Material feature detection method and system fusing spatio-temporal visual features and process parameters

Abstract

This invention relates to the field of materials testing technology, specifically disclosing a method and system for detecting materials features by fusing spatiotemporal visual features and process parameters. The method includes acquiring video stream data of the polymer to be tested under controlled operating conditions, simultaneously acquiring process parameters; performing spatiotemporal processing on the video stream data based on a preset dual-stream network, outputting spatial feature vectors and temporal feature vectors; obtaining a material flowability index and a material viscosity characterization index based on the spatial and temporal feature vectors, and determining the state identification result of the polymer. This invention combines fluid mechanics principles with a computer vision deep learning model to achieve non-contact, real-time, and automated identification and monitoring of material viscosity and flowability. The original input is a high frame rate video stream; the spatial appearance information and temporal motion information are processed separately through a dual-stream network, and the identification results are fused to obtain material features. The method boasts high identification efficiency and high accuracy.

Description

Technical Field

[0001] This invention relates to the field of materials testing technology, specifically a method and system for detecting materials characteristics that integrates spatiotemporal visual features and process parameters. Background Technology

[0002] In the processing of polymer materials, such as extrusion, injection molding, casting, and compounding, the flow state and viscosity characteristics have a direct impact on the quality of the products. Changes in the flow state of polymers are usually accompanied by changes in the surface morphology, internal flow structure, and processing stability of the material. Therefore, accurate detection and identification of the state of polymers during processing is an important foundation for achieving process optimization and quality control.

[0003] In existing technologies, the detection of polymer viscosity and rheological properties mainly relies on offline testing or indirect monitoring methods. For example, offline testing of material samples is performed using specialized equipment such as rotational rheometers and capillary rheometers to obtain viscosity or rheological parameters; or the flow state of the polymer is indirectly inferred by monitoring process parameters such as pressure, temperature, and rotation speed of the processing equipment. Although these methods have a certain degree of accuracy under experimental conditions, they generally suffer from problems such as long testing cycles, inability to reflect real-time processing status, and deviations from actual working conditions, making it difficult to meet the needs of continuous and intelligent production. How to provide a real-time material characteristic detection scheme to adapt to the needs of continuous and intelligent production is the technical problem that this invention aims to solve. Summary of the Invention

[0004] The purpose of this invention is to provide a material feature detection method and system that integrates spatiotemporal visual features and process parameters to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for detecting material features by integrating spatiotemporal visual features and process parameters, the method comprising:

[0007] Acquire video stream data of the polymer to be tested under controlled operating conditions, and simultaneously collect process parameters; wherein, the video stream data and process parameters share the same time scale;

[0008] Based on a pre-defined dual-stream network, video stream data is spatiotemporally processed to output spatial feature vectors and temporal feature vectors. The dual-stream network includes a spatial stream network and a temporal stream network. The spatial stream network is used to extract the spatial appearance features of the material, including static texture distribution, morphological contours, and bubble morphology. The temporal stream network is used to extract the temporal motion features of the material, including optical flow changes, motion trajectories, local eddy current structures, and flow continuity.

[0009] Based on the preset material rheological characteristic fusion model, the process parameters, spatial feature vectors and temporal feature vectors are processed to obtain the material flowability index and the material viscosity characterization index.

[0010] The state identification result of the polymer is determined based on the material flowability index and the material viscosity characterization index.

[0011] As a further aspect of the present invention: the step of acquiring video stream data of the polymer to be detected under controlled operating conditions and simultaneously acquiring process parameters includes:

[0012] Receive working conditions input by staff, and determine the sensing device based on the working conditions;

[0013] The process parameters are acquired in real time based on the sensing device, and image frames of the polymer to be detected are acquired based on the preset image acquisition device; the process parameters include at least stirring speed and temperature;

[0014] Perform temporal registration between process parameters and image frames to determine the process parameters corresponding to each image frame;

[0015] The image frames are sorted according to time order to obtain video stream data.

[0016] As a further aspect of the present invention: the step of performing spatiotemporal processing on video stream data based on a preset dual-stream network to output spatial feature vectors and temporal feature vectors includes:

[0017] Receive the processing density input by the staff and determine the time point based on the processing density;

[0018] Image frames are extracted from the video stream data based on the time point, and the image frames are input into the spatial stream network to obtain spatial feature vectors; the extracted image frames include the image frames at the latest time point.

