Silo external spiral chute installation control system and method based on machine vision

Abstract

The invention relates to the technical field of large silo equipment installation, and discloses a silo external spiral chute installation control system and method based on machine vision, and the system comprises an image collection unit which is used for collecting target images fixed on two butt joint flange plates in real time; the image processing unit is used for resolving the spatial attitude of the target image according to the target image, and calculating the pure torsion angle deviation of the two flange plates around the theoretical butt joint axis according to the spatial attitude; the calculation unit is used for comparing the deviation with a preset threshold value and generating a control instruction; and the execution unit is used for driving the to-be-mounted chute section to rotate according to an instruction so as to correct deviation, and high-precision and automatic detection and real-time correction of the axial torsion angle deviation of the flange plate are realized through non-contact vision measurement, a special torsion angle calculation algorithm and automatic closed-loop control. The problems that traditional manual installation is low in efficiency and poor in precision and depends on experience are solved, and the installation quality and efficiency of the spiral chute are improved.

Description

Technical Field

[0001] This invention relates to the technical field of large silo equipment installation, and discloses a machine vision-based control system and method for installing external spiral chute in silos. Background Technology

[0002] As a core component of external material conveying systems for silos, spiral chutes are typically assembled on-site from multiple sections connected by flange bolts. The quality of installation directly impacts the smoothness of material conveying and the lifespan of the equipment. Traditional installation methods have several shortcomings. For example, traditional installation relies entirely on the visual observation of installers and the use of simple tools for flange alignment, which is highly subjective and cannot provide real-time, synchronous, and accurate observation of two dynamically suspended flanges in different spatial positions. This is especially inefficient and prone to large errors in industrial environments with poor lighting or limited viewing angles. Existing mechanical measurements can only measure the center alignment or end-face distance of the flanges, failing to effectively separate and quantify the pure torsional angular deviation between the two flanges rotating around their common axis. This pure torsional angular deviation is a direct cause of bolt hole misalignment, poor sealing, and the formation of steps within the chute that obstruct material flow. Traditional techniques lack dedicated, efficient, and non-contact measurement methods for this specific degree of freedom deviation. Adjustments during installation rely entirely on worker experience and judgment. Workers adjusting based solely on intuition are prone to under-adjustment or over-adjustment, resulting in high labor intensity, long installation cycles, and difficulty in ensuring stable deviations. Summary of the Invention

[0003] To address the aforementioned technical problems, the main objective of this invention is to provide a machine vision-based installation control system for external spiral chute installation in silos. This machine vision-based system includes: The image acquisition unit includes a first image sensor and a second image sensor, which acquires images of the first target and the second target of the first chute segment flange and the second chute segment flange in real time. The image processing unit is communicatively connected to the image acquisition unit. It calculates the attitude information of the first target and the second target in space in real time using the images of the first target and the second target, and calculates the real-time torsional angle deviation value of the first flange relative to the second flange about its theoretical docking axis using the attitude information of the first target and the second target. The calculation unit compares the real-time torsion angle deviation value with a preset threshold range and generates control commands; The execution unit receives the control command and drives the first chute segment to rotate around its axis to adjust the torsion angle deviation.

[0004] As a preferred embodiment of the machine vision-based installation control of the external spiral chute of the silo in this invention, wherein: The first image sensor continuously acquires images of the first target fixed on the first flange during the installation process; The second image sensor continuously acquires images of the second target fixed on the second flange during installation.

[0005] As a preferred embodiment of the machine vision-based installation control of the external spiral chute of the silo in this invention, wherein: The image processing unit identifies multiple feature points in the target image and obtains their image coordinates; By using the physical dimensions of the target and the known three-dimensional coordinates of each feature point, a correspondence between the image coordinates and the three-dimensional coordinates of the feature points is established. The perspective projection model is solved by the correspondence to obtain the transformation relationship of each target from the body coordinate system to the camera coordinate system; The parameters characterizing the planar spatial position and orientation of each target are determined and output through the transformation relationship.

[0006] As a preferred embodiment of the machine vision-based installation control of the external spiral chute of the silo in this invention, wherein: The image processing unit determines the theoretical docking axis using pre-stored models of the first chute segment and the second chute segment; Obtain the theoretical center positions of the first flange and the second flange from the design model; In the system coordinate system, a straight line connecting the two theoretical center positions and parallel to the silo's central axis is defined as the theoretical docking axis.

[0007] As a preferred embodiment of the machine vision-based installation control of the external spiral chute of the silo in this invention, wherein: The image processing unit is used to calculate the real-time torsion angle deviation value; The first normal vector and the second normal vector are obtained from the planar parameters of the first target and the second target, respectively. The first normal vector and the second normal vector are projected onto a plane perpendicular to the theoretical docking axis to obtain the first projection vector and the second projection vector, respectively. The angle between the first projection vector and the second projection vector in the vertical plane is calculated. The angle is the real-time torsion angle deviation value.

[0008] As a preferred embodiment of the machine vision-based installation control of the external spiral chute of the silo in this invention, wherein: The image processing unit further sets a first threshold and a second threshold that is less than the first threshold; when the absolute value of the torsion angle deviation is greater than the first threshold, it outputs a continuous rotation control signal; when the absolute value of the torsion angle deviation is not greater than the first threshold but greater than the second threshold, it outputs a pulse rotation control signal; when the absolute value of the torsion angle deviation is not greater than the second threshold and continues for a preset duration, it outputs a lock signal and triggers a prompt.

[0009] As a preferred embodiment of the machine vision-based installation control of the external spiral chute of the silo in this invention, wherein: The calculation unit compares the real-time torsion angle deviation value with a preset threshold range; Establish a dynamic data sequence that includes the current deviation value and multiple historical deviation values; The dynamic data sequence is weighted and calculated to generate a comprehensive deviation evaluation value; The comprehensive deviation evaluation value is compared with a preset static threshold to determine the adjustment stage of the system.

