A slump visual assessment system for concrete mixing

Abstract

This invention discloses a slump visual evaluation system for concrete mixing, relating to the field of concrete production process quality inspection and intelligent control technology. The invention uses the departure of the target mixing component from the target observation area as the starting point of the candidate observation segment. It establishes a backfilling reference area based on the mask covering the target mixing component and the disconnected area of ​​the material surface. It then combines the remaining area ratio to be backfilled, the boundary openness rate, the proportion of major connected components, and the disconnected and connected path of the material surface to generate the backfill completion status. Image frames where the mixing component has left but the swept area still has unfilled gaps, open boundaries, or disconnected material surfaces do not trigger an evaluation window, ensuring that the window's starting position corresponds to the actual backfilling process of the concrete surface.

Description

Technical Field

[0001] This invention relates to the field of quality inspection and intelligent control technology in concrete production process, and in particular to a visual assessment system for slump of concrete mixing. Background Technology

[0002] Concrete slump is a crucial indicator of the fluidity and workability of freshly mixed concrete. Current testing methods typically involve sampling concrete after it has exited the mixer, using a slump cone for loading, lifting, and height measurement. This method requires independent sampling and operation, and by the time the test results are generated, the concrete has already been mixed, making it difficult to continuously reflect the state changes of the same batch of concrete at the end of the mixing process. With the development of automated control and production data recording in mixing plants, existing technologies have incorporated image acquisition devices at observation points on slump test benches, unloading hoppers, unloading troughs, or mixing containers. These devices determine slump values, slump grades, or workability categories using contour, texture, optical flow, depth information, or learning models. For example, invention patent CN103163136B discloses a method for determining concrete workability using images of the mixing process and concrete shape characteristics.

[0003] For visual inspection of the interior of a mixing vessel, patent application CN114266989A discloses a processing method that extracts images from monitoring videos of multiple mixing cycles and combines visual features with a classifier to determine the slump level. It is evident that images of the mixing process can be used to determine workability before the product leaves the machine; however, the specific material response stage of the image frame still affects the evaluation results.

[0004] At the wet mixing end of a twin-shaft mixer, the mixing blades and arms periodically enter the image acquisition area. The material spillage, grooves, and splashes generated by the blade sweeping alternate with the concrete surface. After the mixing components leave the image acquisition area, localized material undergoes processes such as backfilling of gaps, collapse of ridges, attenuation of striations, and re-exposure or re-covering of coarse aggregates by the slurry. Residual slurry on the machine wall, adhered lumps, and splashed slurry points may remain in the field of view for extended periods. Therefore, the disappearance of the mixing components from the image only indicates the end of direct occlusion and does not directly suggest that the concrete surface has reached a comparable state.

[0005] Existing processing methods, such as fixed-time frame extraction, multi-frame averaging, fixed regions of interest, and retaining or deleting image frames based on blade visibility, are ill-suited to adapting to variations in backfilling time caused by changes in rotational speed, loading capacity, mix proportions, and actual slump. Images fixed at the same cycle position may be in stages of strong disturbance, backfilling transition, or surface stabilization; the overall video characteristics may also be dominated by a small number of high-amplitude material-turning or splashing frames, causing abrupt changes in slump results obtained from adjacent mixing cycles.

[0006] Therefore, visual assessment of the wet mixing end of a twin-shaft mixer still needs to address the following issues: how to combine the material backfilling state and surface disturbance decay state after the mixing components leave the image acquisition area to determine the stable observation period that can be used for slump evaluation in each mixing cycle, and how to ensure that the slump results are generated only from the images within this stable observation period, so that the output results are correlated with the corresponding mixing cycle and observation state, and provide continuous state input for water addition, extended mixing and material discharge control. Summary of the Invention

[0007] This invention provides a visual assessment system for slump in concrete mixing, which solves the problem of slump assessment occurring before the material surface has stabilized after the mixing component has left the container.

[0008] This invention provides a visual assessment system for slump in concrete mixing, comprising: The video acquisition unit collects video of the wet mixing end of the twin-shaft mixer and determines the target observation area; The candidate observation segment determination unit uses the image frames between the time the target stirring component leaves the target observation area and the time when any stirring component first enters the area as candidate observation segments. The material surface state determination unit determines the concrete body area within the candidate observation section and generates a reference area to be backfilled based on the covering mask and the material surface disconnection area of ​​the component. When the remaining area ratio to be backfilled is not higher than the backfill threshold, the boundary open rate is not higher than the boundary closed threshold, the proportion of the main connected components is not lower than the connectivity recovery threshold, and no material surface disconnection through path is formed, and the above four conditions are simultaneously met in consecutive effective image frames to reach the backfill holding frame number, a backfill completion state is generated, and a disturbance attenuation state is generated based on the decrease and stability of the material surface movement amplitude, material surface contour displacement, and surface structure change. The evaluable window generation unit collects valid image frames starting from the first valid image frame where both states are simultaneously established. It terminates collection when either state fails or an entry event occurs. An evaluable window is generated when the number of valid image frames reaches the window length threshold. The slump visual assessment unit generates slump visual assessment results based on the concrete body area within the evaluable window. The result output unit outputs the evaluation result.

[0009] In some embodiments, the stirring component includes stirring blades and stirring arms, and the stirring component is identified as a single stirring blade, a single stirring arm, or a combination of stirring blades and stirring arms that move in and out synchronously. The candidate observation segment determination unit compares the motion area with the historical component trajectory, identifies the stirring component based on its contour, direction and repeated trajectory, divides the component cycle according to two adjacent entry events of the same stirring component, associates the candidate observation segment with the component cycle of the target stirring component, and records the stirring component identifier and the position of the two event frames.

[0010] In some embodiments, the material surface state determination unit determines the concrete candidate area that is connected to the preset material surface reference area and moves across frames as the concrete body area, and merges the machine wall residual slurry area, the splash slurry point area and the local void area outside the reference area to be backfilled into an interference area mask. Low-resolution areas with fixed boundaries that cover the image of the mixing component across multiple component cycles are identified as lens contamination areas. When a new material area intersects with the feeding port area, moves towards the concrete body area, and its area reaches the feeding area threshold and remains at the feeding hold frame count, a feeding event marker is generated. When the overlap area ratio of the interference area reaches the interference threshold, the visible area ratio of the body is lower than the visible threshold, or the area ratio of the lens contamination area reaches the contamination threshold, it is marked as an invalid frame, and the remaining image frames are valid image frames.

[0011] In some embodiments, adjacent valid image frames are two adjacent frames in a valid image frame sequence; The material surface state determination unit registers adjacent valid image frames, takes the intersection of the concrete body regions of the two frames as the common body region, calculates the material surface movement amplitude and surface structure change within the common body region, calculates the material surface contour displacement based on the set of valid material surface contour points of the two frames, and forms a corresponding sequence. The amplitude of the material surface motion is calculated by the optical flow displacement modulus, and the material surface contour displacement is calculated by the bidirectional nearest distance. Both are normalized according to the inter-frame time and the diagonal pixel length of the target observation area. The change in surface structure is obtained by weighting the difference in local texture description, the change in edge density, and the change in the proportion of textured area.

