A dynamic monitoring method for grain pile morphology based on intelligent vision system

Abstract

The present application provides a method for dynamic monitoring of grain pile morphology based on an intelligent vision system, comprising: extracting the velocity field and direction field of grain flow motion from video data of a falling grain flow using an optical flow field analysis method to obtain initial information on the grain flow trajectory; analyzing the initial diffusion range of the grain flow at the bottom of a silo based on the trajectory change information to determine whether the diffusion pattern is uniform; if the diffusion is uneven, marking the corresponding area as an irregular diffusion area; establishing a dynamic diffusion model based on the trajectory change information, processing the irregular diffusion area using the dynamic diffusion model, and predicting the stacking speed and morphological change trend of the grain pile based on the velocity field and direction field of the grain flow motion; and adjusting the operating parameters of the conveying equipment based on the predicted stacking speed and morphological change trend of the grain pile, the operating parameters including the conveyor belt speed and angle.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for dynamically monitoring grain pile morphology based on an intelligent visual system. Background Art

[0002] During grain silo loading, conveying equipment continuously introduces grain into the silo. Grain then freely falls from the end of the conveyor belt, forming a dynamic grain flow. The grain flow follows a specific trajectory during its fall, influenced by factors such as the physical properties of the grain particles, the height and angle of the conveying equipment, and air flow within the silo. Upon contacting the silo floor, the grain flow begins to spread and gradually accumulates, forming a grain pile. The grain pile formation process exhibits distinct dynamic characteristics, with the spreading pattern dependent on the grain's fluidity, the shape of the silo floor, and the height of the pile. As the grain pile grows, the grain accumulation pattern gradually transitions from localized diffusion to an overall uniform distribution. However, this process may lead to localized uneven accumulation, such as excessive or slow accumulation in certain areas, resulting in uneven or tilted grain pile surfaces. During the grain pile formation process, a complex dynamic relationship exists between the grain flow trajectory and its spreading pattern. The falling speed and angle of the grain flow directly influence the initial spreading range of the grain on the silo floor, which in turn determines the accumulation speed and shape of the grain pile. If the grain flow trajectory is unstable or the diffusion pattern is uneven, it may lead to local stacking differences in the grain pile, thereby affecting the storage efficiency and safety of grain in the warehouse. In order to monitor the uniformity of the grain pile formation process in real time, it is necessary to dynamically analyze the falling trajectory and diffusion pattern of the grain flow. Through optical flow field analysis technology, the velocity field and direction field of the grain flow movement can be captured to evaluate the uniformity of the grain pile during the stacking process. However, the complexity of the dynamic characteristics of grain flow makes optical flow field analysis challenging, such as sudden changes in the grain flow trajectory, irregular changes in the diffusion pattern, and interference from environmental factors in the warehouse, all of which affect the accuracy and real-time performance of optical flow field analysis. In addition, the dynamic changes in the grain pile shape have a direct impact on the control strategy of the warehousing operation. If the operating parameters of the conveying equipment are not adjusted in time, the grain pile shape may become uneven, resulting in uneven or tilted grain pile surface, thereby affecting the efficiency of subsequent outbound operations. Summary of the Invention

[0003] The present invention provides a method for dynamically monitoring grain pile morphology based on an intelligent visual system, which mainly includes:

[0004] The optical flow analysis method is used to extract the velocity and direction fields of the grain flow from the video data of the falling grain flow, and the initial information of the grain flow trajectory is obtained.

[0005] According to the velocity field and direction field of the grain flow movement, if the rate of change of the velocity field or direction field in a local area exceeds a preset rate threshold, it is determined that there is a sudden change in the grain flow trajectory, and the area with the sudden change is marked as a trajectory sudden change area;

[0006] The local optical flow field analysis method is used to refine the velocity field and direction field in the trajectory mutation area to obtain trajectory change information. The trajectory change information includes the change amplitude, change direction, change duration and change area range;

[0007] Based on the trajectory change information, the initial diffusion range of the grain flow at the bottom of the silo is analyzed to determine whether the diffusion pattern is uniform. If the diffusion is uneven, the corresponding area is marked as an irregular diffusion area.

[0008] A dynamic diffusion model is established based on trajectory change information. The model is used to process irregular diffusion areas and, combined with the velocity and direction fields of grain flow, the accumulation velocity and morphological change trend of grain piles are predicted.

[0009] Adjust the operating parameters of the conveying equipment based on the predicted grain pile accumulation speed and shape change trend, including the conveyor belt speed and angle;

[0010] Monitor the grain flow trajectory and diffusion pattern of the adjusted conveying equipment in real time to determine whether the grain pile height standard deviation, coverage area uniformity and stacking density distribution have been improved. If so, record the final grain flow trajectory and diffusion pattern data.

[0011] Furthermore, the optical flow field analysis method is used to extract the velocity field and direction field of the grain flow movement from the video data of the grain flow falling process, and the initial information of the grain flow trajectory is obtained, including: according to the grayscale value changes between adjacent frames in the video frame sequence images collected during the grain flow falling process, the edge contour area of the grain flow is identified by a region matching method, the grain flow area in the identification result is divided into multiple grid units, the corresponding relationship between the illuminance value and the grayscale value of each grid unit is calculated to generate a grain flow density distribution map, a grayscale value gradient constraint equation and a brightness constant constraint equation are established between two adjacent frames of images, and the constraint equation group is solved according to the multi-scale iterative method, the constraint equation group refers to the grayscale value gradient constraint equation and the brightness constant constraint equation, and the grain flow optical flow vector field distribution data is obtained from the solution results of the constraint equation group; the grain flow in each grid unit is calculated according to the grain flow optical flow vector field distribution data. The velocity vector of the grain flow is fitted using the least squares method to fit the grid unit velocity vector field to obtain the overall motion trend data of the grain flow; if the angle between the velocity vector of any grid unit and the velocity vector of the adjacent unit in the overall motion trend data of the grain flow is greater than the preset threshold, a tracking point is set at the edge of the grid unit, and the displacement of each tracking point is calculated using the pyramid-based Lucas-Kanade optical flow tracking algorithm; the tracking point displacement data is Gaussian filtered to obtain smoothed trajectory data, and the minimum enclosing rectangle method is used to calculate the coordinate sequence of the grain flow center of mass; a quadratic polynomial fitting function is constructed based on the grain flow center of mass coordinate sequence as the grain flow motion feature description function, and geometric parameters such as the grain flow area, edge length and diffusion radius are calculated from the function coefficients; the Euclidean distance calculation method is used to compare the changes in geometric parameters between the previous and next frames to obtain the grain flow motion parameters, and the initial information of the grain flow falling trajectory is obtained through the grain flow motion parameters.