[0010] As a preferred embodiment of the machine vision-based installation control of the external spiral chute of the silo in this invention, wherein: The computing unit also generates control commands based on the comparison results, including: The first type is a continuous rotation control command, used to drive the execution unit to rotate at a constant speed; The second type is pulse-type micro-motion control commands, which are used to drive the execution unit to perform intermittent angular displacement adjustments; The third type is the lock and confirmation command, which is used to stop the execution unit and issue an indication signal that alignment is complete.

[0011] As a preferred embodiment of the machine vision-based installation control of the external spiral chute of the silo in this invention, wherein: The execution unit includes a servo rotary drive mechanism mounted on the first chute segment hoisting mechanism; the servo rotary drive mechanism receives control commands issued by the image processing and control unit, and converts them into the corresponding drive mode according to the command type, driving the first chute segment to rotate precisely around its own geometric axis to correct the real-time torsion angle deviation value.

[0012] A preferred embodiment of a machine vision-based method for controlling the installation of external spiral chute in silos, wherein: The image acquisition unit acquires images of the first target and the second target, which are fixedly set on the first flange and the second flange, in real time. The image is processed by the image processing and control unit to calculate the attitude information of the first target and the second target in space in real time. Based on the attitude information of the first target and the second target, calculate the real-time torsional angle deviation of the first flange relative to the second flange around the theoretical docking axis; The real-time torsion angle deviation value is compared with a preset threshold, and a control command is generated based on the comparison result; The control command is sent to the execution unit to drive the first chute segment to rotate around its axis in order to adjust the torsion angle deviation.

[0013] The beneficial effects of this application are: This application uses a target and an image sensor to convert the spatial attitude of the flange, which is difficult to measure directly, into processable image information. By calculating and separating the projection angle of the normal vectors of the two flange target planes onto the plane perpendicular to the theoretical docking axis, the deviation of the unique rotational degree of freedom that needs to be controlled can be quantified. This effectively filters out the translational, tilting and other interference movements that accompany the installation process, making the measurement target extremely clear and accurate.

[0014] This application automatically determines the current state and generates control commands by comparing a preset threshold with the real-time deviation. This improves the success rate and final accuracy of adjustments.

[0015] This application uses control commands to automatically drive a servo rotating mechanism to rotate the chute segments, achieving fine adjustments of minute angles. This not only significantly reduces the labor intensity and skill requirements for workers, but also makes the entire adjustment process fast, stable, and repeatable, improving the overall safety and standardization of the installation operation. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating the overall workflow of the machine vision-based installation control system for external spiral chute in silos, as described in this invention. Figure 2 This is a flowchart of the target detection and recognition process of the machine vision-based installation control system for external spiral chute in silos according to the present invention. Figure 3 This is a flowchart of the hierarchical threshold control and instruction generation in the machine vision-based silo external spiral chute installation control system of the present invention; Figure 4 This is a flowchart of the machine vision-based installation control method for external spiral chute in silos according to the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0018] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0019] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0020] Example 1

[0021] like Figure 1 As shown, a machine vision-based installation control system and method for external spiral chute in silos includes: The image acquisition unit includes a first image sensor and a second image sensor, which acquires images of the first target and the second target of the first chute segment flange and the second chute segment flange in real time. The first image sensor continuously acquires images of the first target fixed on the first flange during the installation process; The second image sensor continuously acquires images of the second target fixed on the second flange during installation.

[0022] In this application, a preferred implementation method of an image acquisition unit includes: The image acquisition unit may include two independent image sensors and two visual targets.

[0023] A preferred image sensor includes a first image sensor and a second image sensor; The first and second image sensors are fixedly mounted on the installation platform, located outside the working area of ​​the spiral chute to be docked. During installation, the spatial position and pointing angle of each sensor are adjusted to ensure that the field of view completely covers the movement range of the corresponding target throughout the entire expected adjustment stroke, and that the field of view is not obstructed by the lifting equipment or ropes throughout the installation process.

[0024] Furthermore, the first and second targets are rigid planar marking plates with high contrast and specific preset geometric patterns.

[0025] Furthermore, the first target is securely installed at a predetermined position on the outer edge or end face of the first chute segment flange via a rigid connector; the second target is installed at the corresponding predetermined position on the second chute segment flange in the same manner. The installation positions of the two targets are asymmetrically distributed in the circumferential direction.

[0026] The image acquisition unit is activated, and the first image sensor continuously focuses on the first target, acquiring a sequence of digital images of the first target at a set sampling frequency.

[0027] The second image sensor synchronously focuses on the second target and continuously acquires digital image sequences of the second target at the same sampling frequency.

[0028] The first and second image sensors transmit the acquired image data streams to the image processing and control unit in real time via data cables or industrial wireless networks.

[0029] The image processing unit is communicatively connected to the image acquisition unit. It calculates the attitude information of the first target and the second target in space in real time using the images of the first target and the second target, and calculates the real-time torsional angle deviation value of the first flange relative to the second flange about its theoretical docking axis using the attitude information of the first target and the second target. The image processing unit identifies multiple feature points in the target image and obtains their image coordinates; The image processing unit receives two image data streams from the image acquisition unit. For either the first target image or the second target image, the image processing unit first runs a target detection and recognition algorithm, such as... Figure 2 As shown.

[0030] Specifically, the target detection and recognition algorithm locates the target region in the image using a pre-stored target pattern template. Further, the algorithm identifies multiple pre-defined non-collinear feature points within the target pattern, such as corner points, center points of circles, or center points of coded markers. These non-collinear feature points are explicitly defined during target design, and the coordinates of each non-collinear feature point in the image's two-dimensional coordinate system are calculated.

[0031] A preferred target detection and recognition algorithm implementation method includes: First, each frame of the input image is preprocessed, including: Digital image processing techniques such as Gaussian filtering or median filtering can be used to suppress inherent random noise and environmental interference from image sensors.

