Flexible feeding vision system based on XYZ multi-module motion control system

Abstract

This invention discloses a flexible material feeding vision system based on an XYZ multi-module motion control system. The system includes an XYZ multi-module control subsystem, a fly-grabbing subsystem, an image processing subsystem, a motion correction subsystem, and a flexible material feeding control device. The XYZ multi-module control subsystem plans motion parameters based on preset data and historical material feeding information, driving the multi-module to complete grasping and placement. The fly-grabbing subsystem dynamically acquires grasping and placement positioning images during the movement. The image processing subsystem corrects image offsets based on flexible material feeding characteristic parameters. The motion correction subsystem achieves precision closed-loop control through secondary confirmation and real-time compensation. The flexible material feeding control device adapts to various material specifications with an adjustable tray and controllable vibration, reducing damage. Through the collaborative operation of multiple subsystems, efficient, low-damage, and high-precision material feeding of various specifications is achieved.

Description

Technical Field

[0001] This invention relates to the field of flexible material feeding control, specifically to a flexible material feeding vision system based on an XYZ multi-module motion control system. Background Technology

[0002] In industrial automated production, automated material feeding is a crucial link in ensuring the continuous operation of the production line. Its efficiency and accuracy directly affect the overall production cycle and product quality. As the manufacturing industry shifts towards multi-variety, small-batch, and flexible production, traditional material feeding technologies are gradually revealing insurmountable shortcomings, specifically as follows:

[0003] Traditional vibratory feeder technology has limitations. Currently, most production lines use ordinary vibratory feeders as material feeding devices. Their working principle involves generating periodic mechanical vibrations from a bottom vibration source, causing materials to move and align along a pre-set fixed feed channel, ultimately achieving directional material conveying. However, this technology has significant drawbacks, such as poor compatibility. The size and shape of the vibratory feeder's feed channel must match specific materials. When changing to different material specifications, the feed channel needs to be redesigned or replaced, resulting in long changeover cycles and high costs, making it unsuitable for multi-variety mixed-line production. There is also a high risk of material damage. Materials are arranged through vibration, friction, and collision within the fixed feed channel, easily leading to surface scratches and edge damage, especially affecting precision parts and fragile materials such as plastics and ceramics, reducing product yield. Finally, there is insufficient flexibility. The fixed structure of the feed channel restricts the adjustment of material placement. For scenarios requiring multi-angle gripping or directional placement, additional auxiliary positioning mechanisms are needed, further increasing system complexity.

[0004] To address compatibility issues with vibratory feeders, some production lines employ 4-axis robots in conjunction with vision inspection for material loading. The process involves the robot grabbing material, then stopping at a secondary confirmation station to statically inspect the material's position and angle using a camera – a process known as "stop-and-shoot inspection." Placement parameters are then adjusted based on the inspection results. However, this technology suffers from efficiency limitations: the joint motion characteristics of the 4-axis robot restrict its operating speed to critical limits, and the stop-and-shoot process during secondary confirmation requires additional time, resulting in a minimum grabbing efficiency of only about 30 pieces per minute, insufficient for high-speed production. Furthermore, the lack of coordination between robot motion control and vision inspection disrupts the continuous flow of the production line during the stop-and-shoot process, increasing system response latency and further limiting overall efficiency.

[0005] Therefore, in order to solve the problems of high material loss risk, low gripping and placement efficiency, and poor compatibility of production line channels in the existing technology, this application proposes a flexible material feeding vision system based on the XYZ multi-module motion control system. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the present invention aims to provide a flexible material feeding vision system based on an XYZ multi-module motion control system.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A flexible material feeding vision system based on an XYZ multi-module motion control system includes: The XYZ multi-module control subsystem includes an XYZ multi-module motion control device, which obtains material coordinates based on the preset number of gripping stations and material tray information, combined with historical material data, and controls the XYZ multi-module to grip and place materials according to the material coordinates. The flying camera grabbing subsystem captures images of the grabbing position of the XYZ multi-module motion control device during the grabbing process using a flying camera as the grabbing positioning image, and also captures images of the material placement position as the placement positioning image. The image processing subsystem performs image blur repair on the capture and placement images using a flying image processing strategy, calculates the image offset, and performs image correction based on the image offset and flexible feeding feature parameters to obtain the processed capture and placement images. The motion correction subsystem, based on the captured image and the placement image, performs a secondary confirmation of the captured and placement positions of the XYZ multi-module motion control subsystem through a motion correction strategy, and corrects the captured and placement parameters of the XYZ multi-module motion control device. A flexible feeding control device is used for flexible feeding of materials.

[0008] As a further improvement of the present invention, the XYZ multi-module control subsystem includes preset gripping data and material tray information. The gripping data includes the number of gripping stations and the type of gripped material. The material tray information includes the tray size and tray image. The position of the material in the tray is determined by combining the tray size and tray image with historical material data. The material with the same number of gripping stations is selected and the material coordinates are determined. The gripping motion direction, gripping angle, and gripping motion speed of the XY module motion control device of the XYZ multi-module motion control device are determined by the number of gripping stations and the material coordinates. The placement motion direction and placement motion speed of the XY module are determined according to the placement position. The descent depth of the Z module of the XYZ multi-module motion control device is set.

[0009] As a further improvement of the present invention, the flying camera grasping subsystem includes a synchronization triggering module and an image acquisition device. When the XYZ multi-module motion control device moves along the XY direction at a preset speed, the synchronization triggering module triggers the camera to capture images within a preset time period when the material enters the camera's field of view, based on the real-time position signal of the XY module of the XYZ multi-module motion control device. The shooting frequency is positively correlated with the movement speed of the XYZ multi-module motion control device. The shooting area of ​​the grasping positioning image includes the entire material and edge feature points, and the placement positioning image includes the baseline of the placement area.

[0010] As a further improvement of the present invention, the flexible feeding feature parameters in the image processing subsystem include material surface roughness, used to distinguish the reflectivity differences of different materials; material elastic deformation coefficient, used to correct the positional displacement of easily deformable workpieces caused by gripping pressure; and residual displacement of the material tray vibration, used to compensate for the micro-displacement of the material caused by the vibration of the flexible vibrating tray. When performing image correction, blurring is first eliminated by blur repair processing, and then the image offset is weighted and corrected in combination with the flexible feeding feature parameters.

[0011] As a further improvement of the present invention, the aerial image processing strategy includes: image preprocessing, performing grayscale and noise reduction processing on the capture positioning image and the placement positioning image, while preserving edge features; identifying key feature points of the material through a feature extraction algorithm, and calculating the coordinates of the key feature points in the image coordinate system; calculating the image clarity score based on the flexible feeding feature parameters, and if it is less than a preset threshold, initiating blur repair processing, and re-extracting feature points after repair; comparing the actual coordinates of the key feature points with the theoretical coordinates based on historical feeding data to obtain the translational offset and rotational offset in the XY direction, and compensating for the translational offset and rotational offset by combining them with the material flexible feature parameters.

[0012] As a further improvement of the present invention, the motion correction subsystem includes, during secondary confirmation, comparing the processed captured image with a preset qualified capture template, wherein the qualified capture template includes an allowable range for feature point position deviation; if the deviation exceeds the allowable range, triggering capture parameter correction, wherein the capture parameters include adjusting the descent depth of the Z module and real-time compensation of the rotary motor angle based on the rotation offset; and placement parameters include correcting the placement endpoint coordinates of the XY module according to the correction offset of the placement positioning image and adjusting the placement speed.