[0019] Read the extracted image frame and its previous frame, input them into the temporal flow network, and obtain the temporal feature vector;

[0020] In this system, both the spatial and temporal feature vectors are labeled with the timestamps of the extracted image frames. The spatial flow network is used to extract the spatial appearance features of the polymer during the flow process. These spatial appearance features include, but are not limited to, static texture distribution, morphological contour features, and bubble morphological distribution, which are used to characterize the appearance and structural state of the material at a specific moment. The temporal flow network is used to extract the temporal motion features of the polymer. These temporal motion features include optical flow changes, motion trajectories, local eddy structures, and flow continuity, which are used to characterize the dynamic flow behavior of the material over time.

[0021] As a further aspect of the present invention: the step of processing process parameters, spatial feature vectors, and temporal feature vectors based on a preset material rheological feature fusion model to obtain the material flowability index and the material viscosity characterization index includes:

[0022] Read the specific time points, and read the process parameters, spatial feature vectors, and temporal feature vectors at each time point;

[0023] The process parameters are normalized to obtain a process parameter vector;

[0024] The spatial feature vector and the temporal feature vector are corrected based on the process parameter vector;

[0025] The corrected spatial feature vector and temporal feature vector are fused to obtain the fused vector;

[0026] The material flowability index and material viscosity characterization index are determined based on the fusion vector, which are used to characterize the flowability and viscosity characteristics of the polymer under current operating conditions, respectively.

[0027] As a further aspect of the present invention: the step of determining the state identification result of the polymer based on the material flowability index and the material viscosity characterization index includes:

[0028] The material flowability index and material viscosity characterization index at each time point are compared with the preset state range;

[0029] The state recognition result at each moment is determined based on the comparison results;

[0030] Based on the state identification results at each time point, determine the state identification result of the polymer;

[0031] The acquisition process is recursively adjusted based on the state recognition results at each time point; the recursive adjustment targets include the acquisition frequency of the sensing device, the acquisition frequency of the image acquisition device, and the processing density.

[0032] As a further aspect of the present invention, the method further includes:

[0033] Establish connection channels with each intelligent module in the experimental scenario;

[0034] For each testing process, the material flowability index and material viscosity characterization index are read at each moment;

[0035] Based on the connection channel, all data within a preset time period is acquired to construct a global database;

[0036] The detection process is clustered based on the material flowability index and material viscosity characterization index at each time point;

[0037] For each type of detection process, the data is compared with the global database, the intersection is calculated, and the intersection is used as additional environmental conditions and fed back to the detection party.

[0038] The present invention also provides a material feature detection system that integrates spatiotemporal visual features and process parameters, the system comprising:

[0039] The process parameter acquisition module is used to acquire video stream data of the polymer to be tested under controlled operating conditions and to acquire process parameters simultaneously; wherein, the video stream data and the process parameters share the same time scale;

[0040] The dual-stream recognition module is used to perform spatiotemporal processing on video stream data based on a preset dual-stream network, outputting spatial feature vectors and temporal feature vectors. The dual-stream network includes a spatial stream network and a temporal stream network. The spatial stream network is used to extract the spatial appearance features of the material, including static texture distribution, morphological contours, and bubble morphology. The temporal stream network is used to extract the temporal motion features of the material, including optical flow changes, motion trajectories, local eddy current structures, and flow continuity.

[0041] The parameter processing module is used to process process parameters, spatial feature vectors and temporal feature vectors based on a preset material rheological feature fusion model to obtain material flowability index and material viscosity characterization index.

[0042] The identification result output module is used to determine the state identification result of the polymer based on the material flowability index and the material viscosity characterization index.

[0043] As a further aspect of the present invention: the process parameter acquisition module includes:

[0044] The sensing device determination unit is used to receive the working conditions input by the staff and determine the sensing device based on the working conditions.

[0045] An image acquisition unit is used to acquire process parameters in real time based on the sensing device and to acquire image frames of the polymer to be detected based on a preset image acquisition device; the process parameters include at least stirring speed and temperature;

[0046] The temporal registration unit is used to perform temporal registration of process parameters and image frames to determine the process parameters corresponding to each image frame.

[0047] The image sorting unit is used to sort image frames based on time order to obtain video stream data.

[0048] As a further aspect of the present invention: the dual-stream identification module includes:

[0049] The time point determination unit is used to receive the processing density input by the staff and determine the time point based on the processing density.