[0032] Furthermore, an adaptive histogram equalization enhancement method with contrast limitation is applied to improve the overall image contrast and ensure that the target pattern is clearly visible in both bright and dark areas.

[0033] If the target uses a specific color with high saturation, convert the image from the RGB color space to a color space such as HSV or Lab.

[0034] After preprocessing, the target detection and recognition algorithm performs preliminary target detection and localization: By leveraging the high contrast between the target and the background, such as a black pattern on a white background, the target detection and recognition algorithm binarizes the image using an adaptive thresholding method, dividing the image into two categories of pixels: target and background.

[0035] Furthermore, in the binary image, the target detection and recognition algorithm finds the contours of all connected regions, and then filters the contours based on the geometric characteristics according to the pre-known physical shape of the target, such as rectangle, circle or specific polygon.

[0036] Specifically, the screening criteria include the area range of the contour, the number of vertices of the approximate polygon of the contour, the aspect ratio of the contour's circumscribed rectangle, and the roundness of the contour. Through multi-condition screening, one or more candidate target regions are initially identified.

[0037] Within the coarsely located candidate region, the target detection and recognition algorithm identifies the target to confirm its identity and locate its internal feature points: Within the candidate region, the target detection and recognition algorithm uses pre-stored target pattern templates generated from different perspectives. By calculating similarity measures such as normalized cross-correlation coefficients or squared differences, it performs sliding window matching. The position with the highest matching degree is identified as the target, and a preliminary pose estimate is obtained at the same time.

[0038] A preferred target detection and recognition algorithm extracts local feature descriptors of candidate regions, matches the descriptors with a database of feature descriptors of target standard patterns that have been extracted and stored offline in advance, and confirms the target identity by matching feature point pairs and eliminating false matches using a random sampling consensus algorithm. It also directly obtains a high-precision correspondence between the feature points of the standard pattern and the corresponding points in the current image.

[0039] If the target uses a binary encoded pattern, the located area will be divided into grids and decoded to read its unique ID. By verifying whether the decoded ID is consistent with the expected target ID, the uniqueness of the target can be confirmed and its identity can be identified.

[0040] Once the target is accurately identified and located, the target detection and recognition algorithm then proceeds to extract the coordinates of preset feature points: The multiple feature points are geometric elements in the pattern that are easy to detect and have precise positions, such as: vertices of a specific polygon, the centers of concentric circles, the intersections of cross lines, or the boundary points of black and white squares in a coded mark.

[0041] The target detection and recognition algorithm outputs two-dimensional image coordinates (u,v) for each successfully identified feature point, and ensures that the coordinates form a one-to-one matching pair with the pre-stored three-dimensional coordinates (X,Y,Z) of the feature point in the target body coordinate system.

[0042] By using the physical dimensions of the target and the known three-dimensional coordinates of each feature point, a correspondence between the image coordinates and the three-dimensional coordinates of the feature points is established. The image processing unit has a pre-stored physical size database for each target, which contains the precise three-dimensional coordinates of each identified feature point in the three-dimensional coordinate system of the target body.

[0043] The three-dimensional coordinates are uniquely determined by the design size of the target and the layout of the feature points. For each identified feature point, the system simultaneously possesses its two-dimensional image coordinates and the corresponding three-dimensional body coordinates, establishing a corresponding matching pair of image point spatial points.

[0044] The perspective projection model is solved by the correspondence to obtain the transformation relationship of each target from the body coordinate system to the camera coordinate system; By establishing multiple sets of corresponding matching pairs, the image processing unit constructs and solves a perspective projection model, which describes the mathematical relationship of how three-dimensional points are projected onto a two-dimensional imaging plane through the camera lens. Solving the perspective projection model outputs the optimal rigid body transformation, including rotation and translation, so that the error between the known three-dimensional coordinates of the feature points on the target, after being transformed and projected onto the image plane, is minimized and the actual detected image coordinates are minimized.

[0045] Furthermore, the solution yields a transformation matrix, which describes the rotation and translation relationship between the target body coordinate system and the corresponding image sensor coordinate system, i.e., the spatial pose of the target relative to the camera.

[0046] In this application, a preferred implementation method of a perspective projection model includes: The perspective projection model solves for the complete three-dimensional spatial pose of the target relative to the camera by finding the correspondence between the image coordinates of the target feature points and their known three-dimensional coordinates.

[0047] Define the coordinate system clearly: the origin Oc of the camera coordinate system is located at the optical center of the camera, and the Zc axis coincides with the optical axis and points in the shooting direction.

[0048] The origin o of the image pixel coordinate system is located at the upper left corner of the image, and the u-axis and v-axis are parallel to the rows and columns of the image, respectively.

[0049] The origin Ot and axis system of the target body coordinate system are predefined according to the target design. For example, the origin is located at the center of the target, and the Xt and Yt axes are located in the target plane.

[0050] Perspective projection models are used to construct projection chains from the target body coordinate system to the image pixel coordinate system.

[0051] The perspective projection model describes the target's posture in front of the camera as a rigid body transformation involving rotation and translation. This rigid body transformation can be represented by a 3x3 rotation matrix R and a 3x1 translation vector T. R and T are the unknown parameters to be solved in the model.

[0052] The coordinates of point P in the target body coordinate system are: , .

[0053] The coordinates of point P in the camera coordinate system T is the translation vector; T is used to correct the orientation of the target, and T is used to correct the position of the target.

[0054] Furthermore, for each successfully identified feature point i on the target, it is known that: The three-dimensional coordinates of feature point i in the body coordinate system are: .

[0055] The two-dimensional coordinates of feature point i in the image pixel coordinate system are (ui, vi).

[0056] Based on the perspective projection model, the following projection equation is established: ; ; Where (Xci, Yci, Zci) are the camera coordinates of feature point i after R and T transformations; The camera coordinates are .

