regression analysis;

[0112] The calculation formula of the final corrected slump is as follows:

[0113]

[0114] : final corrected slump, physical dimension is length mm;

[0115] : base slump prediction value, the value is calculated from the previous formula;

[0116] : intrinsic rheological state index, the value is calculated from the previous formula;

[0117] : reference rheological state index, is a reference value representing normal rheological state, dimension is ; the source is to determine the average level of values by conducting a large number of experiments on concrete batches without or with conventional dosage of admixtures and with qualified performance; : the base of the natural logarithm, a mathematical constant;

[0118] : sensitivity coefficient of the correction function, dimension of physical quantity is set as time to ensure the dimensionless of the exponential term ; the target of the value calibration is to minimize the error between the corrected and the measured slump value of the sample containing the admixture;

[0119] the intrinsic rheological state index is constructed as a linear combination of and , which provides a computational framework that can effectively integrate macro and micro rheological information; through offline calibration, the is highly sensitive to the strong non-Newtonian effect introduced by chemical admixtures, but not sensitive to the viscosity change caused by simply adding water, thereby successfully decoupling the rheological state from the traditional visual features; the subsequent S-shaped correction function uses this decoupled index to realize a nonlinear adjustment mechanism that only corrects effectively when the rheological state deviates significantly from the baseline, greatly improving the accuracy of the final slump prediction, especially when dealing with modern high-performance concrete containing various types of high-efficiency admixtures.

[0120] output control instructions, including:

[0121] if the comprehensive risk index is higher than the risk threshold, trigger a high-risk alarm and generate preventive control instructions;

[0122] if the comprehensive risk index is not higher than the risk threshold, do not generate preventive control instructions;

[0123] In this embodiment, the decision logic and instruction generation mechanism based on the comprehensive risk index are elaborated in detail; The calculation formula is as follows:

[0124]

[0125] wherein, : comprehensive risk index, dimensionless;

[0126] : intrinsic rheological state index, the value is calculated from the previous step;

[0127] : baseline rheological state index, the value is a preset parameter;

[0128] : basic slump prediction value, the value is calculated from the previous step;

[0129] Final slump, value calculated from previous steps

[0130] Dimensionless risk coefficient, a constant calibrated by analyzing historical data, used to adjust the overall magnitude of risk index

[0131] The decision process is as follows:

[0132] Risk calculation: the risk assessment unit calculates the value of the comprehensive risk index in real time In the calculation, in order to prevent the denominator from being zero, when the predicted value of the basic slump is less than a preset minimum positive number , special processing should be performed, such as setting the risk index to a preset maximum value or triggering a specific alarm;

[0133] Threshold comparison: compare the calculated with the preset risk threshold ; is a critical value, which comes from the statistical analysis of the value distribution of the concrete batches corresponding to the quality accidents such as pipe blockage and segregation in history, so as to determine the limit that can effectively distinguish high risk from normal state;

[0134] Instruction generation:

[0135] If , it is determined that the current batch of concrete has high construction risk; the system automatically triggers a high-risk alarm and displays warning information on the control interface; at the same time, the instruction generation unit will query the preset rule library, such as generating clear and preventive control instructions, such as rheological state anomaly, suggesting to extend the mixing time by 20 seconds, recheck homogeneity or visual and rheological state is seriously decoupled, pumping risk is high, suggesting to pause pumping and take samples for recheck;

[0136] If , it is determined that the current batch of concrete is within the acceptable range, no preventive control instructions are generated, and the production process continues normally;

[0137] This method converts the vague and qualitative risk concept into a quantifiable and comparable index through a mathematical model that comprehensively considers the degree of rheological state anomaly and the degree of visual prediction deviation , which has a clear physical meaning; based on this index and the threshold value determined by statistics ​The decision making makes the risk control change from relying on artificial experience to automatic and standardized process, which not only greatly improves the timeliness and accuracy of risk identification, but also provides clear disposal suggestions for the operators through generating specific preventive instructions, so as to effectively avoid serious engineering accidents such as pumping pipe blockage and segregation bleeding caused by abnormal concrete flow performance.

