Embedded-free concrete workability detection method based on deep learning

Abstract

The invention relates to the technical field of concrete construction monitoring, and discloses a non-embedded concrete workability detection method based on deep learning, which comprises the following steps: shooting an image of concrete flowing in a mixer truck chute by using an image acquisition device, analyzing image data, and determining the workability of the concrete in combination with concrete mix proportion data, hydration reaction time and mixer truck chute angle. Therefore, the viscosity value and the slump of the concrete are output. According to the method, a computer vision algorithm, deep learning and machine learning are combined to finish capturing the characteristics of concrete flowing out of the mixer truck step by step, and machine learning is performed based on all captured characteristic values so as to output the concrete slump. An algorithm involved in the scheme supports an edge computing device carrying a network communication module to complete data acquisition and analysis, the communication module is utilized to transmit the acquired and analyzed data to a cloud server, and a commercial mixing station is accurately assisted to make a concrete mix proportion adjustment decision according to different engineering actual conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of concrete construction monitoring, and in particular to a non-embedded concrete workability detection method based on deep learning. Background Art

[0002] In recent years, my country's overall economy has entered a stage of high-quality development, and the economic growth rate has slowed down, which is reflected in the increasingly tense approval of land for construction planning. The era of real estate and municipal engineering to drive investment has passed, profit margins have shrunk, market prices of raw materials have risen, and the necessity of full-cycle quality control of building materials has become increasingly prominent. Among them, the use of commercial concrete has been on the rise, but the current digital scale penetration rate of commercial concrete mixing stations is low, and there is a data gap between construction sites and commercial concrete mixing stations due to production relations. However, if there is a problem with concrete pouring, the accidents caused by it will be huge in terms of personnel safety and indirect or direct economic losses, so the digital connection between commercial concrete mixing and smart construction site management centered on the concrete expert library system is imminent.

[0003] The workability of concrete is a basic physical property for evaluating its workability. Before construction, the workability test of fresh concrete mixture (referred to as fresh concrete) is an essential process in the construction quality control process. Workability refers to the ability of concrete mixture to flow and evenly and densely fill every corner of the formwork under the action of its own weight or the vibration of construction machinery. It reflects the workability of fresh concrete. The size of workability reflects the thinness and consistency of the concrete mixture, which can affect the difficulty of construction compaction and the quality of pouring. At present, the method of testing the workability of concrete on the construction site in the market is still to use slump cone testing, which is cumbersome to operate. In the actual construction process on site, workers and site managers do not test the slump of every truck of concrete because the operation is too complicated, which leads to inaccurate construction record data and inadequate management during the process. In addition, the interpretation of the test results by slump cannot perfectly interpret the physical meaning of the fluid properties of the mixture. Slump can only represent fluid tension. The rheological properties of fresh concrete are characterized by the functional relationship between shear rate and shear stress - rheological model. However, the shear stress and shear rate of fresh concrete cannot be measured directly. Currently, the test results of the concrete rheometer (i.e., torque-speed data) are fitted through the conversion equation of the rheological model, and the rheological curve is established based on the calculated model parameter values ​​to characterize its rheological properties. With the maturity of computer vision and digital image processing technology, it has become possible to identify flowing concrete.

[0004] At the construction site, the workability of concrete is still detected using a slump cone, which is a cumbersome operation. During the actual construction process on site, workers and on-site managers do not conduct slump tests on every truckload of concrete because the operation is too complex. This leads to inaccurate construction record data and inadequate management during the process. Moreover, the quality response of freshly mixed concrete at the construction site is too rough. Most of them use the dry-wet condition of the concrete to reflect the mixing plant. The judgment basis for ready-mixed concrete plants is lacking, resulting in a large gap in the quality of the next batch of freshly mixed concrete compared with the previous batch. Therefore, there is a need for a digital control mechanism for concrete quality between the construction site and the commercial concrete mixing plant. The related research inventions are as follows:

[0005] In the Japanese magazine "Soken Report" (ISSN 1881 - 8765) in December 2020, the results of a questionnaire survey on the impact of the novel coronavirus infection on domestic investment plans (surveys in March and June 2020). Taisei Corporation (President: Yoshiyuki Murata) in the *1 system. By applying this system, various information required for in-situ concrete projects can be digitized and centrally managed, thus improving productivity and quality. " / Concrete" digitizes the information of ready-mixed concrete from kneading in the ready-mixed concrete plant to the completion of pouring. All construction can view this information in real time on a WEB server on the Internet. It is a system that can share information among relevant parties, support shipping management, transportation management, acceptance management, placement management, and quality control, aiming to improve the productivity and quality of in-situ concrete work.