[0012] Furthermore, according to the velocity field and direction field of the grain flow movement, if the rate of change of the velocity field or the direction field in a local area exceeds a preset rate of change threshold, it is determined that there is a mutation in the grain flow trajectory, and the area with the mutation is marked as a trajectory mutation area, including: uniformly gridding the grain flow area according to the grain flow velocity field data, using the Sobel operator to calculate the first-order difference values of the velocity data point set in each grid unit in the horizontal and vertical directions, and obtaining the velocity field change rate matrix from the difference values; for the direction vector set in each grid unit in the grain flow direction field data, using the three-point fitting method to calculate the angle change between adjacent direction vectors, and obtaining the direction field change rate matrix through the angle change calculation; using the double threshold comparison method to compare the grain flow area with the direction vector set in each grid unit in the grain flow direction field data. The velocity field change rate matrix and the direction field change rate matrix are tested. If the velocity field change rate in any grid cell exceeds the preset velocity threshold or the direction field change rate exceeds the preset direction threshold, the grid cell is marked as a mutation candidate point. The connectivity analysis of all mutation candidate points is performed using the region growing algorithm. The set of adjacent mutation candidate points is determined through an eight-neighborhood search method. The boundary tracking algorithm is used to extract the regional boundary point sequence. A piecewise cubic spline curve is constructed from the boundary point sequence as the trajectory mutation region boundary curve. Based on the trajectory mutation region boundary curve, the regional area value is obtained using the integral calculation method, and the regional perimeter value is obtained using the arc length calculation method. The trajectory mutation region range parameters are obtained from the area and perimeter values.

[0013] Furthermore, the local optical flow field analysis method is used to refine the velocity field and direction field in the trajectory mutation area to obtain trajectory change information, which includes the change amplitude, change direction, change duration and change area range, including: dividing the optical flow field data into local grids of fixed pixel size according to the trajectory mutation area range, recalculating the local velocity field and direction field using the pyramid optical flow algorithm for each grid unit, and obtaining the velocity vector and direction vector sequence in the grid unit from the local field data; fitting the velocity vector sequence in the grid unit using the cubic polynomial least squares method, calculating the velocity difference between adjacent frames through the velocity change curve, and performing trapezoidal integration operation on the difference to obtain the velocity change amplitude parameter; constructing a curve of direction angle changing with time according to the direction vector sequence in the grid unit, and extracting the direction angle in the direction change curve using the local extreme value detection method. The inflection point coordinate sequence is obtained by quadratic curve fitting of the inflection point coordinate sequence to obtain the direction change angle parameter; for the speed change amplitude parameter, the continuous frame counting method is used to obtain the number of change duration frames, and the change duration parameter is obtained by calculating the number of frames and the sampling frequency; based on the speed change amplitude parameter and the direction change angle parameter, the weighted average method is used to calculate the motion change intensity value in each grid unit, and the motion change intensity value is normalized to obtain the motion trend data; for the motion trend data, the regional centroid calculation method is used to obtain the coordinate sequence of the center point of the mutation area, and the minimum enclosing rectangle method is used to construct the regional boundary outline through the coordinate sequence, and the change area range parameter is calculated from the boundary outline; based on the change area range parameter, change duration parameter, speed change amplitude parameter and direction change angle parameter, the trajectory change feature vector is constructed, and the complete trajectory change information is obtained from the feature vector.

[0014] Furthermore, the initial diffusion range of the grain flow at the bottom of the silo is analyzed based on the trajectory change information to determine whether the diffusion pattern is uniform. If the diffusion is uneven, the corresponding area is marked as an irregular diffusion area, including: according to the change amplitude and change direction data in the trajectory change information, the region growing algorithm based on eight neighborhoods is used to segment the grain flow coverage area at the bottom of the silo, and the grayscale similarity is used as the growth condition to obtain the initial diffusion range boundary point sequence of the silo bottom, and the diffusion range contour curve is constructed from the boundary point sequence; for the area divided by the diffusion range contour curve, the area is divided into uniform grids of fixed pixel size, and the pixel counting method is used to count the number of grain flow pixels in each grid unit, and the volume density value of each grid unit is calculated from the number of pixels; the volume density value of each grid unit is normalized to the maximum and minimum values, and the normalized volume density value is normalized to the minimum and maximum values using a linear mapping method. The results are converted into diffusion intensity values, and a diffusion intensity distribution map is constructed from the diffusion intensity values. Based on the diffusion intensity distribution map, the Sobel operator is used to calculate the diffusion intensity gradients in the horizontal and vertical directions, and the diffusion intensity change rate data is obtained through the gradient value calculation. For the diffusion intensity change rate data, the weighted average method is used to calculate the diffusion uniformity value of the entire region, and the diffusion irregularity parameter is calculated by calculating the difference between the diffusion uniformity value and the ideal uniform distribution. If the diffusion irregularity parameter exceeds the preset uniformity threshold, the connected domain marking algorithm is used to mark the area where the diffusion intensity gradient is greater than the gradient mean, and the diffusion irregular area is obtained from the marking result. For the diffusion irregular area, the boundary tracking algorithm is used to extract the regional contour point sequence, and the irregular area boundary curve is constructed through the contour point sequence, and the irregular area range data is obtained from the boundary curve.

[0015] Furthermore, a dynamic diffusion model is established based on trajectory change information, and the dynamic diffusion model is used to process the irregular diffusion area. Combined with the velocity field and direction field of the grain flow movement, the accumulation speed and morphological change trend of the grain pile are predicted, including: according to the change amplitude and change direction data in the trajectory change information, a double-layer feedforward neural network is used to construct a dynamic diffusion prediction model, and the diffusion rate and direction prediction parameters are obtained through historical data training, and the grain flow diffusion movement trend in the irregular area is calculated from the prediction parameters; for the diffusion movement trend data, a uniform grid with a fixed pixel size is used to spatially divide the irregular area, and the accumulation amount change value in the grid unit is calculated through the velocity field and direction field data, and the accumulation density change rate per unit time is obtained from the accumulation amount change value; according to the accumulation density change rate, a bilinear interpolation method is used to perform density distribution on the grid unit. Reconstruction: Construct a grain pile height field function through density distribution data, calculate the three-dimensional morphological data of the grain pile from the height field function, use the contour line extraction algorithm to obtain a sequence of contour lines at different heights, calculate the stacking angle of each layer through the contour line sequence, and construct the edge contour curve of the grain pile from the stacking angle data; according to the edge contour curve of the grain pile, use open and close operations to perform morphological smoothing on the curve, and calculate the horizontal expansion area of the grain pile through the processed contour curve; for the horizontal expansion area of the grain pile, perform spatial integration operation in combination with the height field function, obtain the function of the change of the grain pile volume with time from the integral result, and obtain the predicted value of the stacking velocity by derivation of the change function; calculate the error between the predicted value of the stacking velocity and the actual observation data, optimize the parameters of the dynamic diffusion prediction model through the back propagation algorithm, and obtain the trend of the stacking morphology change from the optimized model.