[0013] As a further improvement of the present invention, the motion correction strategy includes: pre-generating the motion trajectory correction amount of the XYZ module based on historical material data; comparing the real-time correction offset obtained by the flying camera with the predicted correction amount; if the difference is greater than a preset threshold, triggering real-time compensation; adjusting the motor driver through a signal; and automatically pausing material feeding and issuing an alarm when the correction offset exceeds the preset threshold for several consecutive times, while recording the characteristic parameters of the abnormal material.

[0014] As a further improvement of the present invention, a historical data correction module is also included, which saves the actual offset of the placement position and the gripping position and the correction parameters to the historical material loading database after each loading is completed, for updating the predicted correction amount.

[0015] As a further improvement of the present invention, the flexible feeding control device includes, An adjustable flexible tray is used as a material carrying platform. Its size can be adjusted by the tray adjustment mechanism to adapt to materials of different specifications. The planar area of ​​the adjustable flexible tray is used for the dispersed placement of materials, and its edge is provided with reference marks for positioning. The vibration drive assembly is installed below the adjustable flexible tray to provide vibration with controllable frequency and intensity, so that the material is dispersed in the tray and moves without a fixed path, avoiding frictional damage caused by fixed material channels. The vibration control module is electrically connected to the XYZ multi-module control subsystem and the vibration drive component. It is used to receive motion control signals and adjust the vibration parameters of the vibration drive component. The vibration parameters include vibration frequency and start / stop time. The module stops vibration within a preset time before the XYZ multi-module motion control device performs the grasping action to reduce the micro-displacement of the material caused by vibration. The tray status detection component is used to collect real-time size information and material distribution images of the adjustable flexible tray, and transmit them to the XYZ multi-module control subsystem as part of the tray information to help determine the material coordinates.

[0016] As a further improvement of the present invention, the material tray status detection component transmits the collected real-time size information of the adjustable flexible material tray and the material distribution image to the XYZ multi-module control subsystem as the core input of the material tray information, which is used by the XYZ multi-module control subsystem to determine the material coordinates in combination with historical material data; The vibration control module receives the grabbing preparation signal from the XYZ multi-module control subsystem. Within a preset time period before the flying grabbing subsystem triggers shooting, it controls the vibration drive component to stop vibrating, so that the material is in a stable state to cooperate with the flying grabbing subsystem to obtain a clear grabbing positioning image. According to the grabbing completion signal fed back by the XYZ multi-module control subsystem, the vibration control module adjusts the vibration drive component to resume vibration, so that the remaining material is redispersed, providing a material distribution basis for the next grabbing.

[0017] The beneficial effects of this invention are: (1) Improve flexibility and compatibility to meet the feeding requirements of multiple specifications of materials. Through the adjustable flexible material tray of the flexible feeding control device and the vibration drive component without fixed material channel, it can adapt to materials of different sizes and shapes without changing the hardware. This solves the problem of long changeover cycle and poor compatibility caused by the traditional vibratory feeder relying on fixed material channel, and meets the needs of multi-variety mixed production line.

[0018] (2) Improve feeding efficiency, break through the cycle bottleneck of traditional technology, adopt the flying grabbing subsystem to realize dynamic detection in motion, and combine the parallel grabbing capability of the XYZ multi-module control subsystem to eliminate the downtime waiting time of the secondary confirmation station and meet the requirements of high-speed production cycle.

[0019] (3) Reduce material damage and ensure product quality. Through flexible vibration without fixed material channel, precise descent depth control of Z module and grasping parameter correction based on material elastic deformation coefficient, the problems of surface scratches and corner collisions caused by traditional vibratory feeder material channel friction and robot hard grasping are avoided, significantly improving the product qualification rate.

[0020] (4) Improve positioning and placement accuracy and reduce error accumulation. The image processing subsystem combines the flexible feeding feature parameters to perform weighted correction on the image offset. Combined with the prediction and feedback double closed-loop control of the motion correction subsystem, the position deviation of grasping and placement is controlled within a very small range, which solves the problem of insufficient accuracy caused by ignoring the characteristics of flexible materials in traditional visual inspection. Attached Figure Description

[0021] Figure 1 This is a system block diagram of a flexible material feeding vision system based on an XYZ multi-module motion control system according to the present invention. Figure 2 This is a flowchart of the process of a flexible material feeding vision system based on an XYZ multi-module motion control system according to the present invention. Figure 3 This is a schematic diagram of the flexible feeding control device of the present invention; Figure 4 This is a schematic diagram of the operation of a flexible material feeding vision system based on an XYZ multi-module motion control system according to the present invention. Figure 5 This is a schematic diagram of material capture for a flexible feeding vision system based on an XYZ multi-module motion control system, according to the present invention. Detailed Implementation

[0022] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom surface," "top surface," "inner," and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0023] The first embodiment of this invention proposes a flexible material loading vision system based on an XYZ multi-module motion control system. Its core lies in achieving high efficiency, precision, and compatibility in the flexible material loading process through the coordinated operation of multiple subsystems. The specific technical implementation and collaborative logic of each subsystem are as follows: Figures 1 to 5 As shown: The basic process of this plan includes: Initialization: Preset grabbing parameters, number of workstations (6), material tray size (500×500mm), fly-grabbing parameters, and flexibility feature parameters. Enter Ra value and elasticity coefficient according to material type. Grabbing phase: The XYZ module moves to the grabbing point according to the material coordinates, the Z module descends to the preset depth, and the flying camera subsystem simultaneously captures the grabbing and positioning image; Detection and correction: The image processing subsystem outputs the correction offset, and the motion correction subsystem adjusts the grasping angle and position to complete the grasping. Placement stage: The XYZ module carries the material to the placement area, and the flying camera system takes a picture of the placement and positioning, repeating the correction process in step 3; Closed-loop optimization: After each material loading is completed, the actual offset and correction parameters are stored in the historical database to update the predicted correction amount. For example, the template is updated once every 100 cycles.

[0024] This system includes: The XYZ multi-module control subsystem includes an XYZ multi-module motion control device, which obtains material coordinates based on the preset number of gripping stations and material tray information, combined with historical material data, and controls the XYZ multi-module to grip and place materials according to the material coordinates. The XYZ multi-module control subsystem is the "motion hub" of the entire system. Its core component is the XYZ multi-module motion control device, which consists of a planar motion mechanism composed of an X-axis linear module and a Y-axis linear module, and an Z-axis execution module. The Z-axis is responsible for vertical motion and angle adjustment.

[0025] The X-axis and Y-axis modules achieve high-speed, smooth movement in a plane through a precision transmission structure, such as a combination of ball screws and servo motors. Their motion trajectory can be planned in real time according to the material coordinates. The Z-axis module integrates gripping actuators and rotary drive components, which can grip and release materials through vertical lifting and can also adjust the placement angle of materials through rotation, thereby adapting to different orientation placement requirements.

[0026] The subsystem first receives the preset number of gripping stations, i.e. the number of materials that can be gripped at one time, such as 6, as well as the material tray information, including the size parameters of the material tray and real-time images. Then, it combines historical material data, including the position of materials gripped in the past, success rate, and offset patterns, for comprehensive analysis. It identifies the distribution state of materials through the material tray images, selects the materials to be gripped that match the number of gripping stations, and finally calculates the specific X and Y coordinates and initial angle of each material in the planar coordinate system.