[0006] Zhongshan Aishang Zhitong Information Technology Co., Ltd. publicly invented "A Method and System for Intelligent Mixing Control of Concrete" (Publication No.: CN109676795A). This invention provides a method and system for intelligent mixing control of concrete, realizing the automatic detection of the rheology and homogeneity of products during the concrete mixing process and the intelligent regulation function of the production process. The method is as follows: perform big data analysis on the monitored mixing data and the physical and chemical property data of the completed concrete to generate the optimal mix ratio and the best mixing plan under various usage requirements; the mixer automatically calls the optimal material mix ratio and the best mixing plan according to the concrete use or specific parameter performance requirements, automatically takes materials and measures the performance parameters of the materials, and feeds the performance parameters back to the edge computing platform to judge whether the material performance meets the requirements; after mixing is completed, compare the measured physical and chemical property data of the concrete with the expected data to judge whether to discharge the material or make corrections; the system includes a cloud platform and a mixing terminal, and the mixing terminal includes a human-computer interaction module, a control system, a mixing system, a monitoring system, and an edge computer. Summary of the Invention

[0007] 1. Technical issues to be resolved

[0008] In view of the shortcomings of the prior art, the present invention provides a non-embedded concrete workability detection method based on deep learning, which can realize non-contact and accurate auxiliary commercial concrete mixing stations to make adjustment decisions on concrete mix ratio according to the actual conditions of different projects.

[0009] (II) Technical solution

[0010] To achieve the above object, the present invention provides the following technical solution: a non-embedded deep learning-based concrete workability detection method, comprising the following steps:

[0011] S1. Use the automatic weighing system to record the concrete mix ratio data in each mixing silo of the commercial concrete mixing station, and record the departure time of the mixer truck that transports the corresponding batch of concrete on site;

[0012] S2. Arrange multiple on-site construction monitoring cameras at the construction site in advance. After the mixer truck enters the construction site and before pouring begins, the construction monitoring camera collects images of the construction site conditions, uses the yolov7 deep learning target detection model to identify the on-site concrete mixer truck and pump truck, and locates the target area;

[0013] S3, according to the positioning in step S2, notify the on-site technician to set up an image acquisition device on the mixer truck to acquire images of the concrete in the mixer truck chute. When the mixer truck starts pouring, the time when the mixer truck starts pouring is recorded, and the time when the corresponding vehicle departs from the commercial concrete mixing station is subtracted from the time to obtain the hydration reaction time of the freshly mixed concrete in the mixer truck;

[0014] S4. Using an image acquisition device, take an image of concrete flowing in the chute, use the GMFlow framework to segment the chute structure and the flowing concrete, continuously extract multiple frames of images, extract the ROI area of ​​the DIC detection algorithm from the continuous frame images, and then use the feature point capture method of the DIC detection without punctuation method to capture the concrete feature points in the ROI area, so as to obtain a Lagrangian flow field image or an Euler flow field image of the concrete flow, and mark the displacement vector information on the flow field image;

[0015] S5. Determine the angle of the concrete mixer truck chute using the SIFT algorithm;

[0016] S6. Use the KNN network for machine learning, input the vector displacement information data of each pixel of the flowing concrete as input, fit the concrete mix ratio data, hydration reaction time and the chute angle of the mixer truck to output the concrete viscosity value and slump, feed back the workability of the on-site concrete to the background database, improve the mix ratio based on the data in the background database, and the commercial concrete mixing station mixes the fresh concrete according to the improved mix ratio in the background database and delivers it to the construction site. Repeat this process so that the working performance of the concrete meets the requirements.

[0017] Preferably, the use of the yolov7 deep learning target detection model to identify on-site concrete mixer trucks and pump trucks and locate the target area specifically includes: judging the relative positions of the mixer truck, mixer truck chute and pump truck in the image, and using the material target detection method based on rotating frame positioning to judge the positions of the concrete mixer trucks and pump trucks at different angles on site, understand the pouring area, and judge which mixer truck needs to pour concrete.

[0018] Preferably, the image acquisition device acquires images through a camera, and when the device is set up, the camera is placed perpendicular to the unloading chute of the mixer truck.