[0012] In some embodiments, the material surface state determination unit smoothly processes the three change sequence; When the decrease ratio of the material surface motion amplitude and the decrease ratio of the material surface contour displacement within the downward trend window reach the corresponding thresholds, and the total decrease of the two changes reaches the minimum total decrease, the downward trend is determined to be established; when the three smooth changes do not exceed the corresponding disturbance thresholds within the disturbance holding frame number, a disturbance attenuation state is generated; when the same smooth change exceeds the corresponding recovery threshold in two consecutive valid image frames, or when the feeding event is marked, an entry event occurs, or the growth rate of the splash slurry point area reaches the sudden increase threshold, the disturbance attenuation state fails.

[0013] In some embodiments, the evaluation window generation unit can determine the window start frame when the frame that is first established as a valid image frame simultaneously in the backfill completion state and the disturbance attenuation state. When an entry event occurs, a material feeding event is marked, the number of consecutive invalid frames reaches the corresponding threshold, or any of these states fail, the last valid image frame before the termination condition is met is used as the window termination frame. The valid image frames between the start and end frames are combined into a window valid frame sequence. When the sequence length reaches the window length threshold, an evaluable window is generated. If it does not reach the threshold, the reason for the window cancellation is recorded. When no evaluable window is generated for a preset number of consecutive candidate observation segments of the same batch of concrete, the result output unit outputs an unevaluable state.

[0014] In some embodiments, the slump visual evaluation unit extracts the proportion of textured areas, continuous coverage of slurry, exposure rate of coarse aggregate, and local material surface contour undulation from the effective frame sequence of the window to form a surface state feature sequence, and generates a slump grade or slump range according to the slump calibration relationship. The slump calibration relationship is established using a historical evaluable window with standard slump test results and is associated with mixer identification, camera calibration version, and mix category.

[0015] In some embodiments, the slump calibration relationship includes the interval feature center corresponding to each slump interval; The slump visual assessment unit extracts the mean, end value and slope of change per unit time from the surface state feature sequence to form a window feature vector. It calculates the weighted distance between the window feature vector and the feature center of each interval and takes the slump interval with the smallest weighted distance as the single-cycle assessment result. The distance weight is determined by the historical variance of the corresponding feature within the same collapse interval.

[0016] In some embodiments, the slump visual assessment unit obtains a preset number of recently generated single-cycle assessment results, with the difference between the maximum and minimum values ​​of the slump interval number not exceeding 1 as a consistency condition. The average overlap area ratio of the interference region in the effective frame sequence of the window is used as the overlap area ratio of the interference region of the window. The product of the number of effective image frames and one minus the overlap area ratio of the interference region of the window is used as the fusion weight. The fusion weight of the cumulative collapse interval is used as the collapse interval with the largest cumulative fusion weight as the visual evaluation result of collapse. If the consistency condition is not met, all fusion weights are 0, or there are tied maximum values, output an unevaluable state and generate a result conflict flag.

[0017] In some embodiments, a closed-loop interface unit connected to the mixer controller via a data interface is also included; The output unit generates associated records, which include the stirring cycle identifier, the start and end frame positions of the window, the generation positions of the two states, the invalid frame marker, the reason for window cancellation, and the result conflict marker. When the slump visual evaluation result is within the target slump range, a signal to allow material release is sent; when it is below the lower limit of the range, a signal to continue stirring is sent; if it is still below the lower limit after the additional stirring time expires, a signal to add water for verification is sent; when it is above the upper limit of the range, a signal to continue observation is sent; and when an unevaluable state is received, a signal to send an abnormal acquisition signal is sent.

[0018] Through the above technical solution, the present invention can achieve at least the following beneficial effects: This invention uses the departure of the target mixing component from the target observation area as the starting point of the candidate observation segment. A reference area to be backfilled is established based on the mask covering the target mixing component and the disconnected area of ​​the material surface. The backfill completion status is generated by combining the remaining area ratio to be backfilled, the boundary openness rate, the proportion of major connected components, and the disconnected and connected paths of the material surface. Image frames where the mixing component has left but the swept area still has unfilled gaps, open boundaries, or disconnected material surfaces do not trigger an evaluation window, ensuring that the window's starting position corresponds to the actual backfilling process of the concrete surface.

[0019] After the backfill completion state is established, the inertial flow, contour undulation, and surface structure changes of the material surface are characterized by the amplitude of material surface movement, the displacement of material surface profile, and the amount of surface structure change. A disturbance decay state is generated based on the descent and stabilization processes. The evaluation window is only formed when both the backfill completion state and the disturbance decay state are established simultaneously. This ensures that the evaluation input for different stirring cycles corresponds to the material state stage where backfilling is complete and disturbance has decayed, avoiding directly using image differences from different material response stages as slump difference input.

[0020] Local voids within the backfill reference area are used to determine the backfill status. Local empty areas, residual slurry areas on the machine wall, and splashed slurry areas outside the backfill reference area are used to generate interference area masks. Lens contamination areas are identified separately based on their image performance of covering moving objects across cycles and having fixed boundaries. This distinguishes between material surface recovery information, fixed scene attachments, and lens attachments, giving the image areas involved in status calculation and slump assessment a clear physical origin.

[0021] The evaluable windows for different mixing cycles are time-aligned according to the first simultaneous occurrence of the backfill completion state and the disturbance decay state, and the fusion weight is generated by the number of effective image frames in the window and the overlap area ratio of the window's interference region. When the interval span of the single-cycle evaluation result exceeds the consistency range, all fusion weights are zero, or the cumulative fusion weights of multiple intervals are tied for the largest, an unevaluable state is output, so that the closed-loop interface generates control signals only based on the collapse visual evaluation results with a definite observation state and a unique fusion result. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope of the present invention.

[0023] Figure 1 This is a framework diagram of the slump visual assessment system for concrete mixing in the embodiment. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the 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 and not intended to limit the invention.

[0025] All terms used in this invention (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined.

[0026] Example 1:

[0027] like Figure 1 As shown, this embodiment employs a slump visual evaluation system for concrete mixing, including a video acquisition unit, a candidate observation segment determination unit, a material surface state determination unit, an evaluable window generation unit, a slump visual evaluation unit, and a result output unit.

[0028] The video acquisition unit captures video of the wet mixing end of the twin-shaft mixer and determines the target observation area. The mixing components include the mixing blades and the mixing arms. The candidate observation segment determination unit identifies the departure event of the mixing component and the subsequent first entry event, and determines the image frame between the two events as the candidate observation segment.

[0029] The material surface state determination unit identifies the concrete body region within the candidate observation segment and generates a reference region to be backfilled based on the component overlay mask and the material surface disconnection area of ​​the departure event. When the remaining area ratio and boundary openness ratio to be backfilled decrease to the corresponding threshold, the proportion of the main connected components reaches the corresponding threshold, no material surface disconnection through path is formed, and the backfill holding frame number is continuously maintained, a backfill completion state is generated. The material surface state determination unit also generates a disturbance attenuation state based on the cross-frame surface changes of the concrete body region.