[0027] Based on the material coordinates, the subsystem automatically plans the gripping motion direction of the XY module, such as the optimal path from the edge of the tray to the center, the gripping angle, and the angle matching the initial placement posture and movement speed of the material. The movement speed takes into account both efficiency and stability. At the same time, based on the preset placement area position, the placement motion direction and speed of the XY module are determined, and the descent depth of the Z-axis module is set, which is the contact distance between the gripper and the material during gripping, to avoid overpressure damage to the material.

[0028] Specifically, such as Figures 1 to 5 As shown, the XYZ multi-module control subsystem includes preset gripping data and tray information. The gripping data includes the number of gripping stations and the type of gripped material. The tray information includes the tray size and tray image. Based on the tray size and tray image, and combined with historical material data, the position of the material in the tray is determined. The material with the same number of gripping stations is selected, and the material coordinates are determined. Based on the number of gripping stations and the material coordinates, the gripping motion direction, gripping angle, and gripping motion speed of the XY module of the XYZ multi-module motion control device are determined. The placement motion direction and placement motion speed of the XY module are determined according to the placement position, and the Z module descent depth of the XYZ multi-module motion control device is set.

[0029] The "preset capture data" are basic parameters configured based on production needs before system startup, serving as the benchmark for subsequent motion control. Specifically, they include: The number of gripping stations refers to the number of materials the system can grip simultaneously in a single operation. Its value matches the number of Z-axis modules in the XYZ multi-module motion control device; for example, 6 gripping stations correspond to 6 independent Z-axis modules. This parameter setting needs to consider the material size to avoid collisions during gripping and the production cycle requirements. For instance, 6 stations can grip 6 materials at a time, directly supporting an efficiency target of 60 materials per minute. This parameter can be dynamically adjusted through the system interface to adapt to different batch production needs.

[0030] The material type to be grasped refers to the set of physical characteristics of the material to be grasped, including but not limited to the material's size range (length, width, height), shape characteristics (e.g., round, square, irregular shape), material properties (e.g., metal, plastic, rubber), and related flexible characteristic parameters such as hardness and surface roughness, as well as orientation requirements (e.g., whether a specific angle is required). This information is used to optimize the grasping parameters. For example, for easily deformable rubber parts, the system will automatically reduce the grasping pressure of the Z-axis gripper; for smooth metal parts, the gripper material will be adjusted to prevent slippage.

[0031] The "material tray information" refers to real-time data reflecting the status of the material carrying platform, providing a spatial reference for material positioning, specifically including: The size of the tray refers not only to the current physical dimensions of the adjustable flexible tray, but also to its adjustable range. This information is collected in real time by the tray status detection component. On the one hand, it is used to determine whether the tray is suitable for the current type of material being grasped. On the other hand, it serves as the basis for establishing a coordinate system, using the tray edge reference mark as the origin to determine the range of planar coordinates.

[0032] The tray image refers to the real-time image of the tray captured by the vision sensor in the tray status detection component. It includes the distribution of materials in the tray. The image recognition algorithm distinguishes the materials from the tray background and initially locates the approximate area of ​​the materials, providing raw data for subsequent accurate coordinate calculation.

[0033] Determining material location by combining tray size, tray image, and historical material handling data is the core step in achieving precise grasping. The system first performs noise reduction and enhancement on the tray image, highlighting edge features such as corners and holes. Then, it uses feature templates of similar materials from historical material handling data, employing a template matching algorithm to identify all materials to be grasped in the image, excluding impurities or substandard materials. Historical material handling data can include common distribution areas and offset patterns of this type of material from the past 100 grasps, used to optimize recognition results. For example, if historical data shows that material tends to accumulate in a certain area due to vibration, the system will prioritize materials in sparsely distributed areas to reduce the risk of grasping interference; if a certain type of material frequently exhibits an X-axis offset of ±0.5mm, this compensation amount will be automatically added after initial positioning. The system selects a set of materials with matching quantities and reasonable spacing from the identified materials based on the number of gripping stations. Then, it establishes a Cartesian coordinate system with the edge reference mark of the material tray as the origin. By calculating the X and Y coordinate values ​​of the material feature points, usually such as the geometric center, in this coordinate system, the precise position of each material is determined. At the same time, combined with the posture information of the material in the material tray image, usually the angle with the X-axis of the coordinate system, the initial angle of the material is recorded.

[0034] Based on the number of gripping stations and the coordinates of the materials, the system performs targeted planning of the motion parameters of the XYZ multi-module motion control device to ensure a balance between efficiency and precision.

[0035] The grasping motion direction is calculated using a path planning algorithm to find the shortest path from the current position to the coordinates of each material. The path must avoid other identified materials and prioritize linear motion to reduce motion time.

[0036] The gripping angle is determined based on the initial angle of the material and the preset gripping posture, such as the difference between the length of the material and the X-axis. The rotation angle of the XY module or the compensation angle of the Z-axis rotary motor is determined to ensure that the gripper and the material posture match during gripping.

[0037] The grasping motion speed adopts a segmented speed adjustment strategy, moving at high speed when far away from the material area and slowing down when approaching the material to avoid position overshoot caused by inertia. This speed value will be dynamically adjusted according to the material quality; the greater the quality, the lower the speed in the high-speed segment.

[0038] Based on the preset placement area location, such as the positioning slot of the assembly line vehicle, plan the movement direction from the gripping position to the placement position, prioritizing straight lines or smooth curves, and set the placement speed, which is usually lower than the gripping speed, to ensure stability during placement.

[0039] Based on the material height and gripper thickness, calculate the vertical distance from the initial height of the Z-axis module to the material in contact. For example, if the material height is 10mm and the gripper thickness is 5mm, then the descent depth is set to 15mm + 0.5mm buffer, which ensures reliable gripping while avoiding excessive pressure from the gripper that could cause material deformation or damage.

[0040] The flying camera grabbing subsystem captures images of the grabbing position of the XYZ multi-module motion control device during the grabbing process using a flying camera as the grabbing positioning image, and also captures images of the material placement position as the placement positioning image. The aforementioned flying camera capture subsystem is key to solving the efficiency bottleneck of traditional "stop-shoot detection". Its core lies in realizing "detection in motion", that is, completing image acquisition during the movement of the XYZ multi-module without stopping and waiting.

[0041] Unlike traditional "stop-and-shoot" photography, "flying photography" involves the XYZ multi-module motion control device moving along the XY direction while the image acquisition device simultaneously captures the image. This process relies on the real-time linkage between the synchronization trigger module and the motion status. When the XY module carrying material enters the camera's field of view, the synchronization trigger module, based on the real-time position signal of the XY module, will trigger the camera to capture the image within a preset time window, such as the instant the material completely enters the field of view. This ensures that the material is in the center of the field of view during the capture, guaranteeing image integrity.

[0042] The captured positioning image must completely include the material as a whole and its edge feature points, such as geometric markers like corners and holes, for subsequent identification of the material's actual position and angle. The placement positioning image must include the baseline of the placement area, such as preset positioning scales or coordinate marks, to determine whether the material placement meets the positional requirements. Furthermore, the shooting frequency is positively correlated with the movement speed of the XY module; the higher the speed, the higher the shooting frequency, to avoid missed or blurred images due to excessively fast movement.

[0043] Specifically, such as Figures 1 to 5 As shown, the flying camera grasping subsystem includes a synchronization trigger module and an image acquisition device. When the XYZ multi-module motion control device moves along the XY direction at a preset speed, the synchronization trigger module triggers the camera to capture images within a preset time when the material enters the camera's field of view, based on the real-time position signal of the XY module of the XYZ multi-module motion control device. The shooting frequency is positively correlated with the movement speed of the XYZ multi-module motion control device. The shooting area of ​​the grasping positioning image includes the entire material and edge feature points, and the placement positioning image includes the baseline of the placement area.