[0019] Preferably, the GMFlow framework consists of three main parts:

[0020] A. Transformer for feature enhancement;

[0021] B. Correlation and softmax layers for global matching;

[0022] C. Self-attention layer for optical flow propagation.

[0023] Preferably, the feature point capture method using DIC detection without punctuation method to capture concrete feature points specifically includes:

[0024] The results of optical flow estimation are used to segment the area of ​​interest of concrete flow separately, and the key points of the flowing concrete are extracted. By obtaining a subset of reference image pixels, a one-to-one correspondence is obtained between the material points in the reference (initial undeformed image) and current (subsequent deformed image) configurations, and a grid containing displacement and strain information related to the reference configuration is obtained, which is divided into Eulerian grid and Lagrangian grid.

[0025] Preferably, the use of the sift algorithm to determine the angle of the concrete mixer truck chute includes scale space extreme value detection, key point positioning, direction determination and key point description, setting a chute reference object at a fixed place on site, and calculating the inclination angle of the real chute by comparing the reference object with the real chute.

[0026] Preferably, in step S6, the relevant data collected from S1 to S5 can be used to collect the real viscosity and slump of the measured flow rate concrete, prepare a training set, and use the KNN network to perform machine learning on the displacement vector information of the concrete flow pixel set, the main factors affecting viscosity in the mix ratio parameters, the hydration reaction time, and the inclination angle of the mixer truck chute as inputs to output the concrete viscosity and slump.

[0027] Preferably, the image acquisition device comprises:

[0028] The data acquisition and processing unit includes a camera, a central processing unit, a network communication module and an edge computing module. The image data collected by the camera is transmitted to the central processing unit and processed by the edge computing module. The network communication module transmits the processing structure to the cloud server.

[0029] Digital display controller, used to display and control image acquisition equipment;

[0030] A power supply unit, which is a lithium battery, is used to provide power to the image acquisition device;

[0031] The external device unit includes a fixing clip for adjusting the camera angle and an aluminum alloy shell for carrying the entire device.

[0032] Another technical problem to be solved by the present invention is to provide a computable and readable storage medium, which implements the above-mentioned monitoring method when the program is executed by a processor.

[0033] (III) Beneficial effects

[0034] Compared with the prior art, the present invention provides a non-embedded concrete workability detection method based on deep learning, which has the following beneficial effects:

[0035] This method combines computer vision algorithms, deep learning, and machine learning to capture the characteristics of concrete flowing out of a mixer truck in steps, and outputs the slump of concrete based on machine learning of all captured feature values. The algorithm involved in this solution supports edge computing devices equipped with network communication modules to complete data collection and analysis, and uses the communication module to transmit the collected and analyzed data to the cloud server, accurately assisting commercial concrete mixing stations to make adjustments to the concrete mix ratio based on the actual conditions of different projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a schematic diagram for demonstrating the process of the present invention;

[0037] Figure 2 It is a schematic diagram of the process of the present invention;

[0038] Figure 3 It is a structural schematic diagram of the algorithm training method of the present invention;

[0039] Figure 4 It is a flowchart of the algorithm training method of the present invention;

[0040] Figure 5 A schematic diagram of system control of an image acquisition device of the present invention;

[0041] Figure 6 It is a schematic diagram of on-site installation of the image acquisition device of the present invention;

[0042] Figure 7 This is an effect diagram of the flowable concrete identification of the present invention. DETAILED DESCRIPTION

[0043] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] See also Figure 1-4 The present invention proposes a non-embedded deep learning-based concrete workability detection method, which uses image acquisition equipment to capture images of concrete flowing from the chute of a mixer truck, analyzes the image data, and combines the concrete mix ratio data, hydration reaction time, and the chute angle of the mixer truck to output the concrete viscosity value and slump.

[0045] Specifically, the acquisition of concrete mix ratio data and hydration reaction time can utilize the automatic weighing system to record the concrete mix ratio data in each mixing silo of the commercial concrete mixing station, utilize the license plate recognition to confirm the departure time of the mixer truck that transports the corresponding batch of concrete on site, record the departure time of the mixer truck that transports the corresponding batch of concrete on site, and then pre-arrange multiple on-site construction monitoring cameras at the construction site. After the mixer truck enters the construction site and before pouring begins, the construction monitoring camera collects images of the construction site conditions, uses the yolov7 deep learning target detection model to identify the on-site concrete mixer truck and pump truck, and locates the target area. According to the positioning, the on-site technicians are notified to set up image acquisition equipment on the mixer truck to collect images of the concrete in the mixer truck chute. When the mixer truck starts pouring, the time when the mixer truck starts pouring is recorded, and the time when the corresponding vehicle departs from the commercial concrete mixing station is subtracted from this time to obtain the hydration reaction time of the freshly mixed concrete in the mixer truck.