[0138] It should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the present application.

Claims

1. A method of visual monitoring of a concrete mixing main unit, characterized by, The method comprises the following steps: S1, collecting a video image sequence of the movement of concrete in a mixing host in real time; S2, extracting a dynamic visual feature vector and a static feature vector based on the video image sequence; S3, calculating an intrinsic rheological state index based on the dynamic visual feature vector and a preset dynamic rheology-visual feature coupling model; S4, calculating a basic slump prediction value based on the static feature vector and a preset static correlation model; S5, determining a final corrected slump by a preset correction function in combination with the intrinsic rheological state index and the basic slump prediction value; S6, calculating a comprehensive risk index according to the final corrected slump, the basic slump prediction value and the intrinsic rheological state index; S7, outputting a control instruction based on the comprehensive risk index and a preset risk threshold. The dynamic visual feature vector comprises a global shear strength and a structural entropy rate. The determination of the global shear strength comprises the following steps: calculating an optical flow field between continuous image frames in the video image sequence; calculating a mean value of a spatial gradient norm of the optical flow velocity field as the global shear strength based on the optical flow field. The static feature vector comprises a surface flatness index and an average optical flow velocity amplitude. The intrinsic rheological state index is calculated by a linear combination of the global shear strength and the structural entropy rate. The calculation formula of the basic slump is as follows: wherein, : base slump, dimension of physical quantity is length ; : surface flatness index, dimensionless quantity : average optical flow velocity amplitude, dimension of ; : weight coefficient, dimension of physical quantity is length ; : bias constant, dimension of physical quantity is length ; : weight coefficient, dimension of physical quantity is calibrated as time ; Surface flatness indicator as input The formula for calculating the surface flatness indicator is: wherein, : surface flatness index, a dimensionless scalar; : average gradient, a spatial average of the gradient magnitude of all pixels in the region of interest (ROI); : a normal constant calibrated by experiment, which has the same dimension as the inverse of the dimension of the average gradient . and average gradient The formula for calculating the average gradient is: in, It is the total number of pixels involved in the calculation; The first one in the image Gradient magnitude of each pixel; Refers to the spatial coordinates of a pixel.

2. The method of claim 1, wherein the method further comprises: The determination of the structural entropy rate comprises the following steps: calculating a gray level co-occurrence matrix of a single image in the video image sequence to obtain a Shannon entropy; calculating a first-order derivative in a time dimension as the structural entropy rate based on the Shannon entropy.

3. The method of claim 1, wherein the method further comprises: The output control instruction comprises the following steps: if the comprehensive risk index is higher than the risk threshold, triggering a high-risk alarm and generating a preventive control instruction; if the comprehensive risk index is not higher than the risk threshold, not generating a preventive control instruction.

4. A concrete mixing main engine visual monitoring system based on the concrete mixing main engine visual monitoring method according to any one of claims 1-3, characterized in that, The method comprises the following steps: an image sequence collecting unit for collecting a video image sequence of the movement of concrete in a mixing host in real time; a feature extracting unit for extracting a dynamic visual feature vector and a static feature vector based on the video image sequence; a state identifying unit for calculating an intrinsic rheological state index based on the dynamic visual feature vector and a preset dynamic rheology-visual feature coupling model; a basic predicting unit for calculating a basic slump prediction value based on the static feature vector and a preset static correlation model; a correction calculating unit for determining a final corrected slump by a preset correction function in combination with the intrinsic rheological state index and the basic slump prediction value; a risk evaluating unit for calculating a comprehensive risk index according to the final corrected slump, the basic slump prediction value and the intrinsic rheological state index; an instruction generating unit for outputting a control instruction based on the comprehensive risk index and a preset risk threshold.

Images