[0057] fx, fy, cx, cy are the camera's intrinsic parameters, including focal length and principal point coordinates. These parameters have been accurately measured and stored through an independent camera calibration process before the system is used, and are considered as known constants in this model.

[0058] By substituting the correspondence of each feature point, we can obtain a set of equations about the unknowns R and T.

[0059] Furthermore, since each point provides two equations u and v, at least four non-coplanar feature points are needed to solve the pose with six degrees of freedom, namely three rotations and three translations.

[0060] Furthermore, the image processing unit uses numerical algorithms to solve the above nonlinear equations to obtain the optimal R and T. By introducing intermediate variables, the problem is transformed into solving a linear equation system to obtain an initial solution, which is then optimized through iteration.

[0061] After solving for the rotation matrix R and translation vector T, the transformation relationship from the target coordinate system to the camera coordinate system is obtained. The image processing unit can further extract easily understood and used geometric parameters from this transformation relationship: Position: The translation vector T itself represents the three-dimensional position of the target origin in the camera coordinate system. .

[0062] Normal vector: Since the target is a plane, its plane normal vector is known in the target body coordinate system, for example... The normal vector is transformed using the rotation matrix R: This gives us the direction (Nx, Ny, Nz) of the normal vector in the camera coordinate system. The normal vector is the parameter that represents the orientation of the target plane in space.

[0063] The parameters characterizing the planar spatial position and orientation of each target are determined and output through the transformation relationship.

[0064] After obtaining the transformation matrix, the image processing unit further extracts geometric parameters used to characterize the spatial state of the target plane.

[0065] Specifically, since the target is a rigid plane, the equation of the rigid plane in space can be uniquely determined by a three-dimensional spatial point and a three-dimensional vector, where the three-dimensional spatial point is the position and the three-dimensional vector is the normal vector.

[0066] Specifically, the translation vector T directly contains the coordinates of the origin of the target coordinate system in the camera coordinate system. The design origin of the target is defined on a physical point on the target plane, such as its geometric center, a specific corner point, or a pattern reference point.

[0067] Furthermore, the image processing unit directly reads the three components of the translation vector T. These three components That is, it is resolved into the three-dimensional spatial coordinates of the target origin in the camera coordinate system.

[0068] The image processing unit parses the three-dimensional position coordinates of the target plane center or a specific reference point in the camera coordinate system from the transformation matrix, and at the same time calculates the direction of the unit normal vector perpendicular to the target plane in the camera coordinate system.

[0069] The target is a rigid plane, and in its body coordinate system, the direction of its normal vector is a pre-known fixed value. Typically, the target plane is defined as a plane where Zt=0, then its unit normal vector in the body coordinate system is... .

[0070] Furthermore, the image processing unit applies the rotation matrix R to the known unit normal vector of the body coordinate system. .

[0071] A preferred specific calculation method is vector-matrix multiplication. Calculation results A 3D vector represents the direction of the same normal vector in the camera coordinate system.

[0072] Furthermore, since the rotation matrix R is an orthogonal matrix, the magnitude of the result after an orthogonal matrix is ​​applied to a unit vector is 1. To eliminate the small errors that may be introduced in the numerical calculation and to ensure the accuracy of subsequent calculations, the image processing unit... Perform normalization.

[0073] Normalization is achieved by calculating the magnitude of the vector and then dividing each component of the vector by the magnitude.

[0074] Furthermore, the image processing unit executes the above process in parallel, simultaneously outputting two sets of independent planar spatial parameters.

[0075] For the measurement branch consisting of the first image sensor and the first target, the image processing unit executes the complete process described above: input the feature point data of the first target, solve for the transformation matrix R1 and T1, and then parse out the three-dimensional position coordinates P1 and the unit normal vector N1 of the first target plane.

[0076] For the branch formed by the second image sensor and the second target, the image processing unit uses the feature point data of the second target to independently solve for the transformation matrix R2 and T2, and analyzes the three-dimensional position coordinates P2 and the unit normal vector N2 of the second target plane.

[0077] The image processing unit synchronously outputs independent sets of geometric parameters {P1,N1} and {P2,N2}. The parallel processing mechanism ensures real-time and synchronous calculation of the attitudes of the two flanges, providing accurate and time-aligned input data for subsequent calculation of the relative torsion angle deviation between them.

[0078] The image processing unit determines the theoretical docking axis using pre-stored models of the first chute segment and the second chute segment; Obtain the theoretical center positions of the first flange and the second flange from the model; Furthermore, a straight line connecting the two theoretical center positions and parallel to the silo's central axis is defined as the theoretical docking axis.

[0079] Furthermore, the design 3D models of the first and second chute segments are imported into the image processing unit. These 3D models are derived from the product manufacturing drawings and include the complete design dimensions, geometry, and spatial positioning relationships of each flange.

[0080] The theoretical installation center points of the first flange and the second flange in their respective segment design models are known and clearly defined.

[0081] Furthermore, the three-dimensional coordinates of the theoretical center point of the first flange in the design coordinate system and the three-dimensional coordinates of the theoretical center point of the second flange in the design coordinate system are extracted. The two coordinate points represent the predetermined spatial position of the docking center of the two flanges under ideal design conditions.

[0082] A calibration procedure is used to establish a fixed world coordinate system, which is typically associated with the silo body or a fixed measurement reference point. Using known calibration points, a spatial transformation relationship is established between the design coordinate system and the world coordinate system. This transformation relationship is then used to convert the coordinates of the two extracted theoretical center points to the world coordinate system, obtaining their theoretical spatial positions within the current installation site's world coordinate system.

[0083] After obtaining the positions of the two theoretical center points in the world coordinate system, the axis generation logic is executed. The core of the axis generation logic is that the direction of the theoretical docking axis is not simply defined as the straight line connecting the two theoretical center points, but is forcibly defined as the direction parallel to the silo design center axis.

