High-precision material edge deviation rectifying method and controller

Abstract

The invention relates to a high-precision material edge correction method. The method comprises the following steps: S1, deploying a machine vision system; s2, deploying a sensor network on a material transmission path; s3, preprocessing the collected material edge image, performing edge detection on the preprocessed image, and extracting an accurate contour of the material edge; an edge detection result and sensor data are accurately aligned and fused in time and space, and a multi-dimensional data set is obtained; s4, according to material characteristics and transmission system dynamics, establishing a physical model of material movement for predicting a future position of a material edge; s5, taking the multi-dimensional data set as the input of a physical model to train the physical model; s6, predicting the edge position of the material by using the trained physical model, comparing the predicted edge position of the material with the edge position of the material detected in real time, and calculating a deviation value; and generating a specific control instruction based on the deviation value, and correcting the deviation value.

Description

Technical Field

[0001] The present invention relates to the field of automation technology, and more specifically, to a high-precision material edge correction method and a controller. Background Art

[0002] In the field of modern industrial automation, material transmission and positioning are key links in the production line, and their accuracy and efficiency directly affect product quality and production efficiency. Especially in high-speed, high-precision production environments, accurate material edge correction is particularly important. However, traditional material correction methods often rely on mechanical structures, fixed-parameter control algorithms or simple sensor detection. These methods often seem to be inadequate when faced with complex and changeable material shapes, transmission speeds and environmental conditions.

[0003] First, the diversity of material shapes poses a severe challenge to the accuracy of the correction. Different materials may have different edge features, such as straight lines, curves, and unevenness. Traditional mechanical structures are difficult to flexibly adapt to these changes, resulting in limited correction accuracy.

[0004] Secondly, changes in transmission speed also increase the difficulty of correction. During high-speed transmission, the position and shape of the edge of the material may change rapidly due to factors such as inertia and vibration. Traditional fixed parameter control algorithms are difficult to track in real time and accurately adjust the correction strategy. Summary of the invention

[0005] The purpose of this invention is to provide a high-precision material edge correction technology solution based on multimodal fusion and real-time feedback, aiming to solve the problems of insufficient accuracy and poor adaptability of existing material correction methods when facing complex and changeable material shapes, transmission speeds and environmental conditions. By integrating machine vision, sensor data and real-time physical models, high-precision, real-time detection and correction of material edges can be achieved, thereby improving the overall efficiency of the production line and product quality.

[0006] In order to achieve the above invention purpose, the technical solution adopted is:

[0007] A high-precision material edge correction method comprises the following steps:

[0008] S1. Deploy a machine vision system to collect edge images of materials;

[0009] S2. A sensor network is deployed on the material transmission path to collect the physical position, speed and acceleration data of the material; the data collected by the sensor network is collectively referred to as sensor data;

[0010] S3. Preprocess the collected material edge image, perform edge detection on the preprocessed image, and extract the precise contour of the material edge; accurately align and fuse the edge detection result with the sensor data in time and space to obtain a multidimensional data set;

[0011] S4. Based on the material characteristics and the dynamics of the transmission system, a physical model of the material movement is established to predict the future position of the material edge;

[0012] S5. Using the multidimensional data set as input of the physical model to train the physical model, the training process is as follows:

[0013] The physical model outputs the predicted material edge position; the predicted material edge position is compared with the detected material edge position, the deviation is calculated, and the parameters of the physical model are dynamically adjusted based on the deviation;

[0014] S6. Use the trained physical model to predict the edge position of the material, compare the predicted edge position of the material with the real-time detected edge position of the material, and calculate the deviation; generate specific control instructions based on the deviation to correct the deviation.

[0015] Preferably, the step S3 pre-processes the collected material edge image, specifically including the following operations: denoising, contrast enhancement, and edge sharpening.

[0016] Preferably, the specific process of edge detection on the preprocessed image is as follows:

[0017] Use Gaussian pyramid to construct a multi-scale space of the image, where each scale corresponds to a resolution version of the image;

[0018] At each scale, edge detection is performed using an edge detection algorithm;

[0019] The edge detection results at each scale are fused into the original edge map using a confidence-based fusion method.

[0020] Continuous contour lines are extracted from the original edge map, and the extracted contour lines are evaluated to check whether they meet the expected material shape and size. If they meet the requirements, the original edge map is output as the final edge detection result; if not, the contour is optimized using a graph theory-based method; the optimized contour is merged with the original edge map to generate the final material edge contour map for output.