[0050] The spatial vector output unit is used to extract image frames from the video stream data based on the time point, input the image frames into the spatial stream network, and obtain spatial feature vectors; the extracted image frames include the image frames at the latest time point.

[0051] The temporal vector output unit is used to read the extracted image frame and its previous frame, input them into the temporal flow network, and obtain the temporal feature vector.

[0052] In this system, both the spatial and temporal feature vectors are labeled with the timestamps of the extracted image frames. The spatial flow network is used to extract the spatial appearance features of the polymer during the flow process. These spatial appearance features include, but are not limited to, static texture distribution, morphological contour features, and bubble morphological distribution, which are used to characterize the appearance and structural state of the material at a specific moment. The temporal flow network is used to extract the temporal motion features of the polymer. These temporal motion features include optical flow changes, motion trajectories, local eddy structures, and flow continuity, which are used to characterize the dynamic flow behavior of the material over time.

[0053] As a further aspect of the present invention: the parameter processing module includes:

[0054] The data reading unit is used to read a specific point in time, and to read the process parameters, spatial feature vector, and temporal feature vector at each point in time.

[0055] The normalization processing unit is used to normalize the process parameters to obtain a process parameter vector;

[0056] The vector correction unit is used to correct the spatial feature vector and the temporal feature vector based on the process parameter vector, respectively.

[0057] The vector fusion unit is used to fuse the corrected spatial feature vector and temporal feature vector to obtain a fused vector.

[0058] The index generation unit is used to determine the material flowability index and the material viscosity characterization index based on the fusion vector, which are used to characterize the flowability and viscosity characteristics of the polymer under the current operating conditions, respectively.

[0059] Compared with the prior art, the beneficial effects of the present invention are: the present invention combines the principles of fluid mechanics with computer vision deep learning models to achieve non-contact, real-time and automated identification and monitoring of material viscosity and its flowability. The original input is a high frame rate video stream, and the spatial appearance information and temporal motion information are processed separately through a dual-stream network. The identification results are fused to obtain material features. This process can be carried out in real time and has high identification efficiency and high accuracy. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0061] Figure 1 A flowchart of a material feature detection method that integrates spatiotemporal visual features and process parameters.

[0062] Figure 2 A block diagram of the structure of a material feature detection system that integrates spatiotemporal visual features and process parameters. Detailed Implementation

[0063] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0064] Figure 1 The flowchart of a material feature detection method that integrates spatiotemporal visual features and process parameters is shown in this embodiment of the invention. The method includes:

[0065] Step S100: Acquire video stream data of the polymer to be tested under controlled operating conditions, and simultaneously collect process parameters; wherein, the video stream data and process parameters share the same time scale;

[0066] Controlled operating conditions refer to preset test conditions, which are generally determined by the actual working scenario. After determining the controlled operating conditions, the obtained state recognition result is the test result under the corresponding test conditions. The test is carried out under controlled operating conditions. The video stream data of the polymer to be tested under controlled operating conditions is acquired by the vision acquisition device, and the process parameters are acquired simultaneously. The process parameters include at least one or more of stirring speed, time, and temperature. For the acquired video stream data, the video stream is processed by denoising, frame synchronization and time alignment to form standardized input data that can be used for feature extraction.

[0067] Step S200: Perform spatiotemporal processing on the video stream data based on a preset dual-stream network, and output spatial feature vectors and temporal feature vectors; wherein, the dual-stream network includes a spatial stream network and a temporal stream network. The spatial stream network is used to extract the spatial appearance features of the material, including static texture distribution, morphological contours and bubble morphology; the temporal stream network is used to extract the temporal motion features of the material, including optical flow changes, motion trajectory, local eddy structure and flow continuity.

[0068] The video stream data is subjected to spatiotemporal processing. The processing is carried out with the help of a pre-set dual-stream network. The dual-stream network processes the video stream data in the temporal domain and the spatial domain respectively to obtain spatial feature vectors and temporal feature vectors.

[0069] Step S300: Based on the preset material rheological characteristic fusion model, process parameters, spatial feature vectors and time feature vectors are processed to obtain the material flowability index and the material viscosity characterization index;

[0070] For the obtained spatial and temporal feature vectors, feature extraction is performed using a pre-defined material rheological feature fusion model to obtain two parameters: the material flowability index and the material viscosity characterization index. These are data that need to be measured from the polymer. Process parameters are used to determine the model parameters of the material rheological feature fusion model itself, enabling the model to jointly model the spatial and temporal feature vectors under process parameter constraints. By introducing process parameters to conditionally constrain, normalize, or weight the visual features, the influence of different operating conditions on the amplitude and representation of the visual features is eliminated. Based on the fusion modeling results, the material flowability index and the material viscosity characterization index are output. These indices characterize the flowability and viscosity properties of the polymer under the current operating conditions, and are quantitative indicators that reflect the material rheological properties.