[0044] The aforementioned aerial capture subsystem achieves efficient detection of material status during movement through a collaborative design of synchronous triggering and dynamic image acquisition. Its technical details are as follows: The synchronous trigger module is the core component connecting the XYZ multi-module motion control device and the image acquisition device. Its core function is to generate a shooting trigger signal in real time based on the motion state. This module receives real-time position signals from the XY modules, such as displacement pulses output by a grating ruler or encoder, accurately calculates the time window for the material to enter the field of view of the image acquisition device, and outputs an electrical signal within this window to trigger shooting. This ensures that the shooting action is strictly synchronized with the material's movement state, avoiding incomplete images due to the material not fully entering the field of view because of movement ahead of time, or missed shots due to the material moving out of the field of view because of movement lag.

[0045] Image acquisition devices refer to hardware equipment used to collect images of materials. Their technical characteristics need to be adapted to "flying photography" scenarios. They typically use global shutter cameras, which can complete the exposure of the entire image in a very short time, at the microsecond level, to avoid motion blur caused by material movement. At the same time, they are equipped with high-resolution lenses to ensure that the detailed features of the material, such as edge contours and tiny holes, are clearly captured, providing high-quality raw data for subsequent image processing.

[0046] When the XYZ multi-module motion control device moves along the XY direction at a preset speed, which is comprehensively set by the XYZ multi-module control subsystem based on production cycle and image clarity requirements, balancing efficiency and quality, the synchronous trigger module's workflow is a closed loop of "position monitoring, time prediction, and signal output": Position monitoring tracks the movement trajectory of the XY modules in real time, and analyzes the current coordinates of the material through position signals, such as the distance from the left boundary of the camera's field of view; Based on the preset speed and current position, it calculates the moment when the material completely enters the field of view, i.e., the instant when the material edge aligns with the field of view boundary, and reserves a preset time, such as 0.1-0.5 seconds, the specific value of which is dynamically adjusted according to the material size and speed, as a shooting preparation window to ensure that the shooting is completed when the material is in the center area of ​​the field of view, at which time the image distortion is minimal; At the predicted moment, it outputs a trigger signal to drive the image acquisition device to complete one shooting.

[0047] The shooting frequency is positively correlated with the movement speed of the XYZ multi-module motion control device. Its technical essence is to dynamically adjust the number of shots to match the dwell time of the material in the field of view. When the movement speed increases, the time for the material to pass through the field of view is shortened. At this time, the shooting frequency needs to be increased to ensure that at least one effective shot is completed in a short period of time. When the movement speed decreases, the dwell time of the material in the field of view is extended, and the shooting frequency can be reduced to reduce data redundancy and avoid excessive image processing pressure.

[0048] This dynamic adaptation ensures the integrity of image acquisition while avoiding resource waste caused by invalid shooting.

[0049] The calculation of the preset time for material to enter the camera's field of view includes: the preset time equals the distance from the edge of the material (the side closest to the field of view) to the center of the camera's field of view divided by the current instantaneous motion speed of the XYZ module, plus a camera trigger delay compensation value. The distance can be obtained through offline calibration, the instantaneous motion speed is fed back in real time by the encoder, and the trigger delay compensation value is obtained by measuring the time difference from the trigger signal's emission to the start of image exposure using an oscilloscope; a typical value is 0.002 to 0.005 seconds. The system automatically measures and calibrates this compensation value every 10 minutes.

[0050] The shooting frequency is positively correlated with the movement speed of the XYZ multi-module motion control device. Specifically, through a piecewise linear mapping principle, when the movement speed does not exceed 100mm / s, the shooting frequency is fixed at a minimum value of 50Hz; when the movement speed is between 100mm / s and 800mm / s, the shooting frequency increases linearly with the speed, with a scaling factor set to increase by approximately 0.214Hz per millimeter per second; when the movement speed reaches or exceeds 800mm / s, the shooting frequency stabilizes at a maximum value of 200Hz. This mapping relationship ensures that the material is photographed at least twice within the field of view to provide redundant alternative images. The real-time position signal received by the synchronous trigger module is provided by a grating ruler or encoder, with a resolution requirement of not less than 0.01mm and a signal update cycle of not more than 1ms. To eliminate signal transmission and processing delays, the system adopts a feedforward compensation mechanism: extrapolating the position of the next sampling cycle based on the movement speed of the previous three sampling points, a trigger signal is generated in advance to compensate for the delay (typically 0.5ms).

[0051] The area captured in the positioning image must completely encompass the entire material to ensure the identification of its global orientation and "edge feature points," such as at least three non-collinear corners or holes, which serve as the benchmark for subsequent coordinate calculations. The positional data of these feature points in the image coordinate system are the core basis for determining the deviation between the actual grasping position and the theoretical coordinates. For example, by comparing the actual coordinates of the feature points with the preset template coordinates, the translational and rotational offsets in the X and Y directions can be directly calculated.

[0052] The area captured by the positioning image must include a "placement area baseline," such as a preset physical scale line or optical positioning mark, with accuracy matching the system's positioning requirements. The baseline serves to provide a spatial reference for material placement. By identifying the relative position of the material's edge to the baseline, it can be determined whether the material is placed within the preset area and whether the angle meets the requirements, such as whether the material's long side is parallel to the baseline, providing a direct basis for correcting placement parameters.

[0053] Through the above design, the flying camera grabbing subsystem breaks through the speed limit of traditional "stop-shoot detection". While ensuring image quality, it integrates material detection into the motion process, providing key support for the system to achieve efficient material loading.

[0054] The image processing subsystem performs image blur repair on the capture and placement images using a flying image processing strategy, calculates the image offset, and performs image correction based on the image offset and flexible feeding feature parameters to obtain the processed capture and placement images. The image processing subsystem is key to achieving "visual guidance." It analyzes the captured and placed positioning images using a flying image processing strategy to provide accurate deviation data for motion correction.

[0055] The execution logic of the aerial image processing strategy is as follows: First, the image is preprocessed, grayscale processing is performed to eliminate color interference, noise reduction processing filters out the influence of ambient light, and the edge features of the material are preserved to provide a basis for subsequent feature recognition; then, the key feature points of the material are identified through feature extraction algorithms, and the coordinates of these feature points in the image coordinate system are calculated; if the image produces a blur due to high-speed motion, i.e., the blur exceeds the standard, blur repair processing is initiated, the motion path is restored to eliminate the blur, and feature points are re-extracted after repair to ensure data reliability.

[0056] Offset correction is achieved by introducing specific parameters for flexible material feeding scenarios. Specifically: Material surface roughness is used to distinguish the reflectivity differences of different materials such as metal and plastic. For example, strong reflectivity of metal surfaces can easily lead to feature point recognition errors, and the image threshold needs to be adjusted by roughness parameters. The material elastic deformation coefficient is used to correct the minute deformation of easily deformable workpieces caused by gripping pressure, and to avoid misjudgment of position due to deformation. The residual displacement of the vibrating disc is used to compensate for the small displacement of the material after the flexible vibrating disc stops vibrating. After the vibration stops, the material may move slightly due to inertia. This parameter can quantify and correct this offset.

[0057] By using the above parameters to perform weighted correction on the image offset and the difference between the actual and theoretical coordinates of feature points, a processed image that truly reflects the state of the material is obtained.