[0016] Furthermore, the operating parameters of the conveying equipment are adjusted according to the predicted grain pile stacking speed and shape change trend, and the operating parameters include the conveyor belt speed and angle, including: according to the predicted grain pile stacking speed and shape change trend data, an adaptive control algorithm is used to establish a conveyor belt parameter response function, the input layer includes the stacking speed and shape characteristics, the hidden layer uses a hyperbolic tangent activation function, and the output layer generates a speed adjustment amount and an angle adjustment amount; for the speed adjustment amount and the angle adjustment amount, an adjustment amount limiting function is constructed based on the preset equipment operation constraints, the adjustment amount is range-constrained by the limiting function, the conveyor belt operation parameter boundary value is obtained from the constraint result, and the speed and angle parameters are linked and analyzed using a fuzzy rule reasoner, the input variables are divided into three fuzzy sets of low speed, medium speed and high speed, and the output variables are divided into small angles , medium angle and large angle three fuzzy sets; for the fuzzy reasoning results, the centroid method is used for defuzzification, the speed adjustment step and angle adjustment step are calculated through the defuzzification results, and the parameter adjustment path curve is constructed from the adjustment step data; according to the parameter adjustment path curve, the mean square error is used as the objective function, the speed response curve and the angle response curve are calculated through gradient descent iteration, the belt speed increment sequence and the angle increment sequence are obtained from the response curve, the piecewise linear mapping method is used for discretization within the set control interval, and the conveyor belt control instruction sequence is obtained through the discretization results; according to the conveyor belt control instruction sequence, the linear interpolation algorithm is used to smooth the instruction data, the conveying equipment operating parameters are adjusted in real time through the processed instruction data, and the real-time operating parameters of the conveyor belt are obtained from the adjustment data.

[0017] Furthermore, the real-time monitoring of the grain flow trajectory and diffusion pattern of the adjusted conveying equipment determines whether the grain pile height standard deviation, coverage area uniformity and stacking density distribution of the grain pile are improved. If improved, the final grain flow trajectory and diffusion pattern data are recorded, including: according to the operating status of the adjusted conveying equipment, using an infrared height measurement sensor array to perform a grid scan on the surface of the grain pile, constructing a three-dimensional height distribution map of the grain pile through the scanning data, extracting the height data of each measuring point from the height distribution map, calculating the average height value using the least squares method, summing the deviations of each measuring point from the average value, and obtaining the grain pile height standard deviation index by dividing the sum of the deviations by the number of measuring points; according to the three-dimensional height distribution map of the grain pile, using a regional growing algorithm based on height gradient to segment the grain pile coverage area, merging areas where the height difference of adjacent points is less than a preset threshold into the same category, and obtaining the coverage area rotation by regional merging. The area is divided into regular hexagonal grids, and the local uniformity value is obtained by calculating the ratio of the height change rate to the area in each grid cell, and the area uniformity parameter is calculated from the local uniformity value; according to the regular hexagonal grid division result, a laser density detector array arranged vertically downward is used to perform density scanning on each grid cell, and the stacking density distribution map is obtained through the scanning data. The Gaussian kernel function is used for convolution operation, and the density difference between adjacent grid cells is calculated through the convolution result, and the density gradient distribution data is obtained from the density difference; if the height standard deviation index is lower than the preset standard deviation threshold and the area uniformity parameter is higher than the preset uniformity threshold, four cameras are used to collect the grain flow trajectory from multiple angles, and the complete trajectory sequence is obtained through image stitching. The convolution neural network is used to extract trajectory features, and the features are classified through the fully connected layer. The diffusion pattern database is constructed from the classification results, and the database content is structured and stored.

[0018] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0019] The present invention discloses a method for dynamic monitoring of grain pile morphology based on an intelligent visual system. The method processes video data of the grain flow falling process through optical flow field analysis technology, extracts the velocity field and direction field of the grain flow movement, and identifies the trajectory mutation area. The local optical flow field analysis method is further used to refine the trajectory change information, and analyze the initial diffusion range and uniformity of the grain flow at the bottom of the warehouse. Based on this, the present invention establishes a dynamic diffusion model to predict the stacking speed and morphological change trend of the grain pile, and adjusts the operating parameters of the conveying equipment accordingly. Through real-time monitoring and feedback adjustment, the present invention can effectively improve the height uniformity, coverage area uniformity and stacking density distribution of the grain pile, realize the optimized control of grain storage, and improve storage efficiency and quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1The present invention is a flow chart of a method for dynamically monitoring grain pile morphology based on an intelligent visual system.

[0021] Figure 2 This is another flow chart of a method for dynamically monitoring grain pile morphology based on an intelligent visual system according to the present invention. DETAILED DESCRIPTION

[0022] To help those skilled in the art better understand the technical solutions in this specification, the following will provide a clear and complete description of the technical solutions in the embodiments of this specification, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. All other embodiments derived by those skilled in the art based on the embodiments in this specification without creative effort shall fall within the scope of protection of this specification.

[0023] like Figure 1-2 In this embodiment, a method for dynamically monitoring grain pile morphology based on an intelligent visual system may specifically include:

[0024] S101. During the falling grain flow, an intelligent vision system is used to collect video data and optical flow field analysis technology is used to extract the velocity field and direction field of the grain flow movement, generating the initial information of the grain flow trajectory. At the same time, a multi-level analysis method is used to calculate the geometric characteristics and motion parameters of the grain flow area to support subsequent diffusion assessment.

[0025] In this embodiment of the present invention, the grain flow is guided into the silo by conveying equipment as it falls. An intelligent vision system uses a camera to capture a sequence of video frames in real time, which is then used to analyze the dynamic characteristics of the grain flow. Optical flow analysis technology captures the motion characteristics of the grain flow between frames, generating velocity and direction field data, which provides a basis for trajectory monitoring.

[0026] S1011. By analyzing the grayscale value variations between adjacent frames in a video frame sequence, a region matching method is used to identify the grain flow edge contour area and divide it into grid cells. The relationship between the illuminance and grayscale value of each cell is calculated to generate a grain flow density distribution map. A set of constraint equations is then established based on the density distribution map, and the optical flow vector field distribution data is solved. In this embodiment of the present invention, the captured video frame sequence is first preprocessed to extract the grayscale value variation characteristics between adjacent frames. The grain flow edge contour is located using a region matching method, and the grain flow area is divided into multiple grid cells, for example, a 32×32 pixel grid. For each grid cell, the corresponding relationship between illuminance and grayscale value is calculated to generate a density distribution map reflecting the grain flow distribution pattern. Next, based on the grayscale gradient and brightness constancy characteristics in the density distribution map, grayscale gradient constraint equations and brightness constancy constraint equations are established, respectively. This set of constraint equations is solved using a multi-scale iterative method to obtain the grain flow optical flow vector field distribution data, describing the movement of the grain flow within each grid cell. Taking a grain unloading scene in a granary as an example, the time interval between adjacent frames is 0.04 seconds. The vector distribution in the central area is uniform, while the edge area shows a speed gradient change.

