CN-121275554-B - Real-time measurement system for rheological parameters of new concrete material

CN121275554BCN 121275554 BCN121275554 BCN 121275554BCN-121275554-B

Abstract

The invention relates to the technical field of building materials, and discloses a real-time measurement system for rheological parameters of a new concrete material, which comprises an image acquisition module, an image processing module, a characteristic extraction module, a real-time inversion module and a real-time inversion module, wherein the image acquisition module is used for acquiring a dynamic image sequence of the new concrete material in a motion state in real time, the image processing module is used for carrying out image registration on a intercepted effective analysis area sequence to obtain a preprocessed effective dynamic image sequence, the characteristic extraction module is used for inputting the preprocessed effective dynamic image sequence into a fusion model of a graph convolution neural network and a three-dimensional convolution neural network based on attention enhancement, extracting a space-time fusion characteristic vector representing rheological behavior of the concrete, and the real-time inversion module is used for inputting the space-time fusion characteristic vector into a rheological parameter inversion model constrained by physical information to invert and output yield stress and plastic viscosity of the concrete material in real time.

Inventors

Li Dingkui
ZHANG KE
WANG JIA
FENG YU
WANG LILI

Assignees

成都建工第九建筑工程有限公司
成都建工集团有限公司

Dates

Publication Date: 20260505
Application Date: 20251211

Claims (9)