[0084] In this application, a preferred mandatory definition is parallel to the central axis of the silo. The specific implementation method is as follows: Obtain the direction of the central axis of the silo body in the design model and transform this direction to the world coordinate system.

[0085] In the world coordinate system, a spatial straight line is defined, the direction of which is parallel to the direction of the silo's central axis, and the position of the silo's central axis passes through one of the theoretical center points or the midpoint between the two theoretical center points.

[0086] This defined straight line is the theoretical docking axis upon which this installation control is based.

[0087] The definition of the axis is based entirely on objective design drawings, eliminating on-site measurement errors that vary from person to person. This provides a stable and accurate geometric reference for calculating the torsion angle. The defined axis is parallel to the central axis of the silo system, ensuring that the goal of installation correction is to make the chute segments rotate around the correct design axis, achieving installation that is consistent with the overall system. From data retrieval and coordinate transformation to axis generation, the entire process requires no manual intervention or input, realizing the automation of benchmark establishment and improving the overall level of the system.

[0088] The image processing unit is used to calculate the real-time torsion angle deviation value; The first normal vector and the second normal vector are obtained from the planar parameters of the first target and the second target, respectively. The first normal vector and the second normal vector are projected onto a plane perpendicular to the theoretical docking axis to obtain the first projection vector and the second projection vector, respectively. The angle between the first projection vector and the second projection vector in the vertical plane is calculated. The angle is the real-time torsion angle deviation value.

[0089] In this application, a preferred method for calculating the torsion angle deviation value includes: To obtain the normal vector and axis direction, the image processing unit first acquires three pre-calculated basic vector data: The first normal vector is used to represent the unit vector of the orientation of the first target plane in space.

[0090] The second normal vector is used to represent the unit vector of the orientation of the second target plane in space.

[0091] The theoretical docking axis direction vector is used to represent the unit vector of the spatial direction of the theoretical docking axis.

[0092] The first normal vector, the second normal vector, and the direction vector of the theoretical docking axis have been expressed in the same world coordinate system.

[0093] To eliminate the interference in angle calculation caused by the possible coplanar tilt of the two flanges, i.e., the different components of the normal vector in the axial direction, a preferred projection filtering operation transforms the normal vector to a specific two-dimensional plane for analysis.

[0094] Specifically, a virtual projection plane, or normal plane, is defined in space using the direction vector of the theoretical docking axis. The normal plane's normal direction is completely consistent with the direction of the theoretical docking axis; therefore, the normal plane is perpendicular to the theoretical docking axis.

[0095] Furthermore, the first and second normal vectors are orthogonally projected onto the virtual projection plane, respectively. The component of each original normal vector along the theoretical docking axis is subtracted from it, resulting in two new vectors: The first projection vector lies within the projection plane and is formed by removing the axial component from the first normal vector.

[0096] The second projection vector lies within the projection plane and is formed by removing the axial component from the second normal vector.

[0097] This application transforms two three-dimensional spatial vectors that may not be coplanar due to installation tilt into two two-dimensional vectors that are necessarily located in the same plane. These two projected vectors uniquely retain the rotation information of the original normal vector about the theoretical docking axis, while eliminating the tilt information along the axis direction.

[0098] Calculation of the included angle in the projection plane: After obtaining the first projection vector and the second projection vector, the calculation of the torsion angle deviation is simplified to solving the included angle between the two vectors in the same plane.

[0099] Specifically, the image processing unit calculates the plane angle between the two projection vectors, which reflects the angular difference between the first target normal vector and the second target normal vector in the direction of rotation around the theoretical docking axis.

[0100] The difference in angle value is the required real-time torsion angle deviation value. The real-time torsion angle deviation value is used to indicate the angle by which the first flange needs to rotate around the theoretical docking axis, so that the projection of the normal vector of the target plane of the first flange in the projection plane is consistent with the corresponding projection direction of the second flange, thereby achieving alignment in the pure torsion direction.

[0101] This application effectively isolates the two coupled spatial attitude changes of torsion around the axis and tilting by projecting onto a plane perpendicular to the axis. This ensures that the measurement results only reflect the deviation of the torsional degree of freedom that needs to be controlled and corrected, avoiding misjudging tilting errors as torsional errors. The calculation process is based on linear vector operations, which is stable and reliable and is not affected by the singularity of the flange's spatial orientation. This ensures the continuity and accuracy of the output results under various installation orientations. The final deviation angle has a clear and direct physical meaning and engineering guidance value. That is, alignment can be achieved by driving the actuator to rotate the first flange around the theoretical axis by this angle.

[0102] The image processing unit further sets a first threshold and a second threshold that is less than the first threshold; when the absolute value of the torsion angle deviation is greater than the first threshold, it outputs a continuous rotation control signal; when the absolute value of the torsion angle deviation is not greater than the first threshold but greater than the second threshold, it outputs a pulse rotation control signal; when the absolute value of the torsion angle deviation is not greater than the second threshold and continues for a preset duration, it outputs a lock signal and triggers a prompt.

[0103] In this application, the image processing unit integrates hierarchical threshold control logic to generate differentiated control commands based on the real-time calculated torsion angle deviation value, driving the actuator to achieve adjustment. A preferred implementation of the control logic includes: Figure 3 As shown; When configured by the installation process, the operator or installation system automatically sets the angle threshold, including a first threshold and a second threshold; Specifically, the first threshold is a relatively large angle value, used to distinguish working conditions that require significant adjustments.

[0104] The second threshold is a finer angle value that is smaller than the first threshold, used to define the high-precision fine-tuning stage and the conditions for alignment completion.

[0105] These two thresholds divide the entire adjustment process into three distinct control phases and serve as the basis for control decisions.

[0106] The image processing unit continuously monitors the real-time torsion angle deviation value and compares the absolute value with a preset first threshold and a second threshold. Based on the comparison result, it triggers the corresponding control signal. Phase 1: Continuous rotation control signal; The trigger condition is when the absolute value of the real-time torsion angle deviation is greater than the first threshold.