[0027] S1012. Calculate the grain flow velocity vector in each grid unit based on the optical flow vector field distribution data, use the least squares method to fit the velocity vector field to generate the overall grain flow movement trend data, and calculate the grain flow area and diffusion radius through the quadratic polynomial fitting function. At the same time, use tracking points and filtering technology to refine the motion parameters to obtain the initial information of the grain flow trajectory.

[0028] In an embodiment of the present invention, the grain flow velocity vector is calculated for each grid cell based on the optical flow vector field distribution data, and the overall grain flow motion trend data is obtained through least squares fitting. If the angle between the velocity vector of a grid cell and an adjacent cell exceeds a preset threshold, such as 25 degrees, a tracking point is set at the edge of the area, and the displacement of the tracking point is calculated using the pyramid Lucas-Kanade optical flow tracking algorithm. A Gaussian filter is applied to the displacement data to generate smoothed trajectory data, and the minimum bounding rectangle method is used to calculate the coordinate sequence of the grain flow center of mass. A quadratic polynomial fitting function is constructed based on the center of mass sequence, whose coefficients reflect the grain flow motion characteristics. This function is used to calculate geometric parameters such as the grain flow area, edge length, and diffusion radius. The Euclidean distance method is further used to compare the changes in geometric parameters between the previous and next frames to quantify the grain flow motion parameters. For example, when the Euclidean distance exceeds 8, it indicates a significant change in the grain flow trajectory. Taking the actual grain unloading process as an example, the vertical displacement of the center of mass is approximately 15 pixels per frame, and the horizontal offset is controlled within 5 pixels. The fitting function accurately describes the trajectory curvature. In this embodiment of the present invention, the falling grain flow exhibits a jet-like distribution, with the grayscale value of the edge region varying between 110 and 180. Through the above analysis, the generated initial trajectory information includes the velocity field, direction field, and geometric parameters, laying the foundation for subsequent diffusion pattern determination. This step does not impose excessive restrictions on the specific algorithm details, and technical personnel can optimize it based on the scenario. In practical applications, these parameters are derived from experimental statistics and can effectively characterize the movement patterns of grain flow, ensuring accurate and real-time monitoring.

[0029] S102. During the falling grain flow, the trajectory mutation situation is determined based on the velocity field and direction field data of the grain flow extracted by the intelligent vision system. If the rate of change of the velocity field or direction field in the local area exceeds the preset threshold, it is determined as a trajectory mutation and the corresponding area is marked as a trajectory mutation area. At the same time, the boundary and range parameters of the mutation area are determined through grid analysis and curve fitting technology to support subsequent diffusion analysis.

[0030] In this embodiment of the present invention, the velocity and direction fields of grain flow reflect its dynamic characteristics. By monitoring local changes in these two fields, trajectory anomalies can be promptly detected. Abrupt trajectory changes are typically caused by grain flow blockage, deflection, or external interference, and require precise identification to optimize grain pile morphology.

[0031] S1021. Divide the grain flow area into uniform grids based on the grain flow velocity field data. Use the Sobel operator to calculate the first-order difference values of the velocity data in the horizontal and vertical directions in each grid cell and generate a velocity field change rate matrix. At the same time, calculate the change in the angle between adjacent vectors for the direction field data to construct a direction field change rate matrix, providing a quantitative basis for mutation detection.

[0032] In an embodiment of the present invention, the grain flow area is divided into a uniform grid of 16×16 pixels to ensure analysis accuracy. The Sobel operator is used to perform differential calculations on the velocity data to obtain the change characteristics in the horizontal and vertical directions respectively. For example, in a certain granary unloading scene, the normal grain flow velocity change rate remains below 0.15, while it can rise to above 0.4 when blocked. The directional field analysis calculates the angle change between adjacent vectors through the three-point fitting method to generate a directional field change rate matrix. In normal fall, the angle change is usually less than 15 degrees, while it can reach more than 35 degrees during deflection. This quantification method clearly reflects the degree of abnormality in grain flow movement.

[0033] S1022. Perform dual threshold detection on the velocity field change rate matrix and the direction field change rate matrix. If the velocity field change rate of a grid cell exceeds a preset velocity threshold or the direction field change rate exceeds a preset direction threshold, it is marked as a mutation candidate point. Then, the connectivity of the candidate points is analyzed through the region growing algorithm and the initial mutation region boundary is extracted to further refine the mutation range.

[0034] In this embodiment of the present invention, dual-threshold detection sets a velocity field threshold of 0.35 and a direction field threshold of 30 degrees. When a grid cell exceeds either threshold, it is marked as a candidate mutation point. For example, during a material feeding inspection, three velocity outliers and two direction outliers were identified within a 350×400 pixel area. A region growing algorithm was used to cluster adjacent candidate points using an eight-neighborhood search to form an initial mutation region. This method effectively integrates scattered outliers and ensures the integrity of the mutation region.

[0035] Boundary extraction is a key step in identifying mutation regions. For the initial mutation region, a boundary tracking algorithm is used to trace the edges point by point, generating a sequence of boundary points. For one example, the sequence obtained contained 86 points. A smooth boundary curve with a length of 320 pixels is then constructed using piecewise cubic spline interpolation. This interpolation technique smooths boundary noise and improves the accuracy of region description.

[0036] S1023. Calculate the area and perimeter parameters of the mutation area based on the boundary curve of the mutation area. Obtain the area value through the integration method and calculate the perimeter value using the arc length formula. Analyze the geometric shape characteristics of the mutation area based on the area-to-perimeter ratio, thereby providing data support for determining the cause of the abnormal grain flow trajectory.

[0037] In an embodiment of the present invention, boundary curve analysis further quantifies the characteristics of the sudden change region. Through integral calculation, a sudden change region has an area of 4,800 square pixels and a perimeter of 360 pixels, with an area-to-perimeter ratio of 13.3. This long, narrow shape indicates that the grain flow may be deflected by an obstacle. In practical applications, such parameters can be correlated with the operating status of the feeding equipment. For example, a high ratio may indicate a conveyor belt angle deviation or interference from foreign objects in the bin. Monitoring these geometric characteristics can provide a basis for subsequent adjustments.

[0038] The identification and parameterization of trajectory mutation areas lays the foundation for grain pile morphology optimization. The dual analysis of velocity and direction fields ensures the comprehensiveness of mutation detection, while grid division and curve fitting techniques improve the accuracy of boundary extraction. In actual scenarios, such as the unloading process of a grain transfer station, the above method can quickly locate the problem area when grain flow anomalies first appear, providing reliable data support for real-time control of transportation parameters. This step does not impose too many restrictions on threshold settings or algorithm details, and can be optimized and adjusted by technical personnel according to specific application scenarios.

[0039] S103. During the grain fall, local optical flow analysis is used to extract detailed velocity and direction field data for areas with sudden changes in trajectory. This generates trajectory change information, including the magnitude, direction, duration, and regional extent of the change. A feature vector is constructed through polynomial fitting and weighted calculation to support grain pile morphology prediction. Local optical flow analysis captures the microscopic motion characteristics of areas with sudden changes in trajectory, providing precise data support for subsequent diffusion pattern analysis. These sudden changes are typically caused by obstruction or deflection of grain flow, and detailed analysis can reveal their impact and changing trends.