1. A system for real-time measurement of rheological parameters of a new concrete material, the system comprising: the image acquisition module is used for acquiring a dynamic image sequence of the new concrete material in a motion state in real time; The image processing module is used for eliminating interference of uneven environmental illumination and shadows on the texture characteristics of the concrete surface by homomorphic filtering, automatically identifying and intercepting a continuous area of concrete flow in an image by combining an optical flow method with background subtraction to obtain an effective analysis area, and carrying out image registration on the intercepted effective analysis area sequence to obtain a preprocessed effective dynamic image sequence; The feature extraction module is used for inputting the preprocessed effective dynamic image sequence into a fusion model of the graph convolution neural network and the three-dimensional convolution neural network based on attention enhancement, and extracting space-time fusion feature vectors representing concrete rheological behaviors; The real-time inversion module is used for inputting the space-time fusion feature vector into a rheological parameter inversion model constrained by physical information, and inverting and outputting the yield stress and the plastic viscosity of the concrete material in real time; The image processing module includes: the conversion sub-module is used for converting the acquired dynamic image sequence into a gray image, carrying out logarithmic transformation on the gray image, and carrying out two-dimensional Fourier transformation on the image after logarithmic transformation to obtain a frequency domain image; A filtering sub-module, configured to suppress a low-frequency component in the frequency domain image and retain a high-frequency component through a gaussian high-pass filter, and perform two-dimensional inverse fourier transform on the filtered frequency domain image; the restoration submodule is used for carrying out exponential transformation on the image after the inverse transformation, restoring the image to an original gray value range, and obtaining a gray image sequence for eliminating uneven illumination and shadow interference and retaining the texture characteristics of the concrete surface; The computing sub-module is used for computing an optical flow field of each frame of image by adopting a Lucas-Kanade optical flow method, selecting 30 frames of static images when concrete does not flow, computing a mean value and a standard deviation of gray values of each pixel, and generating an initial background frame; the first determining submodule is used for carrying out differential operation on each frame of image and the initial background frame to obtain a foreground differential image, and carrying out binarization processing on the foreground differential image to determine a region with a gray level difference value larger than a first threshold value as a candidate foreground region; The intersection operation sub-module is used for performing intersection operation on the region with the motion vector module value larger than the second threshold value in the optical flow field and the candidate foreground region to obtain a continuous region where concrete flows; The cutting sub-module is used for cutting pixels in the minimum circumscribed rectangular coordinate range in each frame of image according to the minimum circumscribed rectangular coordinate of the continuous area where the concrete flows, so as to obtain an effective analysis area sequence only comprising the concrete flowing area; The space alignment sub-module is used for selecting a first frame of the effective analysis area sequence as a reference frame, realizing the space alignment of the frame to be registered and the reference frame by adopting a SIFT feature extraction algorithm for each frame to be registered in the effective analysis area sequence, and finishing the registration for all frames in the effective analysis area sequence to obtain the preprocessed effective dynamic image sequence.
2. A system for real-time measurement of rheological parameters of a new concrete material according to claim 1 wherein the calculation submodule comprises: Extracting a corner point with high contrast from a gray level image of a current frame by using a Shi-Tomasi corner point detection algorithm as a feature point; And screening effective feature points through angular point response value threshold values, establishing a pixel gray level change equation for the feature points in two adjacent frames of images, and solving horizontal and vertical motion vectors of the feature points through a least square method to obtain an optical flow field of each frame of image.
3. A system for real-time measurement of rheological parameters of a new concrete material according to claim 1 wherein the spatial alignment sub-module comprises: constructing a scale space in the effective areas of the frame to be registered and the reference frame, and detecting extreme points under each scale; Positioning and screening the extreme points, calculating the main direction of each feature point, and generating feature descriptors; calculating the similarity between the frame to be registered and the reference frame feature descriptors through Euclidean distance, and screening feature point pairs with the similarity smaller than a third threshold value; Randomly selecting 8 pairs of characteristic points, calculating an affine transformation matrix, verifying the reprojection errors of other characteristic point pairs of the affine transformation matrix, and reserving the characteristic point pairs with the errors smaller than a fourth threshold value; And iteratively determining an optimal affine transformation matrix, carrying out pixel remapping on an effective area of the frame to be registered according to the optimal affine transformation matrix, converting each pixel coordinate of the effective area of the frame to be registered into a corresponding coordinate under a reference frame coordinate system, and calculating a pixel gray value of the converted coordinate by adopting a bilinear interpolation method to realize the spatial alignment of the frame to be registered and the reference frame.
4. A system for real-time measurement of rheological parameters of a new concrete material according to claim 1 wherein the feature extraction module comprises: the first convolution sub-module is used for stacking the preprocessed effective dynamic image sequence according to the time dimension, inputting a three-dimensional convolution neural network comprising 3 convolution layers and 2 pooling layers, and repeating the convolution and pooling processes to obtain an apparent motion characteristic vector; The second convolution sub-module is used for constructing a graph structure based on the preprocessed effective dynamic image sequence, inputting a graph convolution neural network comprising 2 graph convolution layers and obtaining a spatial structure relation feature vector; And the fusion submodule is used for converting the apparent motion feature vector and the spatial structure relation feature vector into feature vectors with the same dimension through two full-connection layers, fusing the feature vectors through a cross-modal attention mechanism and generating a space-time fusion feature vector through a ReLU function.
5. A system for real-time measurement of rheological parameters of a new concrete material according to claim 4 wherein the first convolution sub-module comprises: Sliding on the space and time dimensions of the effective dynamic image sequence by using a 3 multiplied by 3 three-dimensional convolution kernel, and carrying out convolution operation on the space pixel difference and the time pixel change of each local area to generate an initial space-time characteristic diagram; And (3) selecting the maximum value in each pooling window through a maximum pooling layer of 2 multiplied by 2, repeating the rolling and pooling processes, and converting the feature map into an apparent motion feature vector through a global average pooling layer, wherein the apparent motion feature vector is used for representing the flow speed and the surface morphology change of the concrete.
6. A system for real-time measurement of rheological parameters of a new concrete material according to claim 4 wherein the second convolution sub-module comprises: the first layer of graph convolution kernel updates node characteristics through weighted summation, captures local spatial relationships, the second layer of graph convolution kernel strengthens characteristics of key areas through an attention mechanism, and the node characteristics are integrated into spatial structural relationship characteristic vectors through a fully-connected layer, wherein the spatial structural relationship characteristic vectors are used for representing distribution and contact relationships between cement paste and aggregate.
7. A system for real-time measurement of rheological parameters of a new concrete material according to claim 1 wherein the real-time inversion module comprises: The standardized processing submodule is used for carrying out Z-score standardized processing on the space-time fusion feature vector, inputting the standardized feature vector into a rheological parameter inversion model comprising 3 full-connection layers, and outputting a preliminary rheological parameter predicted value; the second determining submodule is used for calculating the deviation between the preliminary rheological parameter predicted value and the physical constraint determined based on the Bingham fluid model to obtain a physical rule loss term; and the adjusting sub-module is used for adjusting the preliminary rheological parameter predicted value according to the current physical rule loss item and transmitting the adjusted yield stress and plastic viscosity value in real time.
8. A system for real-time measurement of rheological parameters of a new concrete material according to claim 7 wherein the standardized processing sub-module comprises: the first full-connection layer uses a ReLU activation function to perform nonlinear conversion on the normalized feature vector, the second full-connection layer uses the ReLU activation function to extract key features, and the third full-connection layer outputs a preliminary rheological parameter predicted value.
9. A method for implementing a system for real-time measurement of rheological parameters of a new concrete material according to claim 1, characterized in that it comprises the following steps: acquiring a dynamic image sequence of a new concrete material in a motion state in real time; The method comprises the steps of adopting homomorphic filtering to eliminate interference of uneven ambient illumination and shadows on the texture characteristics of the concrete surface, automatically identifying and intercepting a continuous area of concrete flow in an image by combining an optical flow method with background subtraction to obtain an effective analysis area, and carrying out image registration on an intercepted effective analysis area sequence to obtain a preprocessed effective dynamic image sequence; inputting the preprocessed effective dynamic image sequence into a fusion model of a graph convolution neural network and a three-dimensional convolution neural network based on attention enhancement, and extracting a space-time fusion feature vector representing concrete rheological behavior; And inputting the space-time fusion feature vector into a rheological parameter inversion model constrained by physical information, and inverting and outputting the yield stress and the plastic viscosity of the concrete material in real time.