[0046] Among them, the yolov7 deep learning target detection model is used to identify the on-site concrete mixer trucks and pump trucks, and locate the target area, including: judging the relative positions of the mixer truck, mixer truck chute and pump truck in the image, and using the material target detection method based on rotating frame positioning to judge the locations of concrete mixer trucks and pump trucks at different angles on site, understand the pouring area, and determine which mixer truck needs to pour concrete.

[0047] See also Figure 5-6 The image acquisition device collects images through a camera, which is a depth camera. When the equipment is set up, the camera is placed perpendicular to the unloading chute of the mixer truck. It should be noted that the camera illumination angle for shooting the flowing concrete in the chute of the mixer truck must be a fixed angle (preferably perpendicular to the ground or perpendicular to the chute), and the coordinate points of the four corners of the projection of the chute as a whole in the camera must be fixed, which paves the way for the subsequent machine learning data analysis of the information of the vectorized image.

[0048] The image acquisition device includes a data acquisition and processing unit, a digital display controller, a power supply unit and an external device unit, wherein the data acquisition and processing unit includes a camera, a central processing unit, a network communication module and an edge computing module. The image data collected by the camera is transmitted to the central processing unit and processed by the edge computing module. The network communication module transmits the processing structure to the cloud server; the digital display controller is used to display and control the image acquisition device; the power supply unit is a lithium battery, which is used to provide power to the image acquisition device; the external device unit includes a fixed clamp with adjustable camera angle, the fixed clamp is arranged on the funnel of the concrete mixer truck, and the aluminum alloy shell is used to carry the entire device.

[0049] An image acquisition device is used to capture images of concrete flowing in the chute. The GMFlow framework is used to segment the chute structure and the flowing concrete. Multiple frames of images are continuously extracted, and the ROI area of ​​the DIC detection algorithm is extracted from the continuous frame images. Then, the feature point capture method of the DIC detection without punctuation is used to capture the concrete feature points in the ROI area, thereby obtaining the Lagrangian flow field image or Euler flow field image of the concrete flow, and marking the displacement vector information on the flow field image.

[0050] The GMFlow framework segments the chute structure and flowing concrete. Its purpose is to extract the results of semantic segmentation and use them as the ROI region of interest for DIC detection. Different from the traditional semantic segmentation model that classifies and recognizes a single image, the GMFlow framework consists of three main parts: Transformer for feature enhancement, correlation and softmax layers for global matching, and a self-attention layer for optical flow propagation. Since the optical flow features of flowing concrete are relatively prominent, the optical flow estimation method is used to segment and recognize the flowing fresh concrete.

[0051] Among them, it is configured to use the semantic segmentation results to select the ROI region of interest for DIC detection, and to infer the pixel motion vector information of the flowing concrete based on the pixel interpolation of the ROI region of interest of the previous and next frames, so as to obtain the Euler motion field.

[0052] The results of optical flow estimation are used to segment the area of ​​interest of concrete flow separately, and the key points of the flowing concrete are extracted. By obtaining a subset of reference image pixels, a one-to-one correspondence is obtained between the material points in the reference (initial undeformed image) and current (subsequent deformed image) configurations, and a grid containing displacement and strain information related to the reference configuration is obtained, which is divided into Eulerian grid and Lagrangian grid.

[0053] Specifically, the changing curve of the concrete's movement state in the chute is used to identify the flow characteristics of different slumps at the same chute angle, record the displacement vector information, form a unique Euler vector pixel field for the flowing concrete, and use a fresh concrete viscometer to record the viscosity of the truck's concrete to do a good job of data collection.

[0054] The SIFT algorithm is used to determine the angle of the concrete mixer truck chute, including scale space extreme value detection, key point positioning, direction determination and key point description. A chute reference is set at a fixed place on site, and the inclination angle of the real chute is calculated by comparing the reference object with the real chute.