[0040] In an embodiment of the present invention, the trajectory mutation area is divided into a local grid of 16×16 pixels, with each grid containing 256 pixels to ensure comprehensive detail capture. The pyramid optical flow algorithm is used to recalculate the velocity field and direction field of each grid cell to generate a sequence of velocity vectors and direction vectors. For example, in a grain silo unloading scene, the velocity vector of the normal grain flow changes smoothly, while the vector in the mutation area fluctuates significantly, and the local velocity can surge from 2 pixels per frame to 8 pixels per frame. This algorithm improves adaptability to complex motion through multi-scale decomposition.

[0041] S1031. For the velocity vector sequence within the grid unit, use the cubic polynomial least squares method to fit the velocity change curve and calculate the velocity difference between adjacent frames to obtain the velocity change amplitude. At the same time, analyze the inflection point characteristics through the direction vector sequence to generate the direction change angle, and combine the continuous frame number to calculate the change duration to quantify the mutation characteristics.

[0042] In an embodiment of the present invention, a cubic polynomial is used to fit the velocity vector sequence, which can smooth noise and reflect the trend of change. After calculating the velocity difference between adjacent frames, the velocity change amplitude parameter is obtained by the trapezoidal integration method. In normal fall, the difference is mostly 2 to 3 pixels per frame, while it can reach more than 8 pixels per frame during mutation. For the direction vector, a direction angle change curve is constructed and the inflection point is extracted by local extreme value detection. The direction change angle parameter is generated after quadratic curve fitting. In actual detection, an inflection point angle exceeding 45 degrees often indicates a grain flow deflection. The duration of the change is obtained by counting continuous frames. Calculated at a sampling rate of 50 frames per second, a short disturbance is about 0.2 seconds, and a severe anomaly can exceed 0.5 seconds. These parameters together describe the intensity and duration of the mutation.

[0043] S1032. Calculate the motion change intensity value using the weighted average method based on the speed change amplitude and direction change angle, and normalize it to generate motion trend data. Then, obtain the coordinate sequence of the center point of the mutation area through regional centroid calculation and construct the boundary contour. Finally, combine various parameters to generate a trajectory change feature vector to fully represent the change information.

[0044] In an embodiment of the present invention, the motion change intensity value is calculated with a speed and direction weight of 3:2, highlighting the dominant role of speed mutation. After normalization, motion trend data is generated to facilitate cross-regional comparison. The center point coordinate sequence is calculated using the regional centroid method, and the boundary contour is constructed using the minimum circumscribed rectangle method to obtain the range of the change area. For example, in a certain detection, the area of the normal disturbance area is about 2000 square pixels, and the blocked area exceeds 5000 square pixels. The trajectory change feature vector integrates the speed change amplitude, direction change angle, duration and range parameters to provide a multi-dimensional anomaly assessment basis.

[0045] For example, analysis of the velocity and directional characteristics of a sudden change in a grain processing plant revealed deflections caused by equipment failure. The boundary contours showed a long and narrow shape, suggesting the presence of an obstruction within the warehouse. Weighted calculations and coordinate analysis quantified the spatiotemporal characteristics of the anomaly, providing a reliable reference for adjusting transportation strategies.

[0046] In this embodiment of the present invention, the generated trajectory change information is fully converted from raw data to feature vectors through multi-level analysis. In practical applications, this allows for rapid response to the initial onset of sudden changes in grain flow. For example, when the directional change angle suddenly increases to 45 degrees and persists for 0.5 seconds, combined with the characteristics of an area exceeding 5,000 square pixels, it can be determined to be a severe blockage, triggering subsequent regulatory actions in a timely manner. The resulting feature vectors lay a solid foundation for the establishment of dynamic diffusion models, enhancing the practicality and adaptability of monitoring.

[0047] S104. After the grain flow reaches the bottom of the silo, the initial diffusion range of the grain flow is analyzed based on the trajectory change information and its uniformity is determined. If the diffusion pattern is uneven, it is marked as an irregular diffusion area. Regional segmentation and density distribution calculation are used to obtain the boundary curve of the irregular area to guide subsequent stacking adjustments. The diffusion behavior of the grain flow at the bottom of the silo directly affects the morphological stability of the grain pile. Uneven diffusion can lead to localized over- or under-stacks, affecting storage efficiency. Refining the analysis of diffusion range and intensity provides a basis for controlling conveying equipment.

[0048] S1041. Based on the amplitude and direction of change in the trajectory change information, the eight-neighborhood region growing algorithm is used to segment the grain flow coverage area at the bottom of the silo. A sequence of boundary points of the initial diffusion range is extracted using grayscale similarity as a condition. Subsequently, a grid is formed based on the boundary point sequence, and the volume density value of each grid cell is calculated to construct basic data for the diffusion intensity distribution. In this embodiment of the present invention, the region growing algorithm is based on an eight-neighborhood search, classifying pixels with grayscale value differences of less than 10 as the same region. For example, for a grain flow falling from a height of 10 meters, its boundary point sequence typically contains 200 to 300 points, reflecting the outline of the diffusion range. Next, the coverage area is divided into a 16×16 pixel grid, with each cell covering 256 square pixels. The number of grain flow pixels is counted using a pixel counting method. Under normal circumstances, the number of pixels in each cell ranges from 180 to 220, and the volume density value distribution is concentrated, with a difference of no more than 15%. This step lays the data foundation for quantifying the diffusion intensity.

[0049] In an embodiment of the present invention, the evaluation of diffusion intensity is the core of judging uniformity. The volume density value is normalized to the maximum and minimum values, and the result is linearly mapped to the range of 0 to 1 to generate a diffusion intensity value. Taking actual observation as an example, when the diffusion is uniform, the diffusion intensity distribution diagram is circular or elliptical, gradually changing from the center to the edge. The Sobel operator is used to calculate the diffusion intensity gradient in the horizontal and vertical directions. Under normal conditions, the ratio of the gradients in the two directions is close to 1, indicating that the diffusion direction is balanced. If the gradient ratio deviates significantly, it indicates that the diffusion pattern is abnormal.

[0050] S1042. Calculate the diffusion intensity change rate and overall uniformity value based on the diffusion intensity distribution map. If the diffusion irregularity parameter exceeds the preset uniformity threshold, the connected domain marking algorithm is used to identify the gradient abnormality area and extract the boundary curve of the diffusion irregularity area, thereby quantifying the irregular range to support optimization decisions. It can be understood that the diffusion intensity change rate is obtained by calculating the gradient value, and the diffusion uniformity value is obtained after weighted averaging. A uniformity value greater than 0.85 indicates better diffusion, and the irregularity parameter is calculated by comparing it with the ideal value of 0.95. If the parameter exceeds the threshold, for example, it reaches 0.2 in a certain detection, the connected domain marking is activated. Irregular areas are divided based on the standard of gradient exceeding the mean by 20%. Three abnormal areas were found in a certain detection, with the largest area of 1200 square pixels. The boundary tracking algorithm extracts a sequence of contour points containing 80 to 120 points, and the constructed boundary curve clearly describes the spatial characteristics of the irregular area.