[0071] Step S400: Determine the state identification result of the polymer based on the material flowability index and the material viscosity characterization index;

[0072] The index is compared and analyzed with a preset state range or reference model to determine whether the polymer to be tested is in a target rheological state, an abnormal state, or a transitional state, thereby outputting the state identification result of the polymer. The state identification result can be used to characterize whether the material meets the expected process requirements, or for subsequent process adjustment, quality control, and process monitoring.

[0073] Regarding step S100, the step of acquiring video stream data of the polymer to be detected under controlled operating conditions and synchronously acquiring process parameters includes:

[0074] Receive working conditions input by staff, and determine the sensing device based on the working conditions;

[0075] The process parameters are acquired in real time based on the sensing device, and image frames of the polymer to be detected are acquired based on the preset image acquisition device; the process parameters include at least stirring speed and temperature;

[0076] Perform temporal registration between process parameters and image frames to determine the process parameters corresponding to each image frame;

[0077] The image frames are sorted according to time order to obtain video stream data.

[0078] In one example of the technical solution of this invention, the data acquisition process is described. The system receives operating conditions input by the operator, determines a sensing device based on these conditions, acquires process parameters in real time using the sensing device, and acquires image frames of the polymer to be detected using a preset image acquisition device. The process parameters include at least stirring speed and temperature. Both the sensing device and the image acquisition device have built-in time stamps. The process parameters and image frames are time-domain registered to determine the process parameters corresponding to each image frame. The image frames are then sorted according to time order to obtain video stream data. This process is actually not complicated; it involves acquiring process parameters through a sensing device, acquiring images through an image acquisition device, and registering the acquired data using a time axis to obtain the raw data.

[0079] Regarding step S200, the step of performing spatiotemporal processing on the video stream data based on a preset dual-stream network to output spatial feature vectors and temporal feature vectors includes:

[0080] Receive the processing density input by the staff and determine the time point based on the processing density;

[0081] Image frames are extracted from the video stream data based on the time point, and the image frames are input into the spatial stream network to obtain spatial feature vectors; the extracted image frames include the image frames at the latest time point.

[0082] The extracted image frame and its previous frame are read and input into the temporal flow network to obtain the temporal feature vector.

[0083] In one example of the technical solution of this invention, the process of spatiotemporal processing of video stream data is defined. The processing density is received from the operator, and time points are determined based on this density. The processing density is used to adjust the time interval, for example, how often a time point is selected. The start or end point of each time point also needs to be set. The simplest method is to record the processing time as the tail time when processing is required, and then sequentially obtain time points forward based on the time interval determined by the processing density. Image frames are extracted from the video stream data based on the time points, and the image frames are input into a spatial stream network to obtain spatial feature vectors. The extracted image frames and their preceding frames are read and input into a temporal stream network to obtain temporal feature vectors. It can be understood that spatial feature vectors are essentially used to identify an image at a specific instant, while temporal feature vectors are used to identify an image and its preceding image, determining change information.

[0084] In this system, both spatial and temporal feature vectors are labeled with the timestamps of the extracted image frames. The spatial flow network is used to extract the spatial appearance features of the polymer during the flow process. These spatial appearance features include, but are not limited to, static texture distribution, morphological contour features, and bubble morphological distribution, which are used to characterize the appearance and structural state of the material at a specific moment. The temporal flow network is used to extract the temporal motion features of the polymer. These temporal motion features include optical flow changes, motion trajectories, local eddy structures, and flow continuity, which are used to characterize the dynamic flow behavior of the material over time. Both spatial and temporal flow networks have corresponding image processing algorithms in the prior art, which will not be elaborated here.