[0058] Specifically, such as Figures 1 to 5 As shown, the flexible feeding feature parameters in the image processing subsystem include material surface roughness, used to distinguish the reflectivity differences of different materials; material elastic deformation coefficient, used to correct the positional displacement of easily deformable workpieces caused by gripping pressure; and residual displacement of the material tray vibration, used to compensate for the micro-displacement of the material caused by the vibration of the flexible vibrating tray. When performing image correction, blurring is first eliminated by blur repair processing, and then the image offset is weighted and corrected in combination with the flexible feeding feature parameters.

[0059] Material surface roughness refers to the degree of microscopic unevenness on the surface of a material, and is a core criterion for distinguishing the differences in reflectivity between different materials. The technical principle is that smooth materials such as metals have low surface roughness, resulting in strong reflectivity and potentially creating highlight areas that obscure material edge features; rough materials such as plastics and rubber have high surface roughness, resulting in weak reflectivity but also prone to diffuse reflection, which may lead to low image contrast. The system uses preset image processing strategies corresponding to different roughness levels. For example, for highly reflective, low-roughness materials, it automatically reduces the camera exposure time to avoid overexposure and enhances the anti-interference capability of the edge detection algorithm; for low-reflectivity, high-roughness materials, it increases the light source brightness to enhance contrast and optimizes noise reduction parameters to reduce noise caused by diffuse reflection, thereby ensuring the accuracy of feature point recognition.

[0060] Material surface roughness refers to the degree of microscopic unevenness on the surface of a material, used to distinguish the reflectivity differences between different materials such as metals, plastics, and rubber. This parameter is obtained through offline measurement. During system initialization, the operator selects the corresponding roughness value from a preset database based on the material material. For example, the roughness of mirror-finished metal parts is no more than 0.2 μm, ordinary machined metal parts are 0.8~1.6 μm, injection-molded plastic parts are 1.6~3.2 μm, and rubber parts are 6.3~12.5 μm. If it is a new material and there is no corresponding value in the database, it can be measured using a handheld surface roughness meter and then manually entered. The system automatically adjusts the image acquisition parameters based on the roughness value. When the roughness is no more than 0.8 μm, the camera exposure time is reduced to 60% of the original value and the highlight suppression algorithm is enabled; when the roughness is not less than 3.2 μm, the brightness of the ring light source is increased to 150% of the original value and multi-frame noise reduction is enabled. The elastic deformation coefficient of a material is a dimensionless coefficient, ranging from 0 to 0.2. It is used to quantify the degree of deformation of easily deformable workpieces under gripping pressure. It is calculated by dividing the material's compression under gripping pressure by its original size, and then dividing by the gripping pressure. The calibration of this coefficient is automatically completed through the system's trial gripping mode. During the initialization phase, the system grips the material at a preset low pressure of 0.5N and captures an image, recording the material's original size. Subsequently, it grips and captures the material again at a preset rated gripping pressure of 2.0N, calculating the compression amount—the difference between the original size and the size after compression—and substituting this value into the aforementioned formula to obtain the deformation coefficient. If the material is a rigid body, this coefficient is automatically set to 0.

[0061] The residual displacement of the vibrating disc refers to the minute displacement of the material due to inertia or the elastic recovery of the disc after the flexible vibrating disc stops vibrating. It is calculated by dividing the material's velocity at the instant vibration stops by 2π, multiplying by the vibration frequency, and then multiplying by an exponential factor that decays with time and damping ratio. The material's velocity at the instant vibration stops is obtained by integrating the acceleration during the last vibration cycle before stopping using an accelerometer. The vibration frequency is provided by the vibration control module. The damping ratio is a pre-set empirical value based on the combined characteristics of the material and the disc, with a smaller value for light materials and a larger value for heavy materials. The time interval between the vibration stopping and the camera capturing the image is obtained by system timing. This parameter is calculated in real-time by the system after each vibration stops, without manual intervention.

[0062] The residual displacement of the vibrating disc refers to the minute displacement of the material due to inertia or the elastic recovery of the disc after the flexible vibrating disc stops vibrating. Its calculation is based on the vibration frequency of the vibration drive component, the material mass, and the vibration stopping time. Higher frequencies, greater inertia, and larger masses result in more significant residual displacement; longer stopping times lead to more complete displacement attenuation. The technical application is as follows: the shooting action of the flying camera grasping subsystem usually takes place after vibration stops, but residual displacement may cause a deviation between the material position in the image and its position before vibration stopped. The system pre-calculates the residual displacement by collecting vibration parameters and material mass in real time, and superimposes this value into the image coordinates to compensate for the positional error caused by residual vibration.

[0063] Specifically, such as Figures 1 to 5 As shown, the aerial image processing strategy includes: image preprocessing, performing grayscale conversion and noise reduction on the captured positioning image and the placement positioning image while preserving edge features; identifying key feature points of the material using a feature extraction algorithm and calculating the coordinates of the key feature points in the image coordinate system; calculating the image clarity score based on the flexible feeding feature parameters, and if it is less than a preset threshold, initiating blur repair processing, and re-extracting feature points after repair; comparing the actual coordinates of the key feature points with the theoretical coordinates based on historical material feeding data to obtain the translational and rotational offsets in the XY directions, and compensating for the translational and rotational offsets by combining them with the material flexibility feature parameters.

[0064] The image sharpness score is calculated as follows: ; Where S(I) is the image sharpness score, with a value ranging from 0 to 1, and I(x,y) is the image grayscale matrix function; For a two-dimensional gradient operator; R aδ represents the surface roughness of the material (unit: μm), which is collected and measured by an offline surface profilometer and then entered into the system; δ represents the elastic deformation coefficient (%), obtained from the slope of the material stress-strain curve; σ represents the vibration residual displacement (mm), calculated by integration using an accelerometer; L represents the feature scale (pixels), taken as the half-width of the smallest bounding rectangle of the material; Ω represents the image integration region; H0 represents the zeroth-order Hankel function of the first kind, whose value can be obtained by looking up tables in standard numerical calculation libraries (such as the GNU Scientific Library); λ represents the optical system resolution (μm / pixel), obtained by camera calibration, calculated by taking pictures using a standard resolution test chart and then calculating the actual physical size corresponding to each pixel; tanh represents the hyperbolic tangent normalization function. The numerator is the double integral of the image gradient energy, reflecting the sharpness of the image edges. In the denominator, the H0 function is used to describe the influence of the material surface roughness on the reflective properties. The larger the optical resolution and the rougher the surface, the larger the denominator and the lower the sharpness score. At the same time, the elastic deformation coefficient, vibration residual displacement, and feature scale are introduced for weighted correction. When S < 0.7, the image is considered to be too blurry, triggering blur correction processing. The threshold of 0.7 is set based on the following: 100 manually labeled aerial images with known sharpness are scored, and ROC curve classification analysis is used to select the threshold corresponding to the balance point between precision and recall.

[0065] The blur restoration process obtains blur kernel parameters using a Wiener filtering restoration method with known motion speed and direction. Based on the real-time position signals of the XY modules recorded by the synchronous trigger module, the instantaneous motion speed of the material in the X and Y axes at the moment of shooting is calculated. Combined with the camera exposure time, the motion blur length is calculated. The integral of the motion speed over the exposure time is equivalent to a linear displacement; that is, the blur length equals the magnitude of the motion speed vector multiplied by the exposure time. Simultaneously, the motion direction angle is determined by the ratio of the X and Y velocity components through arctangent calculation. A point spread function for linear motion blur is constructed using this length and direction. Substituting the Fourier transform of the blurred image and the point spread function into the Wiener filtering formula, in the frequency domain, the spectrum of the blurred image is weighted and restored using the ratio of the conjugate of the point spread function to its squared magnitude plus a signal-to-noise ratio parameter. The signal-to-noise ratio parameter is taken as an empirical value of 0.01. The restored clear image is obtained after inverse Fourier transform. The sharpness score of the repaired image is recalculated. If it is still below the threshold of 0.7, the signal-to-noise ratio parameter is gradually increased and the second step is repeated until the threshold is met or the number of iterations exceeds 5. If the threshold is not met after more than 5 iterations, an "image repair failure" alarm is reported, and the material is marked as abnormal. During the repair process, the blur kernel parameter and the repair result are associated with the material ID and stored in the historical material database for subsequent optimization of the repair strategy.