[0107] The control action generates and outputs a continuous rotation control signal for the image processing unit.

[0108] Furthermore, a signal is sent to the execution unit, which instructs the first chute segment to rotate continuously at a relatively constant speed, aiming to quickly eliminate most of the angular deviation and improve coarse adjustment efficiency.

[0109] The second stage is a pulse-type rotation control signal.

[0110] The trigger condition is when the absolute value of the real-time torsion angle deviation is not greater than the first threshold, but is greater than the second threshold at the same time.

[0111] The image processing unit switches to fine-tuning mode, generates and outputs a pulsed rotation control signal. This signal is a series of discrete, short-duration drive pulses. Each pulse command execution unit generates a small, fixed-step angular displacement. The control method can effectively overcome the inertia of heavy loads, prevent repeated over-adjustment and oscillations near the target position, and achieve smooth and accurate approximation.

[0112] Third stage: Lock and confirmation signal The trigger condition is when the absolute value of the real-time torsion angle deviation is not greater than the second threshold, and this state is maintained stably for a preset duration, such as 2 seconds.

[0113] The image processing unit determines that the alignment has met the final accuracy requirements, generates and outputs a locking signal and a prompt signal. The locking signal instructs the execution unit to stop all adjustment actions and maintain the current position. The prompt signal is triggered through audible and visual alarms, display messages, etc., to inform the operator that the current segment has been aligned and subsequent operations such as bolt tightening can be performed.

[0114] This application achieves rapid coarse adjustment through a first threshold and fine adjustment through a second threshold and pulse control. While ensuring high final accuracy, it significantly shortens the overall adjustment time and is an effective technical means to solve the problems of overshoot and oscillation that are prone to occur in the fine adjustment stage of large inertia mechanical systems, thus ensuring the stability and convergence of the adjustment process.

[0115] The calculation unit compares the real-time torsion angle deviation value with a preset threshold range and generates control commands; The calculation unit compares the real-time torsion angle deviation value with a preset threshold range; Establish a dynamic data sequence that includes the current deviation value and multiple historical deviation values; The dynamic data sequence is weighted and calculated to generate a comprehensive deviation evaluation value; The comprehensive deviation evaluation value is compared with a preset static threshold to determine the adjustment stage of the system.

[0116] This application employs a dynamic evaluation mechanism in its implementation method when performing deviation comparison and decision-making through a calculation unit: During operation, the calculation unit not only acquires the current real-time torsion angle deviation value, but also continuously maintains a dynamically updated data sequence to record historical deviation data over a recent period.

[0117] Sequences are typically implemented using a first-in-first-out (FIFO) queue or a circular buffer with a fixed length, for example, storing the deviation values ​​for the most recent N sampling periods. Newly calculated real-time deviation values ​​are added to the end of the sequence, while the oldest historical data is removed, ensuring that the sequence always reflects the latest dynamic trends.

[0118] Each data point in the sequence contains the magnitude and sign of the deviation value, and may also be associated with a corresponding timestamp.

[0119] Furthermore, to avoid control command oscillations caused by single measurement jumps, the computing unit does not directly use instantaneous deviation values ​​for decision-making, but instead performs weighted calculations on the dynamic data sequence to generate a comprehensive deviation evaluation value.

[0120] Furthermore, a weight coefficient is assigned to each historical data point in the sequence. The weight allocation strategy is based on temporal proximity, such as using a time decay model, where data points closer to the current time are given higher weights to reflect the latest changes. Based on the stability of the deviation, cumulative weights are assigned to consecutive deviations in the same direction. A preferred method is to calculate an exponentially weighted moving average for the sequence data.

[0121] In this application, a preferred calculation unit performs a weighted average or weighted summation operation on all deviation values ​​in a dynamic data sequence using preset weighting coefficients, and the result is the comprehensive deviation evaluation value.

[0122] The calculation unit compares the calculated comprehensive deviation evaluation value with a preset static threshold, which is the first threshold and the second threshold defined in the hierarchical control logic.

[0123] Specifically, the comparison logic is similar to that used with instantaneous values. For example, when the comprehensive deviation evaluation value is greater than the first threshold, the system determines that it is in the coarse adjustment stage; when it is between the first threshold and the second threshold, it determines that it is in the fine adjustment stage; when it is less than or equal to the second threshold and continues for a certain period of time, it determines that it is in the alignment completion stage.

[0124] Based on the determination result, the calculation unit generates and outputs the corresponding control command, such as continuous rotation, pulse rotation, or lock signal. Since the decision is based on a smoothed composite value, it can effectively filter out misjudgments caused by brief fluctuations in visual measurement, slight vibrations in the field, or instantaneous changes in lighting, making the output of control commands more stable and reliable.

[0125] This application effectively suppresses the impact of instantaneous noise and outliers on the control system by weighted fusion of historical data, preventing unnecessary frequent start-stops or jitters of the actuator due to false alarms, avoiding frequent control mode jumps caused by instantaneous value fluctuations near the threshold boundary, and improving the stability and continuity of the entire adjustment process. This application provides a foundation for more complex evaluation algorithms through the structure of dynamic data sequences, enhancing the system's adaptability to different operating conditions.

[0126] The computing unit also generates control commands based on the comparison results, including: The first type is a continuous rotation control command, used to drive the execution unit to rotate at a constant speed; The second type is pulse-type micro-motion control commands, which are used to drive the execution unit to perform intermittent angular displacement adjustments; The third type is the lock and confirmation command, which is used to stop the execution unit and issue an indication signal that alignment is complete.

[0127] The first type of instruction is a continuous rotation control instruction. When the system determines that the comprehensive deviation evaluation value based on the dynamic data sequence is greater than the preset first threshold, it indicates that the current torsion angle deviation is large and is in the coarse adjustment stage. The calculation unit generates the first type of instruction, which aims to drive the execution unit to quickly eliminate most of the angle deviation with high efficiency.