[0085] Regarding step S300, the step of processing process parameters, spatial feature vectors, and temporal feature vectors based on a preset material rheological feature fusion model to obtain the material flowability index and the material viscosity characterization index includes:

[0086] Read the specific time points, and read the process parameters, spatial feature vectors, and temporal feature vectors at each time point;

[0087] The process parameters are normalized to obtain a process parameter vector;

[0088] The spatial feature vector and the temporal feature vector are corrected based on the process parameter vector;

[0089] The corrected spatial feature vector and temporal feature vector are fused to obtain the fused vector;

[0090] The material flowability index and material viscosity characterization index are determined based on the fusion vector, which are used to characterize the flowability and viscosity characteristics of the polymer under current operating conditions, respectively.

[0091] After processing by the dual-flow network, spatial and temporal feature vectors are obtained. A specific time point is read, and for each time point, the process parameters, spatial feature vector, and temporal feature vector are read. The process parameters are normalized to obtain a process parameter vector, which is an array of process parameters. Based on the process parameter vector, the spatial and temporal feature vectors are corrected respectively. The corrected spatial and temporal feature vectors are then fused to obtain a fused vector. Based on the fused vector, the material flowability index and material viscosity characterization index are determined, used to characterize the polymer's flowability and viscosity properties under current operating conditions, respectively. The correction process, the fusion process, and the generation process from the fused vector to the index are all included in the material rheological feature fusion model.

[0092] Specifically, examples of the above processing procedures are explained as follows:

[0093] 1. Standardize process parameters. Standardization is necessary to convert the data into dimensionless data. The standardization process can be achieved using a maximum-minimum method. 2. Adjust spatial and temporal features based on process parameters. This requires constructing a weight function with process parameters as independent variables from historical data. The calculated weights are then used to adjust the spatial and temporal feature vectors by direct multiplication. 3. Vector fusion. This process is equivalent to an integration process, concatenating vectors into a higher-dimensional vector, resulting in a dataset. 4. Calculate the rheological index. This involves sequentially extracting the parameters corresponding to the index from the dataset and then superimposing the parameters with preset weights to obtain the index. Each component in the superposition process can use either a linear or exponential architecture.

[0094] Regarding step S400, the step of determining the polymer state identification result based on the material flowability index and the material viscosity characterization index includes:

[0095] The material flowability index and material viscosity characterization index at each time point are compared with the preset state range;

[0096] The state recognition result at each moment is determined based on the comparison results;

[0097] Based on the state identification results at each time point, determine the state identification result of the polymer;

[0098] The acquisition process is recursively adjusted based on the state recognition results at each time point; the recursive adjustment targets include the acquisition frequency of the sensing device, the acquisition frequency of the image acquisition device, and the processing density.

[0099] In one example of the technical solution of the present invention, the process of determining the state identification result is described. The material flowability index and material viscosity characterization index at each time are compared with the preset state interval. In practical applications, three intervals are generally set, such as the target state interval, the transition state interval, and the abnormal state interval. If the index falls into the target interval, it is determined to be the target state. If the index deviates from the target interval but does not exceed the abnormal threshold, it is determined to be the transition state. If the index exceeds the abnormal threshold, it is determined to be the abnormal state. Each state corresponds to a value.

[0100] By performing the same analysis at each time point, we can obtain the state recognition result at each time point. This is actually a collection of state recognition results, which can be used as output and fed back to the tester.

[0101] It should be noted that, based on the state recognition results at each time point, the above content also introduces a recursive adjustment process. The recursive adjustment targets include the acquisition frequency of the sensing device, the acquisition frequency of the image acquisition device, and the processing density. Specifically, the acquisition frequency of the sensing device and the image acquisition device are generally synchronized. In abnormal situations, they need to be increased. The processing density represents how much data is selected for analysis. The more data selected, the more computing resources are consumed, and the more accurate the results are. Under this architecture, increasing the acquisition frequency of the sensing device, the acquisition frequency of the image acquisition device, and the processing density all correspond to the same situation: the existence of abnormal states in the state recognition results. How to determine the existence of abnormal states in the state recognition results requires a pre-defined rule. The simplest case is to accumulate the number of time points corresponding to abnormal states and determine the correction magnitude proportional to the number, thereby increasing the acquisition frequency of the sensing device, the acquisition frequency of the image acquisition device, and the processing density.

[0102] As a preferred embodiment of the technical solution of the present invention, the method further includes:

[0103] Establish connection channels with each intelligent module in the experimental scenario;

[0104] For each testing process, the material flowability index and material viscosity characterization index are read at each moment;

[0105] Based on the connection channel, all data within a preset time period is acquired to construct a global database;

[0106] The detection process is clustered based on the material flowability index and material viscosity characterization index at each time point;

[0107] For each type of detection process, the data is compared with the global database, the intersection is calculated, and the intersection is used as additional environmental conditions and fed back to the detection party.