[0030] The evaluable window generation unit uses the image frame in which both the backfill completion state and the disturbance attenuation state are simultaneously established for the first time as the starting frame, and the earlier of the failure and entry events in either state as the termination event, defining the image frame between the starting and ending frames as the evaluable window. The slump visual evaluation unit generates the slump visual evaluation result based on the concrete body area within the evaluable window. The result output unit outputs the slump visual evaluation result, and outputs an unevaluable state if no evaluable window is generated within a set number of candidate observation segments.

[0031] The wet mixing end stage refers to the mixing phase from the completion of the last water and admixture addition by the twin-shaft mixer until the discharge command is issued. When the mixer controller can provide an addition end signal, the time corresponding to the last addition end signal is taken as the start time of the wet mixing end stage; when no addition end signal is obtained, the time obtained by subtracting the preset end observation time from the preset total wet mixing time is taken as the start time of the wet mixing end stage.

[0032] In one embodiment, the video acquisition unit acquires video of the wet mixing end using an image acquisition device fixed to the observation port of the twin-shaft mixer, and determines the target observation area based on the fixed boundary of the camera coordinate system. The candidate observation segment determination unit tracks the motion region within consecutive image frames, identifies the mixing component based on its contour shape, motion direction, and repeating trajectory, divides the mixing cycle by two adjacent entry events of the same mixing component, associates the mixing cycle identifier to which the departure event belongs, and records the mixing component identifier, the departure event frame position, and the entry event frame position.

[0033] In one implementation, the material surface state determination unit determines the validity of the image frame according to the following process: (1) The preset material surface reference area is a fixed area continuously covered by freshly mixed concrete when there is no obstruction in the target observation area; the preset evaluation area is a fixed sub-area used for image frame validity judgment.

[0034] (2) The target observation area is segmented. The concrete candidate area connected to the preset material surface reference area and moving with the material surface across frames is taken as the concrete body area; the slurry area fixed at the camera coordinate position is taken as the machine wall residual slurry area; the discrete area separated from the concrete body area and moving with the projectile trajectory is taken as the splash slurry point area; the background area inside the machine that is surrounded by the concrete body area and located outside the current reference area to be backfilled is taken as the local void area. The machine wall residual slurry area, splash slurry point area and local void area are merged into an interference area mask.

[0035] The background area inside the machine located within the current reference area to be backfilled is not written into the interference area mask, and is handled by the remaining area to be backfilled, the boundary open rate, and the material surface disconnection and penetration path.

[0036] (3) The overlap area ratio of the interference region is obtained by dividing the intersection area of ​​the interference region mask and the preset evaluation region by the area of ​​the preset evaluation region; the visible area ratio of the body is obtained by dividing the intersection area of ​​the concrete body region and the preset evaluation region by the area of ​​the preset evaluation region. When the former reaches the interference threshold or the latter is lower than the visible threshold, the corresponding image frame is marked as an invalid frame; image frames not marked as invalid frames are used as valid image frames. When the amplitude of material surface movement, the displacement of material surface contour, or the inter-frame time cannot be generated, the corresponding image frame is additionally marked as an invalid frame.

[0037] The material surface status determination unit sets up a feeding detection area above the target observation area that intersects with the projection range of the mixer's feeding port. A feeding event marker is generated when a new material area intersects with the feeding detection area, moves across frames towards the concrete body area, and its area reaches the feeding area threshold and remains continuously for the required number of feeding frames. Splashed slurry points moving from the concrete body area to the feeding detection area do not participate in the generation of feeding event markers.

[0038] The residual slurry area on the machine wall is located on the surface of the mixing chamber scene. When the mixing component or concrete passes through this area, it can block the residual slurry area on the machine wall. The lens contamination area is located in the imaging optical path of the image acquisition device. Its boundary remains at a fixed camera coordinate position during multiple component cycles and continuously covers the image of the mixing component or concrete body area passing through this position.

[0039] The material surface condition determination unit identifies regions that maintain fixed boundaries across multiple component cycles, have local sharpness below a sharpness threshold, and cover images of moving objects passing through their locations as lens contamination areas. The ratio of the lens contamination area to the target observation area is the lens contamination area percentage; when the percentage reaches the contamination threshold, the corresponding image frame is marked as an invalid frame.

[0040] In one implementation, the evaluable window generation unit generates the evaluable window according to the following process: (1) When the image frame in which the backfill completion state and the disturbance attenuation state are simultaneously established for the first time is a valid image frame, the image frame is used as the starting frame of the window.

[0041] (2) When an entry event occurs, a feeding event is marked, the number of consecutive invalid frames reaches the consecutive invalid frame threshold, or any of the states fail, the valid image frame before the corresponding event occurs shall be used as the window termination frame.

[0042] (3) Arrange the valid image frames between the window start frame and the window end frame in chronological order to form a valid window frame sequence. When the length of the valid window frame sequence reaches the window length threshold, an evaluable window is retained; when the threshold is not reached, no evaluable window is generated, and the reason for window cancellation is recorded.

[0043] In one implementation, the slump visual assessment unit extracts the proportion of textured areas, continuous coverage of slurry, exposure rate of coarse aggregate, and local surface contour undulation from the concrete body area of ​​the effective frame sequence of the window, and arranges them into a surface state feature sequence according to the image frame order. The slump visual assessment unit generates a slump grade or slump range based on the surface state feature sequence and the slump calibration relationship.

[0044] In one implementation, the slump visual evaluation unit performs time-series alignment of the evaluable windows for each mixing cycle based on the image frames in which the backfill completion state and the disturbance decay state are simultaneously established for the first time, and generates single-cycle evaluation results.

[0045] If a set number of single-cycle evaluation results fall within the same or adjacent collapse intervals, the single-cycle evaluation results are deemed to meet the consistency condition. The fusion weight is obtained by multiplying the number of image frames in the effective frame sequence of the window by a factor minus the overlap area ratio of the interference region, and the collapse interval with the largest cumulative fusion weight is taken as the visual collapse evaluation result. If the consistency condition is not met, a result conflict marker is generated and an unevaluable state is output.

[0046] In one implementation, the system further includes a closed-loop interface unit. The result output unit generates associated records for the slump visual assessment results or unevaluable states. These associated records include a mixing cycle identifier, the start and end frame positions of the evaluable window, the generation position of the backfill completion state and the disturbance attenuation state, an invalid frame marker, the reason for window cancellation, and a result conflict marker. The closed-loop interface unit compares the slump visual assessment results with the target slump range and sends a signal to the mixer controller to allow material discharge, continue mixing, or add water for verification. Upon receiving an unevaluable state, it sends a signal to continue observation or collect an anomaly signal.

[0047] In this embodiment, the candidate observation segment determination unit designates the stirring component that triggers the departure event as the target stirring component. The first image frame in which the target stirring component and the target observation area change from intersecting to non-intersecting is the departure event frame; after the departure event frame, the first image frame in which any stirring component and the target observation area change from non-intersecting to intersecting is the entry event frame. The image frames from the departure event frame to the entry event frame constitute the candidate observation segment.

[0048] The component period is the time interval between two adjacent entry events corresponding to the same stirring component identifier. Candidate observation segments are associated with the component period of the target stirring component, and the termination position of a candidate observation segment is determined by the first entry event of any stirring component. The stirring component identifiers correspond to a single stirring blade, a single stirring arm, or a combination of stirring blades and stirring arms that simultaneously enter and leave the target observation area. The direction of motion and repeating trajectory of the same stirring component remain corresponding in adjacent component periods.