[0021] Preferably, at each scale, the specific process of edge detection using the edge detection algorithm is as follows:

[0022] Calculate the gradient of each pixel in the image;

[0023] Determine whether the gradient of the current pixel is greater than the set threshold. If so, confirm it as an edge point; obtain preliminary edge detection results;

[0024] The preliminary edge detection results are processed using the non-maximum suppression method, and then the isolated edge points are connected through the edge connection algorithm to form a complete edge line.

[0025] Preferably, the specific process of using the confidence-based fusion method to fuse the edge detection results at each scale into the original edge map is as follows:

[0026] For the edge detection results at each scale, calculate the confidence of each edge point;

[0027] Normalize the confidence calculated at different scales;

[0028] According to the confidence of the edges at each scale, different weights are assigned to the edge detection results at each scale;

[0029] The edge detection results at different scales are fused based on the weighted average operation of the edge point positions.

[0030] Preferably, the edge detection results are precisely aligned and fused with the sensor data in time and space, as follows:

[0031] Time alignment: Use timestamps or external synchronization signals to accurately align edge detection results with sensor data in time;

[0032] Spatial alignment: First, calibrate the camera to obtain the camera's intrinsic and extrinsic parameters; calculate the mapping relationship between sensor data and image pixels based on the sensor's installation position and posture; map the sensor data onto the image plane to achieve pixel-level alignment;

[0033] Fusion: Weighted averaging of data from different sources to obtain a combined result.

[0034] Preferably, in step S5, a gradient descent method is adopted to dynamically adjust the parameters of the physical model.

[0035] Preferably, the specific process of dynamically adjusting the parameters of the physical model by using the gradient descent method is as follows:

[0036] S11.

[0037] Define a loss function:

[0038] Where N is the number of samples, x i is the input of the i-th sample, y i is the corresponding detected material edge position, f(θ; xi ) is the physical model in terms of parameters θ and input x i The predicted value under ;

[0039] S12. Calculate the gradient of the loss function L(θ) with respect to the model parameter θ

[0040] S13. Update the model parameter θ according to the following formula:

[0041]

[0042] Where α is the learning rate;

[0043] S14. Repeat S12 and S13 until L(θ) is less than the set threshold.

[0044] At the same time, the present invention also provides a deviation correction controller, which executes the high-precision material edge deviation correction method described above when running.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] Improve the accuracy of correction: By introducing high-resolution industrial cameras and advanced image processing algorithms, combined with deep learning models, high-precision recognition of material edges can be achieved. At the same time, combined with sensor data, a multi-modal information fusion system is built to further improve the accuracy and reliability of correction.

[0047] Enhance system adaptability: Use real-time physical models to predict material movement, and combine with dynamic feedback control mechanisms to enable the system to adjust the correction strategy in real time to adapt to changes in different material types, transmission speeds and environmental conditions.

[0048] Improve production efficiency: By achieving real-time, high-precision correction of material edges, production interruptions and scrap rates caused by inaccurate correction can be reduced, and the overall efficiency and capacity of the production line can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0050] Figure 1 The figure is a flow chart of a high-precision material edge correction method.

[0051] Figure 2 Schematic diagram of the edge detection algorithm.

[0052] Figure 3 This is the specific process diagram of fusing the edge detection results at various scales into the original edge map using the confidence-based fusion method. DETAILED DESCRIPTION

[0053] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Example 1

[0055] like Figure 1-3 As shown, the specific process of the high-precision material edge correction method is as follows:

[0056] Implementation Environment Preparation

[0057] 1. Machine vision system deployment:

[0058] Install a high-resolution industrial camera next to the material conveyor line to ensure that the image of the edge of the material can be clearly captured.

[0059] Configure the image processing unit, including the necessary software and algorithm library, for subsequent image preprocessing and edge detection.

[0060] 2. Sensor network deployment:

[0061] Position sensors, speed sensors and acceleration sensors are installed at key locations along the material transfer path to ensure that the physical state data of the material can be acquired in real time.

[0062] Calibrate the sensor to ensure data accuracy and consistency.

[0063] Implementation steps

[0064] S1. Machine vision system collects material edge images

[0065] Start the machine vision system to capture images of the material edge in real time.

[0066] S2. Sensor network collects physical data

[0067] The sensor network is started synchronously to collect the physical position, speed and acceleration data of the material in real time.

[0068] S3. Image preprocessing and edge detection

[0069] Preprocessing:

[0070] The collected images are denoised to reduce noise interference in the images.

[0071] Enhance the contrast of the image and make the edges of the material clearer.

[0072] Sharpen the edges of the image to highlight the details of the edges of the material.

[0073] Edge Detection:

[0074] Use Gaussian pyramid to construct a multi-scale space of images.