[0055] The KNN network is used for machine learning. The vector displacement information data of each pixel of the flowing concrete is input as input. The concrete mix ratio data, hydration reaction time and the chute angle of the mixer truck are fitted to output the concrete viscosity value and slump. The workability of the on-site concrete is fed back to the background database, the mix ratio is improved according to the data in the background database, and the Internet of Things technology is used to notify the mixing station to adjust the concrete mix ratio. The commercial mixing station mixes the fresh concrete according to the improved mix ratio in the background database and delivers it to the construction site. This process is repeated to ensure that the working performance of the concrete meets the requirements.

[0056] See also Figure 7 , Figure 7This is the effect diagram of the flowing concrete identification of the present invention. It needs to be further explained that the movement state of the flowing concrete for 10 consecutive frames shows a certain pattern, and the pattern is the characterization of the rheological properties of the concrete. The motion vector information can be used for machine learning to obtain its fixed rheological parameters, so that concrete with different rheological characteristics can be observed and recorded, and included in the cloud database. As the data accumulates, the cloud algorithm model becomes more accurate, and the upgraded content can be remotely delivered to the edge.

[0057] Since the concrete mix design is determined based on the concrete configuration strength, concrete design strength and concrete strength standard deviation, and the values ​​that have an absolute impact on the workability of fresh concrete are water-cement ratio, cement type, admixture parameters and types, sand ratio, aggregate particle size and its proportion, the above method is used to collect relevant data, collect the real viscosity and slump of the measured flow rate concrete, make a training set, and use the KNN network to perform machine learning on the displacement vector information of the concrete flow pixel set, the main factors affecting viscosity in the mix parameters, hydration reaction time and the inclination angle of the mixer truck chute as inputs, and output concrete viscosity and slump. Among them, this data set uses concrete mix ratio data, hydration reaction time data in the mixer truck, and concrete flow rate vector data as inputs, outputs concrete viscosity, and uses the KNN network to perform machine learning on the data, with the focus on the accuracy of the concrete mixing station mix ratio data, and the accurate recording of the mixing time in the mixing silo of the commercial mixing station plus the transportation time in the mixer truck during transportation.

[0058] In order to establish a weighted relationship between the change in viscosity of freshly mixed concrete corresponding to different gradations and the change in the amount of materials used, it is determined whether the amount of materials used in the concrete mixing process needs to be adjusted according to the predicted value of the concrete's working performance in the mixing image, and the amount of materials used in the concrete mixing process is adjusted based on the predicted value of the concrete's working performance, the gradation of aggregates in the concrete mixing process, and the weighted relationship. The above operations are repeated to infer the viscosity of the concrete until the working performance of the concrete meets the requirements.

[0059] This method combines computer vision algorithms, deep learning, and machine learning to capture the characteristics of concrete flowing out of a mixer truck in steps, and outputs the slump of concrete based on machine learning of all captured feature values. The algorithm involved in this solution supports edge computing devices equipped with network communication modules to complete data collection and analysis, and uses the communication module to transmit the collected and analyzed data to the cloud server, accurately assisting commercial concrete mixing stations to make adjustments to the concrete mix ratio based on the actual conditions of different projects.

[0060] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0061] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0062] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0063] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0064] The foregoing description of specific exemplary embodiments of the present invention is for the purpose of illustration and demonstration. These descriptions are not intended to limit the present invention to the precise form disclosed, and it is clear that many changes and variations can be made based on the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical application, so that those skilled in the art can realize and utilize various different exemplary embodiments of the present invention and various different selections and changes. The scope of the present invention is intended to be limited by the claims and their equivalents.