[0051] In practical applications, evenly diffused grain flow ensures balanced coverage of the silo floor and avoids local overloads. Tests in a grain silo showed that timely marking of irregular areas facilitated adjustment of conveying angles, improving stacking uniformity by approximately 10%. This step uses multi-level analysis, from boundary extraction to intensity distribution and then to irregularity marking, to fully characterize diffusion behavior, providing solid support for dynamic monitoring. The threshold and grid size can be adjusted to optimize the effect based on scenario requirements.

[0052] S105. During the grain accumulation process, a dynamic diffusion model is established based on trajectory information. Combined with the velocity and direction fields of the grain flow, the model predicts the accumulation velocity and morphological trends of the grain pile. Neural network training and grid analysis are used to generate a height field function to quantify accumulation characteristics and optimize model parameters. The dynamic diffusion model uses trajectory information to predict grain flow diffusion behavior and its impact on grain pile morphology. The model is trained using historical data to ensure that predictions closely match the actual accumulation process, providing a scientific basis for real-time adjustments.

[0053] S1051. A two-layer feedforward neural network model is constructed based on the amplitude and direction data from the trajectory change information and trained using historical trajectory data to obtain diffusion rate and direction parameters. The model then divides the irregular diffusion area into a uniform grid and calculates the rate of change in bulk density to support morphological prediction. In this embodiment of the present invention, the two-layer feedforward neural network contains 10 input nodes, corresponding to the five characteristic dimensions of amplitude and direction. After processing through two hidden layers, it outputs predicted diffusion rate and direction values. For a granary, for an input drop velocity of 2.5 meters per second, the model predicts a diffusion rate of 0.8 to 1.2 meters per second, with a direction angle less than 15 degrees. Training data is derived from historical trajectories, ensuring the model's adaptability to various scenarios. Next, the irregular area is divided into a 16×16 pixel grid, and the change in bulk volume is calculated using the velocity and direction fields. Under normal circumstances, the bulk volume difference between adjacent grids is approximately 20%, but can reach 50% in uneven areas. The density growth rate in the center is approximately twice that of the edges. This grid-based analysis provides detailed data for subsequent predictions.

[0054] For example, based on the rate of change of stacking density, the bilinear interpolation method is used to reconstruct the grid cell density distribution and generate the height field function. In a 400×400 pixel area, the height field after interpolation is bell-shaped, with a center height of 1.8 meters and gradually lower at the edges. Bilinear interpolation uses neighborhood weighted calculations to ensure a natural density transition, and its smoothness is better than that of a simple linear method. Subsequently, contour lines of different heights are generated through an isoline extraction algorithm, and the stacking angle decreases from 35 degrees at the bottom to 25 degrees at the top, which is consistent with the characteristics of the grain repose angle, adding a physical basis to the morphological description.

[0055] S1052. Perform spatial integral calculation based on the height field function to obtain the function of the grain pile volume changing with time and derive the derivative to obtain the predicted value of the stacking velocity. At the same time, smooth the contour curve through morphological processing and optimize the model in combination with actual data to improve the prediction accuracy.

[0056] In an embodiment of the present invention, spatial integration converts the height field into a volume change function. Taking a certain observation as an example, when the height of the grain pile reaches 1.5 meters, the volume growth rate is about 0.8 cubic meters per second. The predicted value of the stacking velocity obtained after derivation reflects the stacking dynamics. The contour curve is smoothed by opening and closing operations with a radius of 5 pixels, and the horizontal expansion area is calculated after removing the noise. The initial growth rate is about 0.5 square meters per second, and later it is reduced to 0.2 square meters per second. After comparing the predicted value with the actual measurement, the model is optimized through back propagation. After 50 iterations, the error is reduced to within 8%, which improves the reliability of the prediction.

[0057] In real-world scenarios, the model's predicted volume and area trends reveal asynchrony in accumulation, such as rapid growth at the center and slower expansion at the edges. This characteristic suggests the need to adjust conveying parameters. Through refined meshing and neural network analysis, a complete process from trajectory data to morphological prediction is achieved, providing an efficient tool for real-time monitoring and optimization of grain storage. The number of network layers and mesh accuracy can be adjusted to meet specific needs to further enhance performance.

[0058] S106. During the grain pile accumulation process, the operating parameters of the conveyor equipment, including conveyor belt speed and angle, are adjusted based on the predicted accumulation speed and shape change trends. Adaptive control algorithms and fuzzy reasoning are used to generate control command sequences to ensure the equipment responds smoothly to accumulation requirements. Adjusting conveyor parameters is an integral part of dynamic monitoring and directly impacts the optimization of grain pile shape. By combining predicted data with equipment characteristics, precise control is achieved.

[0059] S1061. Based on the predicted grain pile stacking speed and morphological characteristic data, an adaptive control algorithm is used to construct a conveyor belt parameter response function and generate speed adjustment and angle adjustment values. At the same time, a limit function is used to constrain the adjustment range to ensure the safe operation of the equipment.

[0060] In an embodiment of the present invention, the adaptive control algorithm adopts a three-layer neural network structure. The input layer receives the stacking velocity and morphological characteristics, the hidden layer uses the hyperbolic tangent activation function to process the nonlinear relationship, and the output layer generates the adjustment amount. Taking a certain grain warehouse as an example, when the stacking velocity is 0.8 meters per second, the algorithm outputs a speed adjustment amount of 0.15 meters per second and an angle adjustment amount of 2.5 degrees. To ensure safety, the conveyor belt speed is limited to 0.5 to 2.0 meters per second, and the angle is 15 to 45 degrees. If the adjustment amount exceeds, such as the speed reaches 2.2 meters per second, the limiting function will truncate it to 2.0 meters per second. This constraint mechanism avoids equipment overload and improves operational stability.

[0061] S1062. For the speed and angle adjustment amounts, a fuzzy rule reasoner is used to perform linkage analysis and divide the fuzzy sets. Subsequently, the center of gravity method is used to defuzzify and generate the adjustment step size and construct a parameter adjustment path curve to optimize the adjustment process.

[0062] In an embodiment of the present invention, the fuzzy inference engine divides speed into three fuzzy sets: low speed 0.5 to 0.8 meters per second, medium speed 0.8 to 1.2 meters per second, and high speed 1.2 to 2.0 meters per second. The angle is divided into three fuzzy sets: small angle 15 to 25 degrees, medium angle 25 to 35 degrees, and large angle 35 to 45 degrees. According to empirical rules, small angles are preferred at high speeds to slow down accumulation. After defuzzification using the center of gravity method, a speed adjustment step size of 0.1 meters per second and an angle step size of 2 degrees are generated. The adjustment path curve constructed through the step size data ensures that parameter changes are smooth and controllable, avoiding sudden changes that affect the equipment.