[0075] At each scale, an edge detection algorithm (such as the Canny algorithm) is used to calculate the gradient of each pixel in the image, and the edge point is determined based on the gradient size and the set threshold.

[0076] Non-maximum suppression and edge connection algorithms are applied to connect isolated edge points into complete edge lines.

[0077] Through the confidence-based fusion method, the edge detection results at each scale are fused into the final material edge contour map.

[0078] Continuous contour lines are extracted from the original edge map, and the extracted contour lines are evaluated to check whether they meet the expected material shape and size. If they meet the requirements, the original edge map is output as the final edge detection result; if not, the contour is optimized using a graph theory-based method; the optimized contour is merged with the original edge map to generate the final material edge contour map for output.

[0079] S4. Data alignment and fusion

[0080] Time alignment: Use timestamps or external synchronization signals to precisely align edge detection results with sensor data in time.

[0081] Spatial Alignment:

[0082] Perform camera calibration to obtain the camera's intrinsic and extrinsic parameters.

[0083] According to the installation position and posture of the sensor, the mapping relationship between the sensor data and the image pixels is calculated.

[0084] Map sensor data onto the image plane to achieve pixel-level alignment.

[0085] The aligned data were weighted averaged to obtain a comprehensive multidimensional dataset.

[0086] S5. Physical model training

[0087] Build a physical model of material movement based on material characteristics and transport system dynamics.

[0088] Take a multidimensional dataset as input and train a physics model.

[0089] Define a loss function to measure the deviation between the model prediction value and the actual detection value.

[0090] The gradient descent method is used to dynamically adjust the parameters of the physical model to minimize the loss function value.

[0091] The training process is repeated until the loss function value is less than the set threshold or the predetermined number of training rounds is reached.

[0092] S6. Edge position prediction and correction

[0093] The trained physical model is used to predict the edge position of the material.

[0094] The predicted material edge position is compared with the real-time detected material edge position to calculate the deviation.

[0095] Generate specific control instructions based on the deviation, such as adjusting the speed, direction or position of the conveyor belt.

[0096] The control instructions are sent to the actuators (such as motors, cylinders, etc.) to adjust the material position in real time.

[0097] S7. Monitoring and Optimization

[0098] During system operation, the web-edge correction effect is continuously monitored.

[0099] According to actual feedback, necessary optimization and adjustment are made to the physical model and control algorithm to improve the stability and accuracy of the system.

[0100] Example 2

[0101] On an automated packaging line, materials of various shapes and sizes (such as food packaging boxes, medicine bottles, etc.) need to be accurately aligned and packaged. Since the edge position of the material may be offset due to vibration, speed changes or mechanical errors during the transmission process, the subsequent packaging quality is affected. Therefore, a high-precision material edge correction method is introduced to ensure that the material can be accurately aligned before packaging.

[0102] Implementation steps

[0103] 1. System integration and deployment

[0104] Machine vision system: A high-resolution industrial camera is installed at the material entrance of the packaging line, equipped with an image processing unit to capture material edge images in real time.

[0105] Sensor network: Position sensors, speed sensors, and acceleration sensors are deployed on the material conveyor belt to obtain real-time physical status data of the material.

[0106] Control system: A control system that integrates a machine vision system, a sensor network, and actuators (such as pneumatic push rods, servo motors, etc.) to receive processed data and generate control instructions.

[0107] 2. Material edge image acquisition and processing

[0108] When the material enters the field of view of the machine vision system, the camera begins capturing images.

[0109] The acquired images are preprocessed, including denoising, contrast enhancement and edge sharpening, to improve the accuracy of edge detection.

[0110] A multi-scale edge detection algorithm is applied, combined with Gaussian pyramid and Canny edge detector to extract the precise edge contour of the material.

[0111] 3. Data alignment and fusion

[0112] Use timestamps to precisely align image data collected by a machine vision system with physical data acquired by a sensor network.

[0113] Through camera calibration and sensor position calibration, the image data and physical data are aligned in space.

[0114] The aligned data are fused to form a data set containing multi-dimensional information such as material edge position, speed and acceleration.

[0115] 4. Physical model prediction and parameter adjustment

[0116] Build a physical model of material movement based on material characteristics and transport system dynamics.

[0117] The fused dataset is used as input to train a physical model to predict the future position of the material edge.

[0118] The gradient descent method is used to dynamically adjust the parameters of the physical model to minimize the deviation between the predicted value and the actual detected value.

[0119] 5. Real-time deviation correction control

[0120] Before the material reaches the packaging area, the trained physical model is used to predict its edge position.

[0121] Compare the predicted position with the real-time detected position and calculate the deviation.