[0049] The material surface condition determination unit generates a reference area to be backfilled based on the component covering mask and the material surface disconnection area: , in, In order to be with the first The reference area to be backfilled corresponds to each mixing component; For marking the mixing components; According to the expansion distance The morphological dilation operation performed; For the first The distance of the mask edge extension of each stirring component; For the first The stirring component covers the mask in the valid image frame before leaving the event frame; For the first The location of the departure event frame for each stirring component; For the first The material surface breaks off when the mixing component leaves; The target observation area; Represents the union of regions; This indicates the intersection of regions. When no material surface break area is detected, the material surface break area is set to an empty set, and a reference area to be backfilled is generated based on the component cover mask. If the component cover mask is missing, the pixel area of ​​the reference area to be backfilled is 0, or the pixel length of the boundary of the reference area to be backfilled is 0, no evaluation window is generated, and the reason for window cancellation is recorded as the initial region being undeterminable.

[0050] The material surface break area is extracted within the adjacent range of the component cover mask and is adjacent to the trailing edge of the component movement. The expansion distance is calibrated based on the component edge positioning error in historical images and the component displacement error in adjacent image frames. Its value covers the main jitter range of the component mask edge and is less than or equal to half of the corresponding component sweep width.

[0051] The valid image frames within the candidate observation segment are numbered in chronological order. , with valid image frame number The corresponding original video frame position is denoted as Subscripts in the formulas for subsequent state quantities and changes , and These represent the current valid image frame, the previous valid image frame, and the previous valid image frame, respectively. One valid image frame; the leave event frame position, enter event frame position, and window start and end frame positions are based on the original video frame positions. Record.

[0052] In each valid image frame of the candidate observation segment, the material surface state determination unit identifies the portion of the reference area to be backfilled that is not covered by the concrete body as the remaining area to be backfilled, and uses the ratio of the pixel area of ​​the remaining area to be backfilled to the pixel area of ​​the reference area to be backfilled as the remaining area ratio to be backfilled. .

[0053] The material surface condition determination unit extracts the boundary segment adjacent to the non-concrete body connected area outside the reference area from the boundary of the reference area to be backfilled, and uses it as the reference open boundary. The boundary openness rate is calculated according to the following formula: , in, This is the boundary of the reference area to be backfilled. This is the set of boundary points adjacent to the non-concrete body connected to the outside of the reference area to be backfilled. The boundary openness ratio ranges from greater than or equal to 0 to less than or equal to 1.

[0054] The material surface state determination unit sets a first support area and a second support area on both sides of the material surface disconnection area according to the movement direction of the mixing component. The material surface disconnection connection path is a path formed by non-concrete body pixels and connecting the two support areas. When the two support areas cannot be determined, the material surface disconnection connection path mark is set to 1; when no material surface disconnection connection path is formed and the boundary open rate is less than or equal to the boundary closure threshold, the boundary closure condition is determined to be met.

[0055] The material surface condition determination unit performs connectivity component analysis of the concrete body region within the reference area to be backfilled and its adjacent range. When At that time, the proportion of the main connected components will be... Set to 0; when When the value is greater than or equal to 1, it is calculated according to the following formula: , in, For the first The proportion of the main connected components corresponding to each stirring component; This represents the number of connected components of the concrete body located in the reference area to be backfilled and its adjacent range in the current image frame. The serial number of the connected components of the concrete body; For the first A single connected component of the concrete body; Number the currently valid image frame The corresponding concrete body area; A positive area lower limit is set to prevent the denominator from being zero. The positive area lower limit is the area corresponding to one pixel. This represents the pixel area of ​​the region. This is for maximum value calculation; the remaining parameters follow the previous definition.

[0056] The material surface status determination unit performs frame-by-frame joint determination of the remaining area ratio, boundary openness ratio, and proportion of main connected components, and requires that the three quantitative criteria and the condition that no material surface disconnection and connection path are formed remain true in multiple consecutive image frames with valid image frame numbers. The joint relationship of the three quantitative criteria is as follows: , in, For the first The backfill completion status of each stirring component in the current image frame is 1, which indicates that the backfill completion status is established. To maintain the frame count for backfilling; To maintain the image frame offset within the interval; The backfill threshold corresponding to the remaining area ratio; This is the boundary closure threshold corresponding to the boundary openness rate; The connectivity recovery threshold is the percentage of the main connected components. For the position of the i-th stirring component in the image frame The corresponding material surface disconnection and penetration path marker takes a value of 1 when a material surface disconnection and penetration path is formed, and a value of 0 when no material surface disconnection and penetration path is formed. This means that all image frames within the interval must satisfy the conditions in parentheses. This indicates that all conditions are met simultaneously; the remaining parameters follow the previous definitions.

[0057] The backfill threshold and boundary closure threshold are both greater than or equal to 0 and less than 1, while the connectivity restoration threshold is greater than 0 and less than or equal to 1.

[0058] The backfill hold frame count is obtained by rounding up the product of the image acquisition frame rate and the shortest stable duration of the historical backfill process, and the backfill hold frame count is at least 2. The backfill hold frame count only counts consecutive valid image frames. If an invalid frame appears before the backfill completion state is generated, the hold count that has not yet been completed is cleared.

[0059] The state calibration version is a record of a set of image processing parameters and state determination parameters used for the same batch of concrete. All candidate observation segments of the same batch of concrete use the same state calibration version. If the position of the image acquisition device, the target observation area, or the feature extraction aperture changes, the generation of evaluable windows is stopped, and the reason for window cancellation is recorded as a state calibration version mismatch.

[0060] After the backfill completion status is generated, the material surface status determination unit still updates the remaining area ratio, boundary openness ratio and main connected component ratio in each valid image frame, and determines whether to cancel the backfill completion status according to the following hysteresis failure rules.

[0061] The area failure threshold is higher than the backfill threshold but less than or equal to 1; the boundary failure threshold is higher than the boundary closure threshold but less than or equal to 1; and the connectivity failure threshold is lower than the connectivity recovery threshold but greater than or equal to 0. These three failure thresholds are calibrated based on the fluctuation range of changes in the historical stable material surface stage and are bound to the same state calibration version as the generation threshold. If the remaining area ratio exceeds the area failure threshold for two consecutive valid image frames, the boundary openness rate exceeds the boundary failure threshold for two consecutive valid image frames, or the proportion of the main connected components is lower than the connectivity failure threshold for two consecutive valid image frames, the backfill completion state is revoked. If a material surface disconnection path is formed or an entry event occurs, the backfill completion state is immediately revoked and the candidate observation segment is terminated.

[0062] Historical annotation data is divided into parameter calibration set and parameter validation set according to the concrete batch. Data from the same batch of concrete is only included in one of the sets. Backfill threshold, boundary closure threshold, connectivity restoration threshold, number of frames to maintain each parameter, descent rate threshold, minimum total descent, perturbation threshold, restoration threshold, and window length threshold are generated from the parameter calibration set and compared on the parameter validation set.