[0063] In an embodiment of the present invention, based on the adjustment path curve, the mean square error is used as the objective function, and the speed and angle response curves are optimized by gradient descent. The weight ratio of speed error to angle error is 6:4. The speed curve is stable after 5 iterations, and the angle curve requires 7 to 8 times. In the obtained incremental sequence, the speed change is controlled within 0.1 meters per second and the angle does not exceed 3 degrees. Subsequently, piecewise linear mapping is used to discretize the increments with 5 feature points to ensure a natural transition. For example, when the speed increases from 0.8 to 1.0 meters per second, the 5-segment mapping generates a stable sequence to avoid shock.

[0064] In this embodiment of the present invention, discrete instructions are interpolated at 0.1-second intervals using a linear interpolation algorithm, ensuring continuous changes in conveyor belt parameters. For example, in one adjustment, the speed increased from 0.8 to 1.0 meters per second and the angle was adjusted from 25 to 28 degrees. The smoothed instruction sequence resulted in a smooth equipment response and an approximately 12% improvement in stacking uniformity. This step achieves a seamless connection between prediction and control through multi-level optimization, from response functions to fuzzy reasoning and then to instruction generation, providing efficient support for grain pile morphology optimization. Fuzzy sets or the number of iterations can be adjusted based on equipment characteristics to further enhance accuracy.

[0065] S107. After adjusting the conveying equipment's operating parameters, monitor the grain flow trajectory and dispersion pattern in real time. Determine whether the grain pile uniformity has improved through height, area, and density analysis. If significant, record the optimized trajectory and dispersion pattern data to support subsequent optimization. Multi-dimensional data analysis ensures that the grain pile height standard deviation, coverage area uniformity, and density distribution meet the desired targets.

[0066] S1071. Based on the adjusted state of the conveying equipment, use an infrared height measurement sensor array to perform a grid scan on the surface of the grain pile and construct a three-dimensional height distribution map to extract height data. At the same time, use a region growing algorithm to segment the covered area and calculate the height standard deviation and uniformity parameters.

[0067] In an embodiment of the present invention, an array of infrared height sensors is arranged at intervals of 20 cm to perform a comprehensive scan of the grain pile surface. Taking a 6m×8m granary as an example, the height distribution map generated by 1,200 measurement points clearly reflects the surface morphology, with a center height of approximately 1.8 meters and a height reduction of 0.8 meters at the edge. The least squares method is used to calculate the average height, and the standard deviation is obtained by summing the deviations and dividing by the number of points. The ideal value is within 0.2 meters, and after actual optimization, it can be stabilized at 0.15 meters. The region growing algorithm based on the height gradient uses a 0.1-meter difference as a threshold to segment the covered area and generate boundary contour lines to ensure segmentation accuracy and reliability of boundary recognition.

[0068] In an embodiment of the present invention, the analysis of density and uniformity further quantifies the stacking state. The coverage area is divided into regular hexagonal grids with a side length of 40 cm, and each grid contains 4 height points, taking into account both data representativeness and computational efficiency. The laser density detector array scans vertically at a frequency of 10 Hz to generate a density distribution map. Under normal circumstances, the density difference between adjacent grids is less than 5%. After smoothing the noise through 3×3 Gaussian kernel convolution, the density gradient distribution is calculated, and the local uniformity is obtained based on the height change rate and area ratio within the grid, and then the overall uniformity parameter is generated. After optimization, the uniformity can be increased from 0.75 to above 0.9, significantly improving the stacking distribution.

[0069] S1072. If the monitoring indicators show that the height standard deviation is lower than the preset threshold and the area uniformity is higher than the target value, the grain flow trajectory sequence is collected through four cameras and a diffusion pattern database is constructed using a convolutional neural network to record the optimization results.

[0070] In this embodiment of the present invention, accumulation is considered improved when the standard deviation of height is less than 0.2 meters and the uniformity exceeds 0.85. Four cameras capture images from multiple angles at 50 frames per second, which are stitched together to form a complete trajectory sequence covering the entire process from descent to accumulation. A convolutional neural network employs a three-layer structure: the first layer extracts morphological features, the second layer analyzes motion trajectories, and the third layer integrates spatiotemporal information. Classification by a fully connected layer generates a diffusion pattern database. This structured database stores optimized data, providing a reference for subsequent regulation.

[0071] In practical applications, after adjustments were made to a grain depot, the height standard deviation decreased from 0.25 meters to 0.15 meters, the density gradient smoothness improved by approximately 15%, and the trajectory sequence showed a more uniform grain flow. The pattern data recorded in the database supports continuous iteration of parameter optimization, ensuring long-term storage stability. This step, through the collaboration of sensors and algorithms, from scanning to analysis and recording, enables comprehensive effect evaluation and experience accumulation. Thresholds or grid density can be adjusted based on the scenario to optimize monitoring accuracy.

[0072] Obviously, those skilled in the art may make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application also intends to include such modifications and variations.

Claims

1. A method for dynamic monitoring of grain pile morphology based on an intelligent visual system, characterized in that: The method comprises: The optical flow analysis method is used to extract the velocity and direction fields of the grain flow from the video data of the falling grain flow, and the initial information of the grain flow trajectory is obtained. According to the velocity field and direction field of the grain flow movement, if the rate of change of the velocity field or direction field in a local area exceeds a preset rate threshold, it is determined that there is a sudden change in the grain flow trajectory, and the area with the sudden change is marked as a trajectory sudden change area; The local optical flow field analysis method is used to refine the velocity field and direction field in the trajectory mutation area to obtain trajectory change information. The trajectory change information includes the change amplitude, change direction, change duration and change area range; Based on the trajectory change information, the initial diffusion range of the grain flow at the bottom of the silo is analyzed to determine whether the diffusion pattern is uniform. If the diffusion is uneven, the corresponding area is marked as an irregular diffusion area. A dynamic diffusion model is established based on trajectory change information. The model is used to process irregular diffusion areas and, combined with the velocity and direction fields of grain flow, the accumulation velocity and morphological change trend of grain piles are predicted. Adjust the operating parameters of the conveying equipment based on the predicted grain pile accumulation speed and shape change trend, including the conveyor belt speed and angle; Monitor the grain flow trajectory and diffusion pattern of the adjusted conveying equipment in real time to determine whether the grain pile height standard deviation, coverage area uniformity and stacking density distribution have been improved. If so, record the final grain flow trajectory and diffusion pattern data.