[0108] In one embodiment of the technical solution of this invention, a potential information extraction scheme is provided. This scheme establishes a connection channel with various intelligent modules in the experimental scenario. Given the high prevalence of existing intelligent modules and the existence of a large amount of data in the laboratory where the test is conducted to characterize the environmental conditions, the scheme reads the material flowability index and material viscosity index at each moment during each test. Based on the connection channel, all data within a preset time period is acquired to construct a global database. The test processes are clustered based on the material flowability index and material viscosity index at each moment. For each cluster of test processes, the intersection is calculated by comparing it with the global database and used as an additional environmental condition, which is then fed back to the testing party. This process implies that if the test results (material flowability index and material viscosity index) are the same, are the environmental parameters also the same? In a specific experimental scenario, the experiment... The reproducibility rate is a crucial indicator. It's highly likely that similar individuals conducting similar experiments can yield vastly different results. A significant reason for this is the difference in environment. An experiment is a complex system. Existing equipment only limits the experimental conditions (the operating conditions in this invention) to achieve the same results. This process relies on the assumption of similar environments, such as a standard laboratory environment, which is difficult to achieve. To address this issue, this invention compares each type of detection process against a comprehensive database, calculates the intersection, and uses this intersection as additional environmental conditions, feeding them back to the testing party. These additional environmental conditions are for reference only; they are not the actual operating conditions. They are only used as a reference for cause analysis in cases of reproduction failure. This is highly beneficial for the reproduction process and is a unique feature of this invention.

[0109] Figure 2 The structural block diagram of a material feature detection system that integrates spatiotemporal visual features and process parameters is shown in this embodiment of the invention. The system 10 includes:

[0110] The process parameter acquisition module 11 is used to acquire video stream data of the polymer to be tested under controlled operating conditions and simultaneously acquire process parameters; wherein, the video stream data and process parameters share the same time scale;

[0111] The dual-stream recognition module 12 is used to perform spatiotemporal processing on video stream data based on a preset dual-stream network, and output spatial feature vectors and temporal feature vectors. The dual-stream network includes a spatial stream network and a temporal stream network. The spatial stream network is used to extract the spatial appearance features of the material, including static texture distribution, morphological contours and bubble morphology. The temporal stream network is used to extract the temporal motion features of the material, including optical flow changes, motion trajectory, local eddy structure and flow continuity.

[0112] The parameter processing module 13 is used to process process parameters, spatial feature vectors and time feature vectors based on a preset material rheological feature fusion model to obtain material flowability index and material viscosity characterization index.

[0113] The identification result output module 14 is used to determine the state identification result of the polymer based on the material flowability index and the material viscosity characterization index.

[0114] Furthermore, the process parameter acquisition module 11 includes:

[0115] The sensing device determination unit is used to receive the working conditions input by the staff and determine the sensing device based on the working conditions.

[0116] An image acquisition unit is used to acquire process parameters in real time based on the sensing device and to acquire image frames of the polymer to be detected based on a preset image acquisition device; the process parameters include at least stirring speed and temperature;

[0117] The temporal registration unit is used to perform temporal registration of process parameters and image frames to determine the process parameters corresponding to each image frame.

[0118] The image sorting unit is used to sort image frames based on time order to obtain video stream data.

[0119] Specifically, the dual-stream identification module 12 includes:

[0120] The time point determination unit is used to receive the processing density input by the staff and determine the time point based on the processing density.

[0121] The spatial vector output unit is used to extract image frames from the video stream data based on the time point, input the image frames into the spatial stream network, and obtain spatial feature vectors; the extracted image frames include the image frames at the latest time point.

[0122] The temporal vector output unit is used to read the extracted image frame and its previous frame, input them into the temporal flow network, and obtain the temporal feature vector.

[0123] In this system, both the spatial and temporal feature vectors are labeled with the timestamps of the extracted image frames. The spatial flow network is used to extract the spatial appearance features of the polymer during the flow process. These spatial appearance features include, but are not limited to, static texture distribution, morphological contour features, and bubble morphological distribution, which are used to characterize the appearance and structural state of the material at a specific moment. The temporal flow network is used to extract the temporal motion features of the polymer. These temporal motion features include optical flow changes, motion trajectories, local eddy structures, and flow continuity, which are used to characterize the dynamic flow behavior of the material over time.