[0063] The premature backfilling misclassification rate is the proportion of image frames whose backfilling is not yet complete being judged as backfilled image frames; the premature disturbance judgment rate is the proportion of image frames whose disturbance is not yet finished being judged as disturbance attenuation image frames; the effective window detection rate is the proportion of manually labeled evaluable candidate observation segments from which evaluable windows are generated. Parameter combinations are first selected to meet the corresponding upper limit of the premature misclassification rate, and then combinations with higher effective window detection rates are chosen. Parameters without fixed values ​​and the quantile positions used in historical distributions are determined through a parameter validation set based on the amount of historical data and the observation duration at the wet mixing end.

[0064] In a twin-shaft mixing mechanism, when multiple mixing blades or arms pass through the same target observation area sequentially, the material surface state determination unit saves the mixing component identifier, departure event frame position, and backfill reference area for each mixing component. When the next mixing component enters the target observation area, the area shrinkage process, boundary closure process, or connectivity restoration process corresponding to the previous mixing component is not continued to the next mixing component. Instead, the previous candidate observation segment is terminated, and the backfill reference area is re-established based on the component coverage mask of the next mixing component.

[0065] Example 2: Based on Example 1, the material surface state determination unit obtains the material surface motion amplitude, material surface contour displacement, and surface structure change within the intersection of the registered concrete body regions of adjacent valid image frames, forming a material surface motion amplitude sequence, a material surface contour displacement sequence, and a surface structure change sequence, respectively. The material surface motion amplitude is determined based on the optical flow displacement modulus, and the material surface contour displacement is determined based on the bidirectional nearest distance of the valid material surface contour points. Both are normalized based on the time interval between adjacent image frames and the diagonal pixel length of the target observation area. The surface structure change is determined by weighting the differences in local texture description, edge density changes, and the proportion of textured areas.

[0066] The material surface state determination unit smooths the three change sequence sequences. A downward trend is established when the decrease percentage of the material surface motion amplitude and the decrease percentage of the material surface contour displacement within the downward trend window reach their corresponding thresholds, and the total decrease of both changes reaches the corresponding minimum total decrease. After the downward trend is established, a disturbance attenuation state is generated when the smoothed material surface motion amplitude, smoothed material surface contour displacement, and smoothed surface structure change do not exceed their corresponding disturbance thresholds within the disturbance hold-up frames.

[0067] The disturbance attenuation state fails when any smooth change exceeds the corresponding recovery threshold within two consecutive valid image frames, or when a feeding event is marked, an entry event is detected, the area of ​​the splashing slurry point suddenly increases, or the area of ​​the lens contamination area reaches the contamination threshold.

[0068] In a preferred embodiment of this invention, the material surface state determination unit registers adjacent valid image frames in the valid image frame sequence, using the intersection of the concrete body regions of the two frames as the common body region. The material surface motion amplitude and surface structure change are calculated within the common body region; the material surface contour displacement is calculated based on the set of valid material surface contour points after registration in the two frames. The material surface state determination unit registers the concrete body regions in adjacent valid image frames, using the intersection of the concrete body region of the current image frame and the concrete body region of the previous valid image frame mapped to the coordinate system of the current image frame as the common body region.

[0069] When the effective pixel area of ​​the common body region reaches the minimum visible area of ​​the concrete body region, the diagonal pixel length of the target observation area is greater than 0, and the time interval between adjacent effective image frames is greater than 0, the material surface state determination unit uses the median of the optical flow displacement modulus within the common body region to calculate the material surface motion amplitude: , in, This represents the normalized amplitude of the material surface motion corresponding to the current image frame. This refers to the current image frame position within the candidate observation segment; It is the intersection of the concrete body region in the current image frame and the concrete body region of the previous valid image frame that has been registered and mapped to the coordinate system of the current image frame. The pixel positions within the intersection; pixel position The optical flow displacement vector from the previous valid image frame to the current image frame; Let L be the L2 norm of the vector; For median operations; The diagonal pixel length of the target observation area; This represents the time interval between adjacent image frames. The minimum visible area of ​​the concrete body is determined by the product of the preset evaluation area and the visibility threshold. If any calculation condition is not met, the current image frame is marked as an invalid frame, and no corresponding material surface movement amplitude is generated.

[0070] The material surface condition determination unit also extracts the outer contour of the concrete body area facing the inside of the mixing chamber, and calculates the material surface contour displacement based on the bidirectional nearest distance between contour points in adjacent image frames. The outer contour may include the boundary contour between the concrete body area and the background inside the machine, or it may include the internal raised contour obtained by extracting height undulation features; the same contour extraction aperture is used within the same candidate observation segment.

[0071] For the effective material surface contour points in both frames, resample at equal intervals according to the contour arc length, so that the two sets of contour points use the same sampling interval, and then calculate the bidirectional nearest distance. When the number of effective material surface contour points in the current image frame and the previous effective image frame both reach the contour point number threshold, the diagonal pixel length of the target observation area is greater than 0, and the time interval between the two image frames is greater than 0, the material surface contour displacement is calculated according to the following formula: , in, This represents the normalized material surface contour displacement per unit time corresponding to the current image frame. This is the set of valid material surface contour points in the current image frame coordinate system; This is the set of valid material surface contour points of the previous valid image frame that have been registered and mapped to the current image frame coordinate system; The number of valid material surface contour points in the current image frame; This represents the number of valid material surface contour points in the previous valid image frame. The position of the contour point in the current image frame; The position of the contour points after registration in the previous valid image frame; Position of contour points To the set of contour points The minimum Euclidean distance; Position of contour points To the set of contour points The minimum Euclidean distance; the remaining parameters follow the definitions above.

[0072] Valid material surface contour points are located within the concrete body area or at its boundaries, excluding contour points covered by interference area masks. If any calculation condition is not met, the current material surface contour displacement is marked as unavailable, the image frame is marked as invalid, and the smoothing results, descending trend window, and perturbation retention count are not updated. The contour point count threshold is calibrated based on the lower quantile value of the number of valid material surface contour points in historical valid image frames and is bound to the state calibration version. The quantile position used is determined through a parameter validation set.

[0073] The material surface state determination unit divides the common body region into local image blocks corresponding to the location, extracts a local texture descriptor from each local image block, such as gray-level co-occurrence statistics, local binary mode histogram, or gradient direction histogram, and calculates the descriptor distance between corresponding local image blocks in adjacent valid image frames; the average descriptor distance between each corresponding local image block is determined as the local texture descriptor difference.

[0074] Edge density is the ratio of the number of edge pixels to the number of pixels in the common body region. The absolute value of the edge density difference between adjacent valid image frames is determined as the edge density change.

[0075] The textured region is a linear edge-connected region in the common body region whose length-to-width ratio reaches the textured shape threshold and whose main direction change is not higher than the direction change threshold. The ratio of the pixel area of ​​the textured region to the pixel area of ​​the common body region is determined as the textured region proportion, and the absolute value of the difference between the textured region proportions of adjacent valid image frames is determined as the change in textured region proportion.

[0076] The method for calculating the change in surface structure is as follows: , , in, This represents the change in surface structure corresponding to the current image frame; The change in local texture descriptor quantity is the amount of change after the original difference in local texture descriptor quantity is converted per unit time and normalized to the historical upper and lower limits; The change in edge density is the amount of edge density change after unit time conversion and normalization to historical lower and lower limits, representing the original difference in edge density. The change in the proportion of the textured area after the original difference in the proportion of the textured area is converted to a unit time and normalized to the historical lower and lower limits; The fusion weights are the fusion weights corresponding to the differences in local texture descriptors. The fusion weights are the values ​​corresponding to the changes in edge density. This represents the fusion weight corresponding to the change in the proportion of the textured area.