2. The method according to claim 1, characterized in that The optical flow field analysis method is used to extract the velocity field and direction field of the grain flow motion from the video data of the grain flow falling process, and the initial information of the grain flow trajectory is obtained, including: The grain flow edge contour area is identified by region matching method, and the grain flow area grid unit division data and grain flow density distribution map are obtained; Establishing a grayscale value gradient constraint equation and a brightness constant constraint equation according to the grain flow density distribution map, and solving the grayscale value gradient constraint equation and the brightness constant constraint equation by a multi-scale iterative method to obtain grain flow optical flow vector field distribution data; Calculating the grain flow velocity vector in the grid unit based on the grain flow optical flow vector field distribution data, and using the least squares method to fit the velocity vector field to obtain the overall grain flow movement trend data; constructing a quadratic polynomial fitting function based on the overall grain flow movement trend data, and calculating geometric parameters of the grain flow area and diffusion radius from the function coefficients; The Euclidean distance calculation method is used to compare the changes in geometric parameters between the previous and next frames to obtain the grain flow motion parameters and the initial information of the grain flow falling trajectory.

3. The method according to claim 1, characterized in that The method of determining that a mutation exists in the grain flow trajectory based on the velocity field and direction field of the grain flow movement and the change rate of the velocity field or the direction field in a local area exceeds a preset change rate threshold is as follows: The grain flow area is uniformly gridded according to the grain flow velocity field data, and the first-order difference values of the velocity data point set in the grid unit in the horizontal and vertical directions are calculated using the Sobel operator, and the velocity field change rate matrix is obtained from the first-order difference values; For the direction vector set in the grid unit, a three-point fitting method is used to calculate the angle change between adjacent direction vectors, and a direction field change rate matrix is obtained from the angle change; Performing dual-threshold detection on the velocity field change rate matrix and the direction field change rate matrix, and marking the grid unit as a candidate mutation point if the velocity field change rate exceeds a preset velocity threshold or the direction field change rate exceeds a preset direction threshold; A region growing algorithm is used to perform connectivity analysis on the candidate mutation points, a boundary tracking algorithm is used to obtain a region boundary point sequence, and a piecewise cubic spline curve is constructed using the boundary point sequence to obtain the trajectory mutation region.

4. The method according to claim 1, wherein The local optical flow field analysis method is used to refine the velocity field and direction field in the trajectory mutation area to obtain trajectory change information. The trajectory change information includes the change amplitude, change direction, change duration and change area range, including: The trajectory mutation area is divided into local grids, and the pyramid optical flow algorithm is used to obtain the velocity vector and direction vector sequence from the divided grids; A speed change curve is fitted using a cubic polynomial least squares method according to the speed vector sequence, and a speed difference between adjacent frames is calculated using the speed change curve to obtain a speed change amplitude parameter and a direction change angle parameter; Calculating the speed change amplitude parameter and the direction change angle parameter using a weighted average method, and obtaining a motion change intensity value from the calculated result; A region centroid calculation method is used according to the motion change intensity value to obtain a coordinate sequence of the center point of the mutation region, and a region boundary contour is constructed through the coordinate sequence to obtain a trajectory change feature vector, which is used to represent the trajectory change information.

5. The method according to claim 1, wherein The initial diffusion range of the grain flow at the bottom of the silo is analyzed based on the trajectory change information to determine whether the diffusion pattern is uniform. If the diffusion is uneven, the corresponding area is marked as an irregular diffusion area, including: The grain flow coverage area at the bottom of the warehouse is segmented using the eight-neighborhood region growing algorithm, and the initial diffusion range boundary point sequence of the warehouse bottom is obtained by using grayscale similarity as the growth condition. Divide the initial diffusion range boundary point sequence of the silo bottom into a uniform grid of pixel size, and use a pixel counting method to obtain the number of grain flow pixels in the grid unit to obtain the volume density value; The volume density values of the grid cells are normalized to their maximum and minimum values, and the diffusion intensity values are obtained through the linear mapping method to construct a diffusion intensity distribution map. If the diffusion irregularity parameter in the diffusion intensity distribution map exceeds a preset uniformity threshold, a connected domain marking algorithm is used to mark the region where the diffusion intensity gradient is greater than the gradient mean to obtain the boundary of the diffusion irregular region.

6. The method according to claim 1, characterized in that The dynamic diffusion model is established based on the trajectory change information, and the irregular diffusion area is processed by the dynamic diffusion model. The velocity field and direction field of the grain flow are combined to predict the accumulation speed and morphological change trend of the grain pile, including: Training a two-layer feedforward neural network model based on historical trajectory data, and obtaining food diffusion rate parameters and diffusion direction parameters from the neural network model; For the diffusion rate parameter and the diffusion direction parameter, the irregular area is divided by a uniform grid, and the packing density change rate data is calculated from the grid cells; Reconstructing the grid cell density distribution by bilinear interpolation according to the bulk density change rate data, and constructing the grain pile height field function from the density distribution; A spatial integral operation is performed on the grain pile height field function to obtain a function of the grain pile volume changing with time, and a predicted value of the stacking velocity is obtained by taking the derivative of the changing function.

7. The method according to claim 1, characterized in that The operation parameters of the conveying equipment are adjusted according to the predicted grain pile accumulation speed and shape change trend. The operation parameters include the conveyor belt speed and angle, including: According to the accumulation speed and morphological characteristic data, an adaptive control algorithm is used to establish a conveyor belt parameter response function, and the speed adjustment amount and the angle adjustment amount are obtained from the response function; A fuzzy rule reasoner is used to perform parameter linkage analysis on the speed adjustment amount and the angle adjustment amount. The fuzzy rule reasoner divides the speed variable into a low-speed fuzzy set, a medium-speed fuzzy set, and a high-speed fuzzy set, and divides the angle variable into a small-angle fuzzy set, a medium-angle fuzzy set, and a large-angle fuzzy set. According to the fuzzy inference results, the centroid method is used to perform defuzzification processing, and the speed adjustment step length and angle adjustment step length are obtained from the defuzzification processing results to construct the parameter adjustment path curve; With respect to the parameter adjustment path curve, a piecewise linear mapping method is used to discretize the belt speed increment sequence and the angle increment sequence, and a conveyor belt control instruction sequence is obtained from the discretization results.

8. The method according to claim 1, characterized in that The real-time monitoring of the grain flow trajectory and diffusion pattern of the adjusted conveying equipment determines whether the grain pile height standard deviation, coverage area uniformity, and stacking density distribution of the grain pile have been improved. If so, the final grain flow trajectory and diffusion pattern data are recorded, including: An infrared height measurement sensor array is used to perform a grid scan on the surface of the grain pile, a three-dimensional height distribution map of the grain pile is constructed based on the scanned data, and height data of the measurement points are obtained from the three-dimensional height distribution map of the grain pile; According to the three-dimensional height distribution map of the grain pile, a region growing algorithm based on height gradient is used to segment the covered area of the grain pile, and a contour line of the covered area is obtained by region merging; For the outline of the coverage area, the area is divided into regular hexagonal grids and density scanning is performed using a laser density detector array, and a stacking density distribution map is obtained from the scanned data; If the height standard deviation index is lower than the preset standard deviation threshold and the area uniformity parameter is higher than the preset uniformity threshold, a trajectory sequence is acquired using four cameras, and a diffusion pattern database is constructed using a convolutional neural network based on the trajectory sequence.

Images