[0122] Generate specific control instructions based on the deviation, such as adjusting the speed and direction of the conveyor belt or triggering the pneumatic actuator for fine-tuning.

[0123] The actuator adjusts the material position in real time according to the control instructions to ensure that the material can be accurately aligned before packaging.

[0124] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0125] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0126] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0127] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or partly contributed to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0128] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high-precision material edge correction method, characterized by: The following steps are involved: S1. Deploy a machine vision system to collect edge images of materials; S2. A sensor network is deployed on the material transmission path to collect the physical position, speed and acceleration data of the material; the data collected by the sensor network is collectively referred to as sensor data; S3. Preprocess the collected material edge image, perform edge detection on the preprocessed image, and extract the precise contour of the material edge; accurately align and fuse the edge detection result with the sensor data in time and space to obtain a multidimensional data set; S4. Based on the material characteristics and the dynamics of the transmission system, a physical model of the material movement is established to predict the future position of the material edge; S5. Using the multidimensional data set as input of the physical model to train the physical model, the training process is as follows: The physical model outputs the predicted material edge position; the predicted material edge position is compared with the detected material edge position, the deviation is calculated, and the parameters of the physical model are dynamically adjusted based on the deviation; S6. Use the trained physical model to predict the edge position of the material, compare the predicted edge position of the material with the real-time detected edge position of the material, and calculate the deviation; generate specific control instructions based on the deviation to correct the deviation.

2. The high-precision material edge correction method according to claim 1, characterized in that: The step S3 pre-processes the collected material edge image, specifically including the following operations: denoising, contrast enhancement, and edge sharpening.

3. The high-precision material edge correction method according to claim 1, characterized in that: The specific process of edge detection on the preprocessed image is as follows: Use Gaussian pyramid to construct a multi-scale space of the image, where each scale corresponds to a resolution version of the image; At each scale, edge detection is performed using an edge detection algorithm; The edge detection results at each scale are fused into the original edge map using a confidence-based fusion method. Continuous contour lines are extracted from the original edge map, and the extracted contour lines are evaluated to check whether they meet the expected material shape and size. If they meet the requirements, the original edge map is output as the final edge detection result; if not, the contour is optimized using a graph theory-based method; the optimized contour is merged with the original edge map to generate the final material edge contour map for output.

4. The high-precision material edge correction method according to claim 3, characterized in that: At each scale, the specific process of edge detection using edge detection algorithm is as follows: Calculate the gradient of each pixel in the image; Determine whether the gradient of the current pixel is greater than the set threshold. If so, confirm it as an edge point; obtain preliminary edge detection results; The preliminary edge detection results are processed using the non-maximum suppression method, and then the isolated edge points are connected through the edge connection algorithm to form a complete edge line.

5. The high-precision material edge correction method according to claim 3, characterized in that: The specific process of using the confidence-based fusion method to fuse the edge detection results at each scale into the original edge map is as follows: For the edge detection results at each scale, calculate the confidence of each edge point; Normalize the confidence calculated at different scales; According to the confidence of the edges at each scale, different weights are assigned to the edge detection results at each scale; The edge detection results at different scales are fused based on the weighted average operation of the edge point positions.

6. The high-precision material edge correction method according to claim 1, characterized in that: The edge detection results are precisely aligned and fused with the sensor data in time and space, as follows: Time alignment: Use timestamps or external synchronization signals to accurately align edge detection results with sensor data in time; Spatial alignment: First, calibrate the camera to obtain the camera's internal and external parameters; calculate the mapping relationship between sensor data and image pixels based on the sensor's installation position and posture; Map the sensor data onto the image plane to achieve pixel-level alignment; Fusion: Weighted averaging of data from different sources to obtain a combined result.

7. The high-precision material edge correction method according to claim 1, characterized in that: In step S5, the parameters of the physical model are dynamically adjusted using a gradient descent method.

8. The high-precision material edge correction method according to claim 7, characterized in that: The specific process of dynamically adjusting the parameters of the physical model using the gradient descent method is as follows: S11. Define a loss function: Where N is the number of samples, x i is the input of the i-th sample, y i is the corresponding detected material edge position, f(θ; x i ) is the physical model in terms of parameters θ and input x i The predicted value under ; S12. Calculate the gradient of the loss function L(θ) with respect to the model parameter θ S13. Update the model parameter θ according to the following formula: Where α is the learning rate; S14. Repeat S12 and S13 until L(θ) is less than the set threshold.

9. A deviation correction controller, characterized in that: When it is running, the high-precision material edge correction method described in any one of claims 1 to 8 is executed.

Images