Claims

1. A non-embedded deep learning-based concrete workability detection method, characterized in that: The following steps are involved: S1. Use the automatic weighing system to record the concrete mix ratio data in each mixing silo of the commercial concrete mixing station, and record the departure time of the mixer truck that transports the corresponding batch of concrete on site; S2. Arrange multiple on-site construction monitoring cameras at the construction site in advance. After the mixer truck enters the construction site and before pouring begins, the construction monitoring camera collects images of the construction site conditions, uses the yolov7 deep learning target detection model to identify the on-site concrete mixer truck and pump truck, and locates the target area; S3, according to the positioning in step S2, notify the on-site technician to set up an image acquisition device on the mixer truck to acquire images of the concrete in the mixer truck chute. When the mixer truck starts pouring, the time when the mixer truck starts pouring is recorded, and the time when the corresponding vehicle departs from the commercial concrete mixing station is subtracted from the time to obtain the hydration reaction time of the freshly mixed concrete in the mixer truck; S4. Using an image acquisition device, take an image of concrete flowing in the chute, use the GMFlow framework to segment the chute structure and the flowing concrete, continuously extract multiple frames of images, extract the ROI area of ​​the DIC detection algorithm from the continuous frame images, and then use the feature point capture method of the DIC detection without punctuation method to capture the concrete feature points in the ROI area, so as to obtain a Lagrangian flow field image or an Euler flow field image of the concrete flow, and mark the displacement vector information on the flow field image; S5. Determine the angle of the concrete mixer truck chute using the SIFT algorithm; S6. Use the KNN network for machine learning, input the vector displacement information data of each pixel of the flowing concrete as input, fit the concrete mix ratio data, hydration reaction time and the chute angle of the mixer truck to output the concrete viscosity value and slump, feed back the workability of the on-site concrete to the background database, improve the mix ratio based on the data in the background database, and the commercial concrete mixing station mixes the fresh concrete according to the improved mix ratio in the background database and delivers it to the construction site. Repeat this process so that the working performance of the concrete meets the requirements.

2. According to claim 1, a non-embedded deep learning-based concrete workability detection method is characterized in that: In the step S2, the target detection model of yolov7 deep learning is used to identify the on-site concrete mixer truck and pump truck, and the target area is located, which specifically includes: judging the relative positions of the mixer truck, mixer truck chute and pump truck in the image, and using the material target detection method based on rotating frame positioning to judge the positions of the concrete mixer truck and pump truck at different angles on site, understand the pouring area, and judge which mixer truck needs to pour concrete.

3. The method for detecting concrete workability without embedded deep learning according to claim 1, characterized in that: The image acquisition device in step S3 acquires images through a camera, and when the device is set up, the camera is placed perpendicular to the unloading chute of the mixer truck.

4. The method for detecting concrete workability without embedded deep learning according to claim 1, characterized in that: The GMFlow framework in step S4 consists of three main parts: A. Transformer for feature enhancement; B. Correlation and softmax layers for global matching; C. Self-attention layer for optical flow propagation.

5. The method for detecting concrete workability without embedded deep learning according to claim 4 is characterized in that: The method for capturing concrete feature points using the feature point capture method without punctuation method using DIC detection in step S4 specifically includes: The results of optical flow estimation are used to segment the area of ​​interest of concrete flow separately, and the key points of the flowing concrete are extracted. By obtaining a subset of reference image pixels, a one-to-one correspondence is obtained between the material points in the reference (initial undeformed image) and current (subsequent deformed image) configurations, and a grid containing displacement and strain information related to the reference configuration is obtained, which is divided into Eulerian grid and Lagrangian grid.

6. The method for detecting concrete workability without embedded deep learning according to claim 1, characterized in that: In step S5, the angle of the concrete mixer truck chute is determined by using the SIFT algorithm, including scale space extreme value detection, key point positioning, direction determination and key point description. A chute reference is set at a fixed place on site, and the inclination angle of the real chute is calculated by comparing the reference object with the real chute.

7. The method for detecting concrete workability without embedded deep learning according to claim 1, characterized in that: In the step S6, the relevant data collected from S1 to S5 can be used to collect the real viscosity and slump of the measured flow rate concrete, and a training set can be prepared. The KNN network is used to perform machine learning on the displacement vector information of the concrete flow pixel set, the main factors affecting viscosity in the mix ratio parameters, the hydration reaction time, and the inclination angle of the mixer truck chute as inputs to output the concrete viscosity and slump.

8. The method for detecting concrete workability without embedded deep learning according to claim 1, characterized in that: The image acquisition device comprises: The data acquisition and processing unit includes a camera, a central processing unit, a network communication module and an edge computing module. The image data collected by the camera is transmitted to the central processing unit and processed by the edge computing module. The network communication module transmits the processing structure to the cloud server. Digital display controller, used to display and control image acquisition equipment; A power supply unit, which is a lithium battery, is used to provide power to the image acquisition device; The external device unit includes a fixing clip for adjusting the camera angle and an aluminum alloy shell for carrying the entire device.

9. A computable and readable storage medium, characterized in that: When the program is executed by a processor, the monitoring method as described in any one of claims 1 to 7 is implemented.

Images