[0077] The differences in local texture description, edge density variation, and the proportion of textured areas are first divided by the time interval between adjacent image frames to obtain the unit time variation. For each unit time variation, the difference between subtracting the historical calibration lower limit and dividing by the difference between the historical calibration upper and lower limits yields the normalized variation. A normalized variation less than 0 is set to 0, and greater than 1 is set to 1. When the historical calibration upper and lower limits are equal, the corresponding variation is marked as unusable.

[0078] All three fusion weights are greater than or equal to 0 and less than or equal to 1. The fusion weights are determined based on historical candidate observation segments with manually marked perturbation decay positions. Among the candidate weight combinations, combinations that control the proportion of prematurely generated perturbation decay states within a set range are first selected, and then the combination with the smallest deviation between the state generation position and the manually marked position is selected from the combinations.

[0079] Within at least three consecutive valid image frames after the start of the candidate observation segment, the material surface condition determination unit calculates the ratio of the effective area of ​​the textured region to the visible area of ​​the concrete body, and uses the median of the ratio to determine whether the textured region is usable. If the median is lower than the textured usability threshold, the change in the proportion of the textured region is set to 0 and the fusion weight is adjusted. Set to 0; when At that time, and Normalized according to the original proportion, when At that time, both were set to 0.5. The fusion weight remained fixed within the same candidate observation segment.

[0080] Before the texture availability determination is completed, the differences in local texture description, edge density changes, and texture area proportion changes are temporarily stored, and no perturbation attenuation state is generated. After the fusion weight is determined, the surface structure changes of the temporarily stored image frames are calculated back according to the image frame order. If no less than 3 consecutive valid image frames are formed before the candidate observation segment terminates, no perturbation attenuation state is generated.

[0081] When local texture descriptor differences or edge density variations are unavailable, the corresponding variations and fusion weights are reset to 0, and the fusion weights of the remaining available components are normalized proportionally. When the sum of the original fusion weights of the remaining available components is 0, no perturbation attenuation state is generated. When both local texture descriptor differences and edge density variations are unavailable, no corresponding perturbation attenuation state is generated, and the reason for window cancellation is recorded as "uncertain surface structure changes".

[0082] To suppress spikes caused by aggregate flipping in a single frame, localized reflections, and jitter at the segmentation edge, the three change sequences corresponding to the material surface motion amplitude, material surface contour displacement, and surface structure change are exponentially smoothed using the same time base: , in, The amplitude of the smooth material surface motion corresponding to the current image frame; This represents the smooth material surface contour displacement corresponding to the current image frame. This represents the amount of smooth surface structure change corresponding to the current image frame; To ensure smooth updates, the coefficients must be greater than 0 and less than or equal to 1. The amplitude of the smooth material surface motion corresponding to the previous valid image frame; This is the smooth material surface contour displacement corresponding to the previous valid image frame; This represents the change in smooth surface structure corresponding to the previous valid image frame; the remaining parameters follow the aforementioned definitions.

[0083] After the fusion weights are determined, the material surface state determination unit selects the first pair of consecutive valid image frames, and then selects the next image frame corresponding to... , and Initialize the smoothing result, with the previous valid image frame only used to calculate adjacent changes. Before the disturbance attenuation state is generated, invalid frames are cleared, the downward trend window is cleared, and the disturbance retention count is maintained. The fusion weights remain unchanged, and the smoothing result is reinitialized by the first pair of consecutive valid image frames.

[0084] The smoothing update coefficient is calibrated based on the effective image acquisition frame rate, decreasing as the frame rate increases and increasing as the frame rate decreases, and is truncated by preset upper and lower limits. The effective image acquisition frame rate is the frame rate corresponding to the effective timestamp interval of a set number of image frames before the start position of the candidate observation segment; if the frame rate cannot be obtained, the nominal frame rate of the image acquisition device is used. The smoothing update coefficient remains fixed within the same candidate observation segment. When the deviation of the time interval between adjacent image frames reaches the frame interval abnormality threshold, the current image frame is marked as an invalid frame; if a continuous set of effective image frames cannot be re-formed before the disturbance attenuation state is generated, the candidate observation segment is canceled.

[0085] The downward trend window includes at least three consecutive valid image frames. The material surface state determination unit counts the number of times the decrease in the smooth material surface motion amplitude and the decrease in the smooth material surface contour displacement of adjacent valid frame pairs within the downward trend window reach the corresponding single-frame decrease threshold, and uses the proportion of these numbers to the total number of adjacent valid frame pairs as the decrease ratio of the material surface motion amplitude and the decrease ratio of the material surface contour displacement. The single-frame decrease threshold is calibrated based on the upper quantile of the absolute value of the difference between adjacent valid frames corresponding to the change in the historical stable material surface stage.

[0086] A downward trend is established when both decreasing ratios reach their corresponding decreasing ratio thresholds, and the total decrease in the amplitude of the material surface movement and the total decrease in the displacement of the material surface profile between the start and end positions of the downward trend window reach their corresponding minimum total decreases. Starting from the first consecutive valid image frame after the end of the downward trend window, it is determined whether the three smoothed change quantities remain within their corresponding perturbation thresholds within the perturbation holding frame number.

[0087] The stirring speed corresponding to the current image frame is preferentially determined by the feedback value from the mixer controller; if the feedback value is unavailable, it is calculated from the repeated trajectory period of the same stirring component. When both speed sources are available, they are compared; if the relative difference in speed reaches the speed source conflict threshold, no disturbance attenuation state is generated, and the reason for window cancellation is recorded as speed source conflict. If neither speed source is available or the stirring speed exceeds the calibrated range, no disturbance attenuation state is generated, and the reason for window cancellation is recorded as speed undetermined.

[0088] The motion perturbation threshold, contour perturbation threshold, and surface structure perturbation threshold are calibrated based on historical candidate observation segments marked with the end position of artificial perturbation. After dividing the rotation speed range according to the stirring speed, the upper quantile value of the stable distribution of the change after the artificially marked position is used as the initial threshold, and it is corrected according to the proportion of image frames where the perturbation has not ended in advance.