[0066] The motion correction subsystem, based on the captured image and the placement image, performs a secondary confirmation of the captured and placement positions of the XYZ multi-module motion control subsystem through a motion correction strategy, and corrects the captured and placement parameters of the XYZ multi-module motion control device. The motion correction subsystem ensures the accuracy of grasping and placement through a closed-loop logic of "detection, comparison, and correction," and its core is the dynamic execution of the motion correction strategy.

[0067] The subsystem compares the captured image output by the image processing subsystem with a preset "qualified capture template," including the allowable deviation range of feature point positions. If the actual deviation is within the allowable range, the capture is deemed qualified; if it exceeds the range, capture parameter correction is immediately triggered. For capture parameters, the correction includes adjusting the Z-module descent depth, such as fine-tuning based on the actual height of the material to avoid excessively deep or shallow contact between the gripper and the material; and real-time compensation of the rotary motor angle based on the rotational offset, i.e., the angular deviation obtained from image processing, to ensure the material capture angle meets preset requirements. For placement parameters, the subsystem corrects the placement endpoint coordinates of the XY modules based on the corrected offset of the placement positioning image, i.e., compensating for planar position deviations, and adjusts the placement speed, decelerating as it approaches the placement point to avoid positional shifts caused by impacts.

[0068] The dual closed-loop control of prediction and feedback incorporates "predictive correction" logic into the motion correction strategy. Based on historical material data, the motion trajectory correction amount is pre-generated. For example, if the material in a certain area often has an X-axis offset of 0.1mm, a compensation amount of 0.1mm is planned in advance. The real-time offset obtained by the aerial camera is compared with the predicted correction amount. If the difference exceeds the threshold, the motor driver is adjusted in real time through pulse signals to achieve dual protection of predictive compensation and real-time correction, which greatly improves the correction accuracy.

[0069] Specifically, such as Figures 1 to 5 As shown, the motion correction subsystem includes, during secondary confirmation, comparing the processed captured image with a preset qualified capture template. The qualified capture template includes an allowable range for feature point position deviation. If the deviation exceeds the allowable range, capture parameter correction is triggered. The capture parameters include adjusting the descent depth of the Z-module and real-time compensation of the rotary motor angle based on the rotation offset. The placement parameters include correcting the placement endpoint coordinates of the XY modules according to the correction offset of the placement positioning image and adjusting the placement speed.

[0070] Specifically, such as Figures 1 to 5As shown, the motion correction strategy includes: pre-generating the motion trajectory correction amount of the XYZ module based on historical material data; comparing the real-time correction offset obtained by the flying camera with the predicted correction amount; if the difference is greater than a preset threshold, real-time compensation is triggered; the motor driver is adjusted by the signal; when the correction offset exceeds the preset threshold for several consecutive times, the feeding is automatically paused and an alarm is issued; at the same time, the characteristic parameters of the abnormal material are recorded.

[0071] The qualified grasp template is a standard image template preset by the system, generated based on historical qualified grasp data. It contains the coordinate set of key feature points of the material under ideal grasp conditions and the relative positional relationships between the feature points. For example, for a square material, the template will record the theoretical coordinates of its four vertices in the image coordinate system, as well as geometric constraints such as the length and angle of the line connecting adjacent vertices, which are used as a benchmark to judge whether the actual grasp is qualified.

[0072] The allowable range for feature point position deviation is based on the material's precision requirements. For example, the positioning accuracy of electronic components needs to be ≤±0.02mm, while for plastic parts it can be relaxed to ±0.1mm, with a dynamically set threshold range. For rigid materials, due to the low risk of deformation, the deviation range can be set to a smaller value; for easily deformable materials, the range can be appropriately widened to avoid frequently triggering unnecessary corrections. Deviation calculation is based on the Euclidean distance between the actual coordinates of the feature point and the template coordinates. If the distance of any feature point exceeds this range, it is judged as "unqualified".

[0073] When a gripping position deviation is detected, the system will reduce the descent depth of the Z module in real time based on the material height deviation value output by the image processing subsystem. This ensures that the contact pressure between the gripper and the material is moderate, avoiding both material compression and deformation caused by excessive descent and unstable gripping caused by insufficient descent.

[0074] Real-time compensation of the rotary motor angle is based on the rotational offset calculated by the image processing subsystem, which is the difference between the actual angle of the material and the preset angle. The system sends an angle compensation command to the Z-axis rotary motor, driving the motor to rotate the corresponding angle in real time during the gripping process, so that the material posture is consistent with the preset gripping angle. The response time of this compensation process needs to be matched with the movement speed of the XY module to avoid new deviations caused by compensation delays.

[0075] The placement positioning image includes baselines for the placement area, such as preset X-axis and Y-axis positioning lines, which serve as the reference for coordinate correction. The system compensates for the original placement endpoint coordinates of the XY modules by identifying the relative positions of material edge feature points to the baselines and combining this with the correction offset output from the image processing subsystem. The correction process must consider the motion inertia of the XY modules, and a small buffer value is reserved for the compensation amount to avoid overshoot.

[0076] To reduce the impact of placement on positioning accuracy, the system adopts a "segmented deceleration" strategy. When far from the placement endpoint, the XY module moves at a higher speed; as it approaches the endpoint, the speed drops to 30% of the original speed; and at the moment of contact with the placement surface, the speed further drops to 5%. For fragile materials, the deceleration ratio can be increased to 20% of the original speed to minimize impact.

[0077] Based on historical data, the system generates predictive corrections for specific materials through statistical analysis. For example, if historical data shows that the average offset of a material in the X direction is +0.1mm, the system will pre-generate a trajectory correction of +0.1mm in the X direction and add this value in advance when the XY module plans the motion trajectory, achieving "predictive compensation." The correction is updated in real time through the historical data correction module to ensure adaptation to changes in material conditions, such as wear and batch differences.

[0078] The difference between the real-time corrected offset obtained from aerial photography and the predicted correction reflects the difference between the actual deviation and the predicted deviation. The setting of the preset threshold needs to balance accuracy and system stability. If the difference is less than or equal to the threshold, it means that the predicted compensation is effective and no additional adjustment is needed; if the difference is greater than the threshold, it means that the actual deviation exceeds the predicted range and real-time compensation needs to be triggered.

[0079] Real-time compensation is achieved by sending pulse signals to the motor driver, with the signal frequency being positively correlated with the compensation amount. For example, when the difference is +0.08mm, the system sends an 800-pulse positive compensation signal to the XY module motor, driving the motor to rotate an additional angle to achieve position correction. This process requires transmission via high-speed buses such as EtherCAT to ensure a delay of ≤1ms, avoiding any impact on motion continuity.