[0128] A preferred instruction is a digital command containing the target rotation direction and preset constant speed or torque parameters. The direction is determined by the sign of the deviation value. The instruction is sent to the servo driver of the execution unit through a communication interface, such as fieldbus or Ethernet. After the driver parses the instruction, it controls the servo motor to perform continuous and uniform rotational motion in the specified direction and speed, driving the first chute segment to quickly approach the alignment position.

[0129] The second type of instruction is a pulse-type micro-motion control instruction. When the comprehensive deviation evaluation value is not greater than the first threshold but greater than the more refined second threshold, the system enters the fine-tuning stage. In order to avoid overshoot and oscillation caused by continuous drive when the large inertial load approaches the target, the calculation unit switches to generating the second type of instruction. The instruction is not a continuous speed command, but a sequence of discrete pulses. Each pulse command contains a fixed, small angular displacement and rotation direction.

[0130] The computing unit sends this series of pulses to the servo driver at a pulse frequency. The driver operates in position control mode or pulse follower mode. Each time it receives a pulse, it controls the motor to rotate a fixed small angle, and then stops, waiting for the next pulse.

[0131] The third type of instruction is the lock and confirmation instruction. When the comprehensive deviation evaluation value is not greater than the second threshold and reaches the preset stabilization time, the system determines that the torsion angle deviation has met the final installation accuracy requirements, the alignment is completed, and the third type of instruction is generated at this time.

[0132] A clear stop or enable signal is sent to the servo drive, instructing it to immediately stop all motion outputs and lock the current position to prevent accidental segment displacement due to external forces.

[0133] An independent digital output signal or communication message is used to activate the peripheral prompting device. The signal can directly drive the audible and visual alarm installed on site or pop up an alignment completion message on the human-machine interface, clearly informing the operator that it is safe to perform subsequent operations such as bolt connection.

[0134] The execution unit receives the control command and drives the first chute segment to rotate around its axis to adjust the torsion angle deviation.

[0135] The execution unit includes a servo rotary drive mechanism mounted on the first chute segment hoisting mechanism; the servo rotary drive mechanism receives control commands issued by the image processing and control unit, and converts them into the corresponding drive mode according to the command type, driving the first chute segment to rotate precisely around its own geometric axis to correct the real-time torsion angle deviation value.

[0136] The servo rotary drive mechanism is mounted on a temporary hoisting mechanism or a dedicated installation and adjustment fixture on the first chute segment. The mounting method ensures that the rotation center of the drive mechanism's output shaft is aligned with the theoretical geometric axis of the first chute segment.

[0137] A preferred servo rotary drive mechanism may include a servo motor, a high reduction ratio precision reducer, an output flange, and a brake device.

[0138] Furthermore, the output flange is rigidly connected to the outer wall or reinforcing structure of the first chute segment via a set of temporary connectors, such as adapter plates, clamps, or temporary bolts.

[0139] The servo rotary drive mechanism establishes a real-time communication connection with the image processing and control unit through its matching servo driver, and the driver continuously receives control command data packets from the control unit.

[0140] The driver parses the instruction type identifier in the data packet, which corresponds to the first, second, and third types of instructions and related parameters, such as speed, pulse count, and direction.

[0141] Based on the parsed instruction type, the driver automatically switches its operating mode internally: When a continuous rotation control command is received, the driver switches to speed control mode and generates a corresponding analog voltage or PWM signal to drive the motor based on the speed value and direction signal in the command.

[0142] When it receives a pulse-type micro-motion control command, the driver switches to position control mode or pulse sequence follower mode. It interprets each received digital pulse as a fixed angular displacement command, driving the motor to complete the angular stepping.

[0143] Upon receiving a lock and confirmation command, the driver immediately stops power output and activates the motor's electromagnetic brake. It may also switch to zero-speed position holding mode, using the motor's holding torque to lock the output shaft.

[0144] After the brake is activated, the output shaft of the mechanism is physically locked to ensure that the chute segment will not undergo unexpected angular displacement due to external force or gravity when the bolts are tightened manually.

[0145] A preferred servo driver can feed back information such as the motor's actual speed, current position, torque output, and fault status to the image processing and control unit in real time. The control unit can then monitor the execution status and, in case of an anomaly, immediately stop and issue an alarm. The mechanism itself can also be equipped with a mechanical limit device to prevent exceeding the safe rotation range.

[0146] Example 2

[0147] Machine vision-based control method for the installation of external spiral chute in silos, such as Figure 4 As shown: including: The image acquisition unit acquires images of the first target and the second target, which are fixedly set on the first flange and the second flange, in real time. The image is processed by the image processing and control unit to calculate the attitude information of the first target and the second target in space in real time. Using the attitude information of the first and second targets, the real-time torsional angle deviation of the first flange relative to the second flange around the theoretical docking axis is calculated. The real-time torsion angle deviation value is compared with a preset threshold, and a control command is generated based on the comparison result; The control command is sent to the execution unit to drive the first chute segment to rotate around its axis in order to adjust the torsion angle deviation.

[0148] It is important to note that the constructions and arrangements of this application shown in several different exemplary embodiments are merely illustrative. Although only two embodiments are described in detail in this disclosure, those who consult this disclosure will readily understand that many modifications are possible without substantially departing from the novel teachings and advantages of the subject matter described in the application. For example, variations in the size, dimensions, structure, shape, and proportions of various elements, as well as parameter values ​​(e.g., temperature, pressure, etc.), mounting arrangements, use of materials, color, orientation, etc. For instance, an element shown as integrally formed may be composed of multiple parts or elements, the position of elements may be inverted or otherwise altered, and the nature or number or position of discrete elements may be changed or altered. Therefore, all such modifications are intended to be included within the scope of the invention. The order or sequence of any process or method steps may be changed or rearranged according to alternative embodiments. Any "device plus function" clause is intended to cover the structure performing the function described herein, and not only structurally equivalent but also equivalent in structure. Other substitutions, modifications, alterations, and omissions may be made in the design, operation, and arrangement of the exemplary embodiments without departing from the scope of the invention. Therefore, the present invention is not limited to the specific embodiments, but extends to various modifications that still fall within the scope of the appended claims.