[0124] Furthermore, the parameter processing module 13 includes:

[0125] The data reading unit is used to read a specific point in time, and to read the process parameters, spatial feature vector, and temporal feature vector at each point in time.

[0126] The normalization processing unit is used to normalize the process parameters to obtain a process parameter vector;

[0127] The vector correction unit is used to correct the spatial feature vector and the temporal feature vector based on the process parameter vector, respectively.

[0128] The vector fusion unit is used to fuse the corrected spatial feature vector and temporal feature vector to obtain a fused vector.

[0129] The index generation unit is used to determine the material flowability index and the material viscosity characterization index based on the fusion vector, which are used to characterize the flowability and viscosity characteristics of the polymer under the current operating conditions, respectively.

[0130] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting material characteristics by integrating spatiotemporal visual features and process parameters, characterized in that, The method includes: Acquire video stream data of the polymer to be tested under controlled operating conditions, and simultaneously collect process parameters; wherein the video stream data and process parameters share the same time scale; Based on a pre-defined dual-stream network, video stream data is spatiotemporally processed to output spatial feature vectors and temporal feature vectors. The dual-stream network includes a spatial stream network and a temporal stream network. The spatial stream network is used to extract the spatial appearance features of the material, including static texture distribution, morphological contours, and bubble morphology. The temporal stream network is used to extract the temporal motion features of the material, including optical flow changes, motion trajectories, local eddy current structures, and flow continuity. Based on the preset material rheological characteristic fusion model, the process parameters, spatial feature vectors and temporal feature vectors are processed to obtain the material flowability index and the material viscosity characterization index. The state identification result of the polymer is determined based on the material flowability index and the material viscosity characterization index.

2. The material feature detection method integrating spatiotemporal visual features and process parameters according to claim 1, characterized in that, The steps of acquiring video stream data of the polymer to be detected under controlled operating conditions and simultaneously acquiring process parameters include: Receive working conditions input by staff, and determine the sensing device based on the working conditions; The process parameters are acquired in real time based on the sensing device, and image frames of the polymer to be detected are acquired based on the preset image acquisition device; the process parameters include at least stirring speed and temperature; Perform temporal registration between process parameters and image frames to determine the process parameters corresponding to each image frame; The image frames are sorted according to time order to obtain video stream data.

3. The material feature detection method integrating spatiotemporal visual features and process parameters according to claim 1, characterized in that, The steps of performing spatiotemporal processing on video stream data based on a preset dual-stream network to output spatial feature vectors and temporal feature vectors include: Receive the processing density input by the staff and determine the time point based on the processing density; Image frames are extracted from the video stream data based on time points, and the image frames are input into the spatial stream network to obtain spatial feature vectors; the extracted image frames include the image frames at the latest time point. Read the extracted image frame and its previous frame, input them into the temporal flow network, and obtain the temporal feature vector; In this system, both the spatial and temporal feature vectors are labeled with the timestamps of the extracted image frames. The spatial flow network is used to extract the spatial appearance features of the polymer during the flow process. These spatial appearance features include, but are not limited to, static texture distribution, morphological contour features, and bubble morphological distribution, which are used to characterize the appearance and structural state of the material at a specific moment. The temporal flow network is used to extract the temporal motion features of the polymer. These temporal motion features include optical flow changes, motion trajectories, local eddy structures, and flow continuity, which are used to characterize the dynamic flow behavior of the material over time.

4. The material feature detection method integrating spatiotemporal visual features and process parameters according to claim 1, characterized in that, The steps for processing process parameters, spatial feature vectors, and temporal feature vectors based on a preset material rheological characteristic fusion model to obtain the material flowability index and the material viscosity characterization index include: Read the specific time points, and read the process parameters, spatial feature vectors, and temporal feature vectors at each time point; The process parameters are normalized to obtain a process parameter vector; The spatial feature vector and the temporal feature vector are corrected based on the process parameter vector; The corrected spatial feature vector and temporal feature vector are fused to obtain the fused vector; The material flowability index and material viscosity characterization index are determined based on the fusion vector, which are used to characterize the flowability and viscosity characteristics of the polymer under current operating conditions, respectively.