[0080] Anomaly Handling and Recording Mechanism: When the correction offset exceeds the preset threshold N times consecutively, the system determines it to be in an "abnormal state". This may be caused by sudden changes in material characteristics, equipment failure, etc., and immediately triggers a three-level response: ① Suspend the feeding operation to avoid batch non-conforming; ② Issue an audible and visual alarm to prompt the operator to intervene; ③ Record the characteristic parameters of the abnormal material, including surface roughness, elastic deformation coefficient, deviation between actual size and theoretical size, etc., and associate them with the corresponding aerial photography images and correction data to provide a basis for subsequent troubleshooting.

[0081] Through the above design, the motion correction subsystem achieves full-process precision control of "predictive compensation, real-time correction, and fault tolerance", which not only ensures positioning accuracy under high-speed motion, but also further enhances the flexibility and reliability of the system by dynamically adjusting to adapt to the characteristics of different materials.

[0082] A flexible feeding control device is used for flexible feeding of materials.

[0083] The flexible feeding control device is the physical carrier for realizing "flexible feeding". Its core lies in getting rid of the limitation of the traditional vibratory feeder's "fixed material channel" and realizing the compatible supply of materials of multiple specifications through adjustable structure and controllable vibration.

[0084] This material tray serves as a material-carrying platform, with retractable adjustment mechanisms, such as slide rails and drive cylinders, on its edges to adjust its size according to the material dimensions. The tray's surface has no fixed material channels; instead, surface texture reduces material slippage, allowing the material to disperse freely under vibration, avoiding the frictional damage of traditional material channels. Furthermore, reference markers on the tray's edge serve as coordinate references for image recognition, aiding in determining the absolute position of the material.

[0085] Vibration drive components, such as electromagnetic vibrators, are installed below the material tray. Their vibration frequency and intensity can be adjusted via the vibration control module. Low-frequency vibration is used for lightweight materials to avoid splashing, while high-frequency vibration is used for easily sticky materials to promote dispersion. More importantly, the vibration control module is linked with the XYZ multi-module control subsystem. Before the gripping action is executed, it receives a gripping preparation signal and stops vibration in advance to reduce the micro-displacement of materials caused by vibration. After gripping is completed, vibration is resumed based on feedback signals, allowing the remaining material to redisperse and providing a uniform material distribution for the next gripping operation.

[0086] The tray status detection components, such as vision sensors or laser rangefinders installed on the edge of the tray, will collect images of the current size of the tray and the material distribution in real time, and transmit this information to the XYZ multi-module control subsystem as the core input of the tray information to help it determine the material coordinates more accurately.

[0087] In summary, through the collaborative logic of "XYZ multi-module motion execution, dynamic detection by the flying camera subsystem, precise analysis by the image processing subsystem, real-time correction by the motion correction subsystem, and adaptive supply by the flexible feeding device", the various subsystems form a complete closed-loop control, ultimately achieving efficient, low-damage, and high-precision feeding of materials of various specifications.

[0088] Specifically, such as Figures 1 to 5 As shown, the material tray status detection component transmits the collected real-time size information of the adjustable flexible material tray and the material distribution image to the XYZ multi-module control subsystem as the core input of the material tray information, which is used by the XYZ multi-module control subsystem to determine the material coordinates in combination with historical material data; The vibration control module receives the grabbing preparation signal from the XYZ multi-module control subsystem. Within a preset time period before the flying grabbing subsystem triggers shooting, it controls the vibration drive component to stop vibrating, so that the material is in a stable state to cooperate with the flying grabbing subsystem to obtain a clear grabbing positioning image. According to the grabbing completion signal fed back by the XYZ multi-module control subsystem, the vibration control module adjusts the vibration drive component to resume vibration, so that the remaining material is redispersed, providing a material distribution basis for the next grabbing.

[0089] Specifically, such as Figures 1 to 5 As shown, it also includes a historical data correction module, which saves the actual offset of the placement position and the gripping position and the correction parameters to the historical material database after each feeding is completed, and is used to update the predicted correction amount.

[0090] Historical data correction module: This refers to a software module integrated into the system control unit, which has the functions of data acquisition, storage, analysis, and prediction correction updates. Through real-time communication with the XYZ multi-module control subsystem and motion correction subsystem, it realizes closed-loop data flow. Its core function is to transform discrete material feeding process data into a reusable empirical model for the system, reducing dependence on initial parameters.

[0091] Actual offset: This refers to the deviation between the actual gripping and placement position of the material and the preset theoretical position after each loading. It includes the translational offset (Δx, Δy) in the XY direction and the rotational offset Δθ. This data is calculated by the image processing subsystem based on the aerial image and is a direct indicator of the actual loading accuracy. For example, if the actual coordinates of the material center during gripping are (100.2, 200.1) and the theoretical coordinates are (100.0, 200.0), then the actual translational offset is Δx = +0.2mm and Δy = +0.1mm.

[0092] Correction parameters: These refer to the specific parameter values ​​adjusted by the motion correction subsystem to eliminate actual offset, including but not limited to: the adjustment amount of the Z-module descent depth, such as correcting it from 15mm to 14.8mm; the angular compensation amount of the rotary motor, such as +2°; the correction value of the XY module placement endpoint coordinates, such as +0.3mm; and the adjustment ratio of the placement speed, such as reducing it from 300mm / s to 90mm / s. These parameters directly reflect the system's response strategy to deviations.

[0093] Historical material database: This refers to a structured data storage unit used to store the aforementioned actual offsets and correction parameters, categorized and indexed by material type and tray size. The database also records the timestamp of data generation, material loading environment parameters such as temperature and humidity, and equipment status such as motor operating current, providing a foundation for subsequent data analysis.

[0094] Its prediction correction update mechanism includes, The historical data correction module uses a sliding window algorithm to dynamically analyze historical data. It uses the most recent N data points (where N can be set according to the feeding cycle time, e.g., 100 qualified feeding data points) as the sample window to calculate the statistical characteristics of the actual offset. Mean, such as the mean value μx of Δx: reflects the systematic deviation trend of this type of material, such as the continuous X-axis deviation caused by the wear of the material tray; Variance, such as the variance σx of Δx: reflects the degree of dispersion of the offset; the smaller the variance, the more stable the offset. Correlation, such as the correlation coefficient between Δx and vibration frequency: identifies key environmental factors affecting the offset.

[0095] The iterative logic for the predicted correction is as follows: based on the above statistical characteristics, the module updates the original predicted correction with weights. The new predicted correction is equal to the original predicted correction multiplied by the weight coefficient plus the window mean multiplied by 1 minus the weight coefficient.

[0096] The weighting coefficient (ranging from 0 to 1) is dynamically determined by the variance of the offset within the current sample window. The weighting coefficient equals 1 divided by (1 plus an empirical coefficient multiplied by the variance). The empirical coefficient is set to 100. The variance is calculated based on the offset of the most recent 100 qualified data captures within the sliding window, such as the offset in the X direction. Windows are maintained separately for each material type. A smaller variance indicates data stability; for example, when the variance is less than 0.0004, the weighting coefficient approaches 0.2, new data has a higher weight, and the system responds quickly to material changes. A larger variance indicates greater data volatility; for example, when the variance is greater than 0.01, the weighting coefficient approaches 0.9, the original correction has a higher weight, and the system remains stable. In practical applications, the weighting coefficient is limited, with a lower limit of 0.2 and an upper limit of 0.9. When the system detects a change in material type, such as switching from metal parts to plastic parts, the module automatically switches the sample window to the historical data of that type of material. If it is a new material and there is no historical data, the initial prediction correction is set to 0, the initial weight coefficient is set to 0.5, and the initial window is quickly built after the first 5 feedings. Starting from the 6th feeding, the system switches to the above dynamic adjustment rules.