[0089] The time interval between adjacent valid image frames is determined based on the image acquisition timestamp; if the image acquisition timestamp is unavailable, the reciprocal of the nominal frame rate of the image acquisition device is used as the time interval between adjacent valid image frames. If the resulting time interval is not greater than 0 or the deviation from the nominal inter-frame time reaches the frame interval abnormality threshold, the current image frame is marked as an invalid frame. The disturbance attenuation state is generated in the order of the descending trend phase first, followed by the stable holding phase, and its joint generation method is as follows: , in, Identification of target mixing components The corresponding candidate observation segment in the current valid image frame number The disturbance attenuation state is defined by a value of 1, which indicates that the disturbance attenuation state is valid. The percentage decrease in the amplitude of the material surface motion within the downward trend window, ending at the previous valid image frame of the first image frame of the stable maintenance phase; This is the threshold for the percentage decrease in the amplitude of material surface movement; The percentage decrease in the material surface contour displacement within the downward trend window, ending at the previous valid image frame of the first image frame of the stable holding phase. The threshold for the percentage decrease in material surface profile displacement; The number of consecutive valid image frames contained in the downward trend window; This represents the minimum total decrease in motion amplitude required within the downward trend window. This represents the minimum total reduction in profile displacement required within the downward trend window. The number of consecutive valid image frames included in the stabilization phase; To stabilize the image frame offset during the maintenance phase; Image frame position The motion disturbance threshold corresponding to the stirring speed; Image frame position The profile disturbance threshold corresponding to the stirring speed; Image frame position The surface structure disturbance threshold corresponding to the stirring speed; Image frame position The corresponding stirring speed; This indicates that each consecutive valid image frame during the stable holding phase satisfies the condition in parentheses. This indicates that all the conditions for judgment are met simultaneously; the remaining parameters follow the definitions above.

[0090] The values ​​of the decrease ratio of the material surface motion amplitude and the decrease ratio of the material surface contour displacement are greater than or equal to 0 and less than or equal to 1, and the corresponding decrease ratio threshold values ​​are greater than 0 and less than or equal to 1. The disturbance hold-up frame count is calculated by rounding up the product of the image acquisition frame rate and the shortest stable duration before the formation of the historical evaluable window, and is at least 2; the downward trend window includes at least three consecutive valid image frames. When an invalid frame appears before the disturbance decay state is generated, the downward trend window and the disturbance hold-up count are cleared, and the count is re-accumulated from subsequent consecutive valid image frames.

[0091] After the disturbance attenuation state is generated, the material surface state determination unit continues to update three smooth change quantities. The motion recovery threshold is higher than the motion disturbance threshold, the contour recovery threshold is higher than the contour disturbance threshold, and the surface structure recovery threshold is higher than the surface structure disturbance threshold but less than or equal to 1. If any smooth change quantity exceeds the corresponding recovery threshold within two consecutive valid image frames, the disturbance attenuation state is revoked.

[0092] When a feeding event is flagged, an entry event occurs, the growth rate of the splash area reaches the splash area surge threshold, or the proportion of the lens contamination area to the target observation area reaches the contamination threshold, the disturbance attenuation state is revoked, the downward trend window is cleared, and the disturbance retention count is reset. The growth rate of the splash area is calculated using the difference in splash area between two consecutive valid image frames as the numerator, and the larger of the splash area in the previous valid image frame and the area of ​​a single pixel as the denominator. The splash area surge threshold and the contamination threshold are calibrated based on historical candidate observation segments and historical image frames with lens contamination flags, respectively.

[0093] After the disturbance attenuation state is generated, invalid image frames do not update the three change quantity sequences. When the number of consecutive invalid frames reaches the consecutive invalid frame count threshold, the disturbance attenuation state is revoked; if the threshold is not reached, subsequent valid image frames continue to update based on the smoothed result of the previous valid image frame. The consecutive invalid frame count threshold is greater than 1, and is calculated based on the image acquisition frame rate and the maximum allowed continuous loss of material surface observation time.

[0094] Example 3: Based on Examples 1 and 2, the slump visual evaluation unit performs time-series alignment on the window's effective frame sequence according to the image frames in which the backfill completion state and the disturbance attenuation state are simultaneously established for the first time.

[0095] Within the concrete body region of each valid image frame, the proportion of the area of ​​the linear edge connected region that meets the texture shape threshold to the area of ​​the concrete body region is determined as the texture region proportion; the regions with local texture variance lower than the slurry texture threshold and interconnected are determined as slurry candidate regions, and the proportion of the area of ​​the largest slurry candidate connected region to the area of ​​the concrete body region is determined as the slurry continuous coverage rate. In the concrete body region, the mask coverage area of ​​interfering areas is excluded. Regions with local texture variance and edge density reaching the corresponding thresholds, area within the coarse aggregate scale range, and moving with the concrete body across frames are selected as coarse aggregate candidate regions. The ratio of the area of ​​the coarse aggregate candidate region to the non-interfering area of ​​the concrete body region is the coarse aggregate exposure rate. The material surface contour of the concrete body region facing the inside of the mixing chamber is extracted. The average distance from the material surface contour point to the material surface contour fitting baseline is divided by the diagonal pixel length of the target observation area to obtain the local material surface contour undulation. Each change is arranged in the image frame order according to the effective frame sequence of the window to form the surface state feature sequence.

[0096] Each slump calibration relationship is associated with the mixer identification, camera calibration version, and mix proportion category. The camera calibration version includes the image acquisition device installation location, imaging scale, target observation area boundary, and lighting configuration; the mix proportion category is jointly identified by the cementitious material dosage range, sand ratio range, maximum nominal coarse aggregate size, and target loading range.

[0097] The evaluation window only calls upon slump calibration relationships that are identical to the current mixer identifier, camera calibration version, and mix proportion category. If no matching slump calibration relationship exists, an "evaluable" status is output, and the reason for window cancellation is recorded as "slump calibration relationship mismatch."

[0098] The slump visual assessment unit groups historical evaluable windows according to the slump interval to which the standard slump test results belong. For each historical evaluable window, the mean, last value, and slope of change per unit time are extracted from the surface state feature sequence to form a window feature vector. The mean of the window feature vectors within each slump interval is calculated to obtain the interval feature center for the corresponding slump interval. For the evaluable window to be evaluated, a window feature vector is generated, and the weighted distance between this window feature vector and the feature centers of each interval is calculated. The interval with the smallest weighted distance is determined as the slump interval. The distance weights of each feature are initialized with the reciprocal of the historical variance of the corresponding feature within the same slump interval and normalized until the sum of all distance weights is 1. When the historical variance is 0, the corresponding feature is not used in the weighted distance calculation.

[0099] For the For each single-cycle evaluation result, the fusion weight is obtained by multiplying the number of image frames in the corresponding window's effective frame sequence by a factor minus the overlap area ratio of the interference region. When a set number of single-cycle evaluation results fall within the same collapse interval or adjacent collapse intervals, the fusion weights corresponding to each collapse interval are accumulated, and the collapse interval with the largest accumulated fusion weight is determined as the collapse visual evaluation result. If the sum of all fusion weights is 0 or the single-cycle evaluation results do not meet the consistency condition, a result conflict flag is generated and an unevaluable state is output.

[0100] The closed-loop interface unit sends a discharge permission signal when the visual slump assessment result is within the target slump range; it sends a continue mixing signal when the visual slump assessment result is below the lower limit of the target slump range and the remaining mixing time has not reached the additional mixing time; and it sends a water addition verification signal when the additional mixing time has been reached. It sends a continue observation signal when the visual slump assessment result is above the upper limit of the target slump range. When the result output unit outputs an unevaluable status, the closed-loop interface unit sends a data acquisition anomaly signal.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

[0102] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of the invention and form different embodiments. For example, all the embodiments above can be used in any combination. The information disclosed in the background section is intended only to enhance the understanding of the general background of the invention and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art.