[0149] Furthermore, in order to provide a concise description of exemplary embodiments, not all features of actual embodiments (i.e., those features that are not relevant to the currently considered best mode for carrying out the invention, or those features that are not relevant to implementing the invention) may be omitted.

[0150] It should be understood that numerous specific implementation decisions can be made during the development of any practical implementation, such as in any engineering or design project. Such development efforts may be complex and time-consuming, but for those of ordinary skill in the art who benefit from this disclosure, the development effort will be a routine task in design, manufacturing, and production without requiring extensive experimentation.

[0151] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A machine vision-based installation control system for external spiral chute in silos, characterized in that, include: The image acquisition unit includes a first image sensor and a second image sensor, which acquires images of the first target and the second target of the first chute segment flange and the second chute segment flange in real time. The image processing unit is communicatively connected to the image acquisition unit. It calculates the attitude information of the first target and the second target in space in real time using the images of the first target and the second target, and calculates the real-time torsional angle deviation value of the first flange relative to the second flange about its theoretical docking axis using the attitude information of the first target and the second target. The calculation unit compares the real-time torsion angle deviation value with a preset threshold range and generates control commands; The execution unit receives the control command and drives the first chute segment to rotate around its axis to adjust the torsion angle deviation.

2. The machine vision-based silo external spiral chute installation control system according to claim 1, characterized in that: The first image sensor continuously acquires images of the first target fixed on the first flange during the installation process; The second image sensor continuously acquires images of the second target fixed on the second flange during installation.

3. The machine vision-based silo external spiral chute installation control system according to claim 1, characterized in that: The image processing unit identifies multiple feature points in the target image and obtains their image coordinates; By using the physical dimensions of the target and the known three-dimensional coordinates of each feature point, a correspondence between the image coordinates and the three-dimensional coordinates of the feature points is established. The perspective projection model is solved by the correspondence to obtain the transformation relationship of each target from the body coordinate system to the camera coordinate system; The parameters characterizing the planar spatial position and orientation of each target are determined and output through the transformation relationship.

4. The machine vision-based silo external spiral chute installation control system according to claim 3, characterized in that: The image processing unit determines the theoretical docking axis using pre-stored models of the first chute segment and the second chute segment; Obtain the theoretical center positions of the first flange and the second flange from the design model; In the system coordinate system, a straight line connecting the two theoretical center positions and parallel to the silo's central axis is defined as the theoretical docking axis.

5. The machine vision-based silo external spiral chute installation control system according to claim 4, characterized in that: The image processing unit is used to calculate the real-time torsion angle deviation value; The first normal vector and the second normal vector are obtained from the planar parameters of the first target and the second target, respectively. The first normal vector and the second normal vector are projected onto a plane perpendicular to the theoretical docking axis to obtain the first projection vector and the second projection vector, respectively. The angle between the first projection vector and the second projection vector in the vertical plane is calculated. The angle is the real-time torsion angle deviation value.

6. The machine vision-based silo external spiral chute installation control system according to claim 1, characterized in that: The image processing unit further sets a first threshold and a second threshold that is less than the first threshold; when the absolute value of the torsion angle deviation is greater than the first threshold, it outputs a continuous rotation control signal; when the absolute value of the torsion angle deviation is not greater than the first threshold but greater than the second threshold, it outputs a pulse rotation control signal; when the absolute value of the torsion angle deviation is not greater than the second threshold and continues for a preset duration, it outputs a lock signal and triggers a prompt.

7. The machine vision-based silo external spiral chute installation control system and method according to claim 1, characterized in that: The calculation unit compares the real-time torsion angle deviation value with a preset threshold range; Establish a dynamic data sequence that includes the current deviation value and multiple historical deviation values; The dynamic data sequence is weighted and calculated to generate a comprehensive deviation evaluation value; The comprehensive deviation evaluation value is compared with a preset static threshold to determine the adjustment stage of the system.

8. The machine vision-based silo external spiral chute installation control system according to claim 1, characterized in that: The computing unit also generates control commands based on the comparison results, including: The first type is a continuous rotation control command, used to drive the execution unit to rotate at a constant speed; The second type is pulse-type micro-motion control commands, which are used to drive the execution unit to perform intermittent angular displacement adjustments; The third type is the lock and confirmation command, which is used to stop the execution unit and issue an indication signal that alignment is complete.

9. The machine vision-based silo external spiral chute installation control system according to claim 8, characterized in that: The execution unit includes a servo rotary drive mechanism mounted on the first chute segment hoisting mechanism; the servo rotary drive mechanism receives control commands issued by the image processing and control unit, and converts them into the corresponding drive mode according to the command type, driving the first chute segment to rotate precisely around its own geometric axis to correct the real-time torsion angle deviation value.

10. A machine vision-based method for controlling the installation of external spiral chute in silos, characterized in that... Includes a machine vision-based installation control system for external spiral chute in silos as described in any one of claims 1-9; wherein: The image acquisition unit acquires images of the first target and the second target, which are fixedly set on the first flange and the second flange, in real time. The image is processed by the image processing and control unit to calculate the attitude information of the first target and the second target in space in real time. Using the attitude information of the first and second targets, the real-time torsional angle deviation of the first flange relative to the second flange around the theoretical docking axis is calculated. The real-time torsion angle deviation value is compared with a preset threshold, and a control command is generated based on the comparison result; The control command is sent to the execution unit to drive the first chute segment to rotate around its axis in order to adjust the torsion angle deviation.

Images