Description

Real-time measurement system for rheological parameters of new concrete material Technical Field The invention relates to the technical field of building materials, in particular to a real-time measurement system for rheological parameters of a new concrete material. Background The concrete rheological parameters directly determine the key construction performances such as pouring fluidity, pumping resistance and the like, the accurate measurement is critical to the guarantee of engineering quality, the existing measurement means mainly comprise contact type, such as a rotary rheometer, and laboratory detection is needed after the concrete is sampled from an engineering site, so that the rheological property of the concrete in an actual motion state cannot be reflected in real time, and the problems that the original state of a material is damaged in the sampling process and the detection result is lagged behind the construction progress exist. Disclosure of Invention The invention aims to solve the problems, and designs a real-time measurement system for rheological parameters of a new concrete material. The first aspect of the invention provides a real-time measurement system for rheological parameters of a new concrete material, comprising: the image acquisition module is used for acquiring a dynamic image sequence of the new concrete material in a motion state in real time; The image processing module is used for eliminating interference of uneven environmental illumination and shadows on the texture characteristics of the concrete surface by homomorphic filtering, automatically identifying and intercepting a continuous area of concrete flow in an image by combining an optical flow method with background subtraction to obtain an effective analysis area, and carrying out image registration on the intercepted effective analysis area sequence to obtain a preprocessed effective dynamic image sequence; The feature extraction module is used for inputting the preprocessed effective dynamic image sequence into a fusion model of the graph convolution neural network and the three-dimensional convolution neural network based on attention enhancement, and extracting space-time fusion feature vectors representing concrete rheological behaviors; And the real-time inversion module is used for inputting the space-time fusion feature vector into a rheological parameter inversion model constrained by physical information, and inverting and outputting the yield stress and the plastic viscosity of the concrete material in real time. Optionally, in a first implementation manner of the first aspect of the present invention, the image processing module includes: the conversion sub-module is used for converting the acquired dynamic image sequence into a gray image, carrying out logarithmic transformation on the gray image, and carrying out two-dimensional Fourier transformation on the image after logarithmic transformation to obtain a frequency domain image; A filtering sub-module, configured to suppress a low-frequency component in the frequency domain image and retain a high-frequency component through a gaussian high-pass filter, and perform two-dimensional inverse fourier transform on the filtered frequency domain image; the restoration submodule is used for carrying out exponential transformation on the image after the inverse transformation, restoring the image to an original gray value range, and obtaining a gray image sequence for eliminating uneven illumination and shadow interference and retaining the texture characteristics of the concrete surface; The computing sub-module is used for computing an optical flow field of each frame of image by adopting a Lucas-Kanade optical flow method, selecting 30 frames of static images when concrete does not flow, computing a mean value and a standard deviation of gray values of each pixel, and generating an initial background frame; the first determining submodule is used for carrying out differential operation on each frame of image and the initial background frame to obtain a foreground differential image, and carrying out binarization processing on the foreground differential image to determine a region with a gray level difference value larger than a first threshold value as a candidate foreground region; The intersection operation sub-module is used for performing intersection operation on the region with the motion vector module value larger than the second threshold value in the optical flow field and the candidate foreground region to obtain a continuous region where concrete flows; The cutting sub-module is used for cutting pixels in the minimum circumscribed rectangular coordinate range in each frame of image according to the minimum circumscribed rectangular coordinate of the continuous area where the concrete flows, so as to obtain an effective analysis area sequence only comprising the concrete flowing area; The space alignment sub-module is used for selecting a fir