Claims

1. A visual assessment system for slump in concrete mixing, characterized in that, include: The video acquisition unit collects video of the wet mixing end of the twin-shaft mixer and determines the target observation area; The candidate observation segment determination unit uses the image frames between the time the target stirring component leaves the target observation area and the time when any stirring component first enters the area as candidate observation segments. The material surface state determination unit determines the concrete body area within the candidate observation section and generates a reference area to be backfilled based on the covering mask and the material surface disconnection area of ​​the component. When the remaining area ratio to be backfilled is not higher than the backfill threshold, the boundary open rate is not higher than the boundary closed threshold, the proportion of the main connected components is not lower than the connectivity recovery threshold, and no material surface disconnection through path is formed, and the above four conditions are simultaneously met in consecutive effective image frames to reach the backfill holding frame number, a backfill completion state is generated, and a disturbance attenuation state is generated based on the decrease and stability of the material surface movement amplitude, material surface contour displacement, and surface structure change. The evaluable window generation unit collects valid image frames starting from the first valid image frame where both states are simultaneously established. It terminates collection when either state fails or an entry event occurs. An evaluable window is generated when the number of valid image frames reaches the window length threshold. The slump visual assessment unit generates slump visual assessment results based on the concrete body area within the evaluable window. The result output unit outputs the evaluation result.

2. The slump visual assessment system for concrete mixing according to claim 1, characterized in that, The stirring components include stirring blades and stirring arms. The stirring component identification corresponds to a single stirring blade, a single stirring arm, or a combination of stirring blades and stirring arms that move in and out simultaneously. The candidate observation segment determination unit compares the motion area with the historical component trajectory, identifies the stirring component based on its contour, direction and repeated trajectory, divides the component cycle according to two adjacent entry events of the same stirring component, associates the candidate observation segment with the component cycle of the target stirring component, and records the stirring component identifier and the position of the two event frames.

3. The slump visual assessment system for concrete mixing according to claim 1, characterized in that, The material surface state determination unit determines the concrete candidate area that is connected to the preset material surface reference area and moves across frames as the concrete body area, and merges the machine wall residual slurry area, the splash slurry point area and the local void area outside the reference area to be backfilled into the interference area mask. Low-resolution areas with fixed boundaries that cover the image of the mixing component across multiple component cycles are identified as lens contamination areas. When a new material area intersects with the feeding port area, moves towards the concrete body area, and its area reaches the feeding area threshold and remains at the feeding hold frame count, a feeding event marker is generated. When the overlap area ratio of the interference area reaches the interference threshold, the visible area ratio of the body is lower than the visible threshold, or the area ratio of the lens contamination area reaches the contamination threshold, it is marked as an invalid frame, and the remaining image frames are valid image frames.

4. The slump visual assessment system for concrete mixing according to claim 3, characterized in that, Adjacent valid image frames are two adjacent frames in a valid image frame sequence; The material surface state determination unit registers adjacent valid image frames, takes the intersection of the concrete body regions of the two frames as the common body region, calculates the material surface movement amplitude and surface structure change within the common body region, calculates the material surface contour displacement based on the set of valid material surface contour points of the two frames, and forms a corresponding sequence. The amplitude of the material surface motion is calculated by the optical flow displacement modulus, and the material surface contour displacement is calculated by the bidirectional nearest distance. Both are normalized according to the inter-frame time and the diagonal pixel length of the target observation area. The change in surface structure is obtained by weighting the difference in local texture description, the change in edge density, and the change in the proportion of textured area.

5. The slump visual assessment system for concrete mixing according to claim 4, characterized in that, The material surface condition determination unit smooths out the three change sequence sequences; When the decrease ratio of the material surface motion amplitude and the decrease ratio of the material surface contour displacement within the downward trend window reach the corresponding thresholds, and the total decrease of the two changes reaches the minimum total decrease, the downward trend is determined to be established; when the three smooth changes do not exceed the corresponding disturbance thresholds within the disturbance holding frame number, a disturbance attenuation state is generated; when the same smooth change exceeds the corresponding recovery threshold in two consecutive valid image frames, or when the feeding event is marked, an entry event occurs, or the growth rate of the splash slurry point area reaches the sudden increase threshold, the disturbance attenuation state fails.

6. The slump visual assessment system for concrete mixing according to claim 3, characterized in that, The evaluable window generation unit determines the window start frame when the frame that is simultaneously established as a valid image frame for the first time in both the backfill completion state and the disturbance attenuation state. When an entry event occurs, a feeding event is marked, the number of consecutive invalid frames reaches the corresponding threshold, or any of the states fails, the last valid image frame before the termination condition is met is used as the window termination frame. The effective image frames between the start and end frames are combined into a window effective frame sequence. When the sequence length reaches the window length threshold, an evaluable window is generated. If the threshold is not reached, the reason for the window cancellation is recorded. When no evaluable window is generated for a preset number of consecutive candidate observation segments of the same batch of concrete, the result output unit outputs an unevaluable state.

7. The slump visual assessment system for concrete mixing according to claim 6, characterized in that, The slump visual assessment unit extracts the proportion of textured areas, continuous coverage of slurry, exposure rate of coarse aggregate, and local material surface contour undulation from the effective frame sequence of the window to form a surface state feature sequence, and generates slump grade or slump range according to the slump calibration relationship. The slump calibration relationship is established using a historical evaluable window with standard slump test results and is associated with mixer identification, camera calibration version, and mix category.

8. The slump visual assessment system for concrete mixing according to claim 7, characterized in that, The slump calibration relationship includes the interval feature center corresponding to each slump interval; The slump visual assessment unit extracts the mean, end value and slope of change per unit time from the surface state feature sequence to form a window feature vector. It calculates the weighted distance between the window feature vector and the feature center of each interval and takes the slump interval with the smallest weighted distance as the single-cycle assessment result. The distance weight is determined by the historical variance of the corresponding feature within the same collapse interval.

9. The slump visual assessment system for concrete mixing according to claim 8, characterized in that, The slump visual assessment unit obtains a preset number of recently generated single-cycle assessment results, with the difference between the maximum and minimum values ​​of the slump interval number not exceeding 1 as a consistency condition. The average overlap area ratio of the interference region in the effective frame sequence of the window is used as the overlap area ratio of the interference region of the window. The product of the number of effective image frames and one minus the overlap area ratio of the interference region of the window is used as the fusion weight. The fusion weight of the cumulative collapse interval is used as the collapse interval with the largest cumulative fusion weight as the visual evaluation result of collapse. If the consistency condition is not met, all fusion weights are 0, or there are tied maximum values, output an unevaluable state and generate a result conflict flag.

10. The slump visual assessment system for concrete mixing according to claim 9, characterized in that, It also includes a closed-loop interface unit that connects to the mixer controller via a data interface; The output unit generates associated records, which include the stirring cycle identifier, the start and end frame positions of the window, the generation positions of the two states, the invalid frame marker, the reason for window cancellation, and the result conflict marker. When the slump visual evaluation result is within the target slump range, a signal to allow material release is sent; when it is below the lower limit of the range, a signal to continue stirring is sent; if it is still below the lower limit after the additional stirring time expires, a signal to add water for verification is sent; when it is above the upper limit of the range, a signal to continue observation is sent; and when an unevaluable state is received, a signal to send an abnormal acquisition signal is sent.

Images