5. The material feature detection method integrating spatiotemporal visual features and process parameters according to claim 1, characterized in that, The step of determining the polymer state identification result based on the material flowability index and the material viscosity characterization index includes: The material flowability index and material viscosity characterization index at each time point are compared with the preset state range; The state recognition result at each moment is determined based on the comparison results; Based on the state identification results at each time point, determine the state identification result of the polymer; The acquisition process is recursively adjusted based on the state recognition results at each time point; the recursive adjustment targets include the acquisition frequency of the sensing device, the acquisition frequency of the image acquisition device, and the processing density.

6. The material feature detection method integrating spatiotemporal visual features and process parameters according to claim 1, characterized in that, The method further includes: Establish connection channels with each intelligent module in the experimental scenario; For each testing process, the material flowability index and material viscosity characterization index were read at each moment; Based on the connection channel, all data within a preset time period is acquired to construct a global database; The detection process is clustered based on the material flowability index and material viscosity characterization index at each time point; For each type of detection process, the data is compared with the global database, the intersection is calculated, and the intersection is used as additional environmental conditions and fed back to the detection party.

7. A material feature detection system integrating spatiotemporal visual features and process parameters, characterized in that, The system includes: The process parameter acquisition module is used to acquire video stream data of the polymer to be tested under controlled operating conditions and simultaneously acquire process parameters; wherein, the video stream data and process parameters share the same time scale; The dual-stream recognition module is used to perform spatiotemporal processing on video stream data based on a preset dual-stream network, outputting spatial feature vectors and temporal feature vectors. The dual-stream network includes a spatial stream network and a temporal stream network. The spatial stream network is used to extract the spatial appearance features of the material, including static texture distribution, morphological contours, and bubble morphology. The temporal stream network is used to extract the temporal motion features of the material, including optical flow changes, motion trajectories, local eddy current structures, and flow continuity. The parameter processing module is used to process process parameters, spatial feature vectors and time feature vectors based on a preset material rheological feature fusion model to obtain material flowability index and material viscosity characterization index. The identification result output module is used to determine the state identification result of the polymer based on the material flowability index and the material viscosity characterization index.

8. The material feature detection system integrating spatiotemporal visual features and process parameters according to claim 7, characterized in that, The process parameter acquisition module includes: The sensing device determination unit is used to receive the working conditions input by the staff and determine the sensing device based on the working conditions. An image acquisition unit is used to acquire process parameters in real time based on the sensing device and to acquire image frames of the polymer to be detected based on a preset image acquisition device; the process parameters include at least stirring speed and temperature; The temporal registration unit is used to perform temporal registration of process parameters and image frames to determine the process parameters corresponding to each image frame. The image sorting unit is used to sort image frames based on time order to obtain video stream data.

9. The material feature detection system integrating spatiotemporal visual features and process parameters according to claim 7, characterized in that, The dual-stream identification module includes: The time point determination unit is used to receive the processing density input by the staff and determine the time point based on the processing density. The spatial vector output unit is used to extract image frames from the video stream data based on the time point, input the image frames into the spatial stream network, and obtain spatial feature vectors; the extracted image frames include the image frames at the latest time point. The temporal vector output unit is used to read the extracted image frame and its previous frame, input them into the temporal flow network, and obtain the temporal feature vector. In this system, both the spatial and temporal feature vectors are labeled with the timestamps of the extracted image frames. The spatial flow network is used to extract the spatial appearance features of the polymer during the flow process. These spatial appearance features include, but are not limited to, static texture distribution, morphological contour features, and bubble morphological distribution, which are used to characterize the appearance and structural state of the material at a specific moment. The temporal flow network is used to extract the temporal motion features of the polymer. These temporal motion features include optical flow changes, motion trajectories, local eddy structures, and flow continuity, which are used to characterize the dynamic flow behavior of the material over time.

10. The material feature detection system integrating spatiotemporal visual features and process parameters according to claim 7, characterized in that, The parameter processing module includes: The data reading unit is used to read a specific point in time, and to read the process parameters, spatial feature vector, and temporal feature vector at each point in time. The normalization processing unit is used to normalize the process parameters to obtain a process parameter vector; The vector correction unit is used to correct the spatial feature vector and the temporal feature vector based on the process parameter vector, respectively. The vector fusion unit is used to fuse the corrected spatial feature vector and temporal feature vector to obtain a fused vector. The index generation unit is used to determine the material flowability index and the material viscosity characterization index based on the fusion vector, which are used to characterize the flowability and viscosity characteristics of the polymer under the current operating conditions, respectively.

Images