[0097] The foregoing has illustrated and described the basic features, principles, and advantages of the present invention. It should be noted that the present invention is not limited to the above embodiments, but only to some embodiments. Any improvements and additions made without departing from the spirit and scope of the present invention are considered to be within the scope of protection of the present invention.

Claims

1. A flexible material feeding vision system based on an XYZ multi-module motion control system, characterized in that, include: The XYZ multi-module control subsystem includes an XYZ multi-module motion control device, which obtains material coordinates based on the preset number of gripping stations and material tray information, combined with historical material data, and controls the XYZ multi-module to grip and place materials according to the material coordinates. The flying camera grabbing subsystem captures images of the grabbing position of the XYZ multi-module motion control device during the grabbing process using a flying camera as the grabbing positioning image, and also captures images of the material placement position as the placement positioning image. The image processing subsystem performs image blur repair processing on the capture and placement images using a flying image processing strategy, calculates the image offset, and performs image correction based on the image offset and flexible feeding characteristic parameters to obtain the processed capture and placement images. The motion correction subsystem, based on the captured image and the placement image, performs a secondary confirmation of the captured and placement positions of the XYZ multi-module motion control subsystem through a motion correction strategy, and corrects the captured and placement parameters of the XYZ multi-module motion control device. A flexible feeding control device is used for flexible feeding of materials.

2. The flexible material feeding vision system based on an XYZ multi-module motion control system according to claim 1, characterized in that, The XYZ multi-module control subsystem includes preset gripping data and tray information. The gripping data includes the number of gripping stations and the type of gripped material. The tray information includes the tray size and tray image. Based on the tray size and tray image, and combined with historical material data, the position of the material in the tray is determined. The material with the same number of gripping stations is selected, and the material coordinates are determined. Based on the number of gripping stations and the material coordinates, the gripping motion direction, gripping angle, and gripping motion speed of the XY module motion control device of the XYZ multi-module motion control device are determined. Based on the placement position, the placement motion direction and placement motion speed of the XY module are determined, and the Z module descent depth of the XYZ multi-module motion control device is set.

3. The flexible material feeding vision system based on an XYZ multi-module motion control system according to claim 1, characterized in that, The flying camera grabbing subsystem includes a synchronization triggering module and an image acquisition device. When the XYZ multi-module motion control device moves along the XY direction at a preset speed, the synchronization triggering module triggers the camera to capture images within a preset time when the material enters the camera's field of view, based on the real-time position signal of the XY module of the XYZ multi-module motion control device. The shooting frequency is positively correlated with the movement speed of the XYZ multi-module motion control device. The shooting area of ​​the grabbing positioning image includes the entire material and edge feature points, and the placement positioning image includes the baseline of the placement area.

4. The flexible material feeding vision system based on an XYZ multi-module motion control system according to claim 1, characterized in that, The flexible feeding feature parameters in the image processing subsystem include material surface roughness, used to distinguish the reflectivity differences of different materials; material elastic deformation coefficient, used to correct the positional displacement of easily deformable workpieces caused by gripping pressure; and residual displacement of the material tray vibration, used to compensate for the micro-displacement of the material caused by the vibration of the flexible vibrating tray. When performing image correction, blurring is first eliminated by blur repair processing to eliminate the motion blur caused by high-speed movement, and then the image offset is weighted and corrected in combination with the flexible feeding feature parameters.

5. A flexible material feeding vision system based on an XYZ multi-module motion control system according to claim 1, characterized in that, The aerial photography image processing strategy includes image preprocessing, which involves grayscale conversion and noise reduction of the captured positioning image and the placement positioning image while preserving edge features. Key feature points of the material are identified using a feature extraction algorithm, and the coordinates of these key feature points in the image coordinate system are calculated. If the image clarity score calculated based on the flexible feeding feature parameters is less than a preset threshold, blur repair processing is initiated, and feature points are re-extracted after repair. The actual coordinates of the key feature points are compared with the theoretical coordinates based on historical feeding data to obtain the translational and rotational offsets in the X and Y directions. These translational and rotational offsets are then combined with the material's flexible feature parameters for compensation.

6. A flexible material feeding vision system based on an XYZ multi-module motion control system according to claim 1, characterized in that, The motion correction subsystem includes, during secondary confirmation, comparing the processed captured image with a preset qualified capture template, wherein the qualified capture template includes an allowable range for feature point position deviation. If the deviation exceeds the allowable range, triggering capture parameter correction, wherein the capture parameters include adjusting the descent depth of the Z module and real-time compensation of the rotary motor angle based on the rotation offset; and placement parameters include correcting the placement endpoint coordinates of the XY module according to the correction offset of the placement positioning image and adjusting the placement speed.

7. A flexible material feeding vision system based on an XYZ multi-module motion control system according to claim 1, characterized in that, The motion correction strategy includes: pre-generating the motion trajectory correction amount of the XYZ module based on historical material data; comparing the real-time correction offset obtained by the flying camera with the predicted correction amount; if the difference is greater than a preset threshold, real-time compensation is triggered; the motor driver is adjusted by signal; when the correction offset exceeds the preset threshold for several consecutive times, the feeding is automatically paused and an alarm is issued; at the same time, the characteristic parameters of the abnormal material are recorded.

8. A flexible material feeding vision system based on an XYZ multi-module motion control system according to claim 7, characterized in that, It also includes a historical data correction module, which saves the actual offset of the placement position and the gripping position and the correction parameters to the historical feeding database after each feeding is completed, and is used to update the predicted correction amount.

9. A flexible material feeding vision system based on an XYZ multi-module motion control system according to claim 1, characterized in that, The flexible feeding control device includes, An adjustable flexible tray is used as a material carrying platform. Its size can be adjusted by the tray adjustment mechanism to adapt to materials of different specifications. The planar area of ​​the adjustable flexible tray is used for the dispersed placement of materials, and its edge is provided with reference marks for positioning. The vibration drive assembly is installed below the adjustable flexible tray to provide vibration with controllable frequency and intensity, so that the material is dispersed in the tray and moves without a fixed path, avoiding frictional damage caused by fixed material channels. The vibration control module is electrically connected to the XYZ multi-module control subsystem and the vibration drive component. It is used to receive motion control signals and adjust the vibration parameters of the vibration drive component. The vibration parameters include vibration frequency and start / stop time. The module stops vibration within a preset time before the XYZ multi-module motion control device performs the grasping action to reduce the micro-displacement of the material caused by vibration. The tray status detection component is used to collect real-time size information and material distribution images of the adjustable flexible tray, and transmit them to the XYZ multi-module control subsystem as part of the tray information to help determine the material coordinates.

10. A flexible material feeding vision system based on an XYZ multi-module motion control system according to claim 9, characterized in that, The tray status detection component transmits the collected real-time size information and material distribution image of the adjustable flexible tray to the XYZ multi-module control subsystem as the core input of the tray information, which is used by the XYZ multi-module control subsystem to determine the material coordinates in combination with historical material data. The vibration control module receives the grabbing preparation signal from the XYZ multi-module control subsystem. Within a preset time period before the flying grabbing subsystem triggers shooting, it controls the vibration drive component to stop vibrating, so that the material is in a stable state to cooperate with the flying grabbing subsystem to obtain a clear grabbing positioning image. According to the grabbing completion signal fed back by the XYZ multi-module control subsystem, the vibration control module adjusts the vibration drive component to resume vibration, so that the remaining material is redispersed, providing a material distribution basis for the next grabbing.

Images