Industrial code reading control method based on equipment collaborative optimization and high-speed response system

Abstract

The invention relates to an industrial code reading control method and a high-speed response system based on device collaborative optimization, and the method comprises the steps: synchronizing image data and three-dimensional coordinates of a plurality of visual devices, and generating a multi-view image sequence with a unified timestamp; extracting two-dimensional bounding box coordinates of the target object to establish a motion trail prediction model; constructing a space projection relation by combining the three-dimensional coordinates of the equipment; fusing the multi-view data to solve the three-dimensional position and attitude of the target object; calculating illumination, object distance and motion blur coefficients in real time based on pose parameters, and dynamically adjusting exposure time, gain and focal length of each device; acquiring an image by using the adjusted equipment, detecting a code system area, and performing geometric correction to generate a standardized image; and outputting a code reading result after decoding. And through multi-device collaborative optimization and dynamic parameter adjustment, the decoding success rate and efficiency under complex working conditions are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the field of industrial machine vision, and in particular relates to an industrial code reading control method and a high-speed response system based on equipment collaborative optimization. Background Art

[0002] With the rapid development of industrial automation and machine vision technologies, intelligent barcode reading systems based on vision sensors are gaining widespread application in logistics sorting, product traceability, and other fields. This technology uses image acquisition devices to capture barcode information on object surfaces and employs decoding algorithms for automatic recognition, improving production process efficiency. Traditionally, industrial barcode reading relies primarily on a single, fixed vision device. The processing flow typically involves installing an industrial camera at a preset location, setting fixed exposure parameters and focal length based on empirical data; triggering a single image capture when the target object passes through the field of view; and locating and decoding the code area in the captured image. For static or low-speed scenarios, this approach can meet basic needs. However, the current single-device barcode reading method has obvious defects: poor dynamic adaptability: high-speed moving objects are prone to motion blur, and fixed exposure parameters cannot adapt to sudden changes in lighting (such as reflections and shadows), resulting in deterioration of image quality; limited viewing angle coverage: a single device is difficult to capture multi-angle code information, and when the barcode is partially blocked or on a curved surface, the decoding failure rate increases significantly; lack of collaborative ability: when multiple devices work independently, there is a lack of data fusion mechanism, and it is impossible to optimize global imaging parameters through posture prediction, resulting in resource waste and recognition delays. Summary of the Invention

[0003] Based on this, it is necessary to provide an industrial code reading control method and a high-speed response system based on equipment collaborative optimization that can solve the above problems.

[0004] In a first aspect, the present application provides an industrial code reading control method based on equipment collaborative optimization, comprising:

[0005] Acquire image data and 3D position coordinates of multiple visual devices, and generate multi-view image sequences with unified timestamps through clock calibration;

[0006] Extract the two-dimensional bounding box coordinates of the target object in the multi-view image sequence, calculate its motion state parameters to establish a motion trajectory prediction model;

[0007] Combine the motion trajectory prediction model with the three-dimensional position coordinates of the visual device to establish the spatial projection relationship of the target object under different viewing angles;

[0008] Based on the spatial projection relationship, the observation data from each perspective is fused to obtain the three-dimensional spatial position and posture parameters of the target object;

[0009] Based on the 3D spatial position and posture parameters of the target object, the lighting environment parameters, imaging object distance and motion blur coefficient are calculated, and the exposure time, gain coefficient and focal length parameters of each visual device are adjusted in real time accordingly;

[0010] Using the adjusted vision devices to capture images, detect the symbology areas in the images and extract geometric features. Based on these geometric features, geometric correction is performed to generate a standardized symbology image. The symbology includes common barcode types, whose geometric features include module size, locator shape, or row and column arrangement rules.

[0011] Decode and recognize the standardized code image to generate the code reading result.

[0012] In one embodiment, after adjusting the exposure time, gain coefficient and focal length parameters of each visual device in real time, the method further includes:

[0013] Determine the initial code reading task allocation plan based on the image acquisition performance parameters of each visual device. Image acquisition performance parameters include resolution, frame rate, and field of view.

[0014] Real-time monitoring of the processing status data of each visual device for the assigned task, including image processing delay, decoding failure rate and resource utilization;

[0015] Calculate the task weight adjustment coefficient of each visual device based on the processing status data;

[0016] The initial task allocation plan is weighted optimized based on the task weight adjustment coefficient to generate an optimized allocation plan.

[0017] In one embodiment, an initial code reading task allocation scheme is determined based on image acquisition performance parameters of each visual device, including:

[0018] Calculate the resolvability of the minimum symbol size based on the resolution parameters of each visual device;

[0019] Calculate the effective coverage area of ​​each visual device based on the relationship between the field of view angle parameters and the spatial position of the target object;

[0020] Fusing resolution and effective coverage area to generate device allocation parameters;

[0021] The code area to be identified is allocated to the corresponding visual device according to the allocation parameters, and the initial task allocation plan is generated.

[0022] In one embodiment, after generating the optimized allocation plan, the method further includes:

[0023] Write the code reading results into the target associated database;

[0024] If there are differences in the decoded content between N consecutive code reading results associated with the same target object ID in the target association database (N≥2), the manual re-inspection process will be triggered;

[0025] Update the exposure time, gain factor, focal length parameters and device allocation parameters of each visual device based on the manual re-inspection results.

[0026] In one embodiment, decoding and identifying a standardized code image to generate a code reading result includes:

[0027] Perform decoding recognition on the standardized code image and calculate the decoding confidence of each visual device;

[0028] If the decoding confidence of any vision device is lower than the preset threshold, the code reading result is generated according to the following steps:

[0029] Eliminating decoding results of visual devices whose decoding confidence is lower than a preset threshold to obtain a preferred decoding result;

[0030] Generate a corresponding weighting coefficient according to the confidence value of the visual device of the preferred decoding result;

[0031] The preferred decoding results are weighted and aggregated based on the weighting coefficients to generate a code reading result including a confidence score.

[0032] In one embodiment, the method further comprises:

[0033] Obtain historical decoding time data, equipment processing performance parameters and code complexity level to form the calculation basis for the preset time window;

[0034] Based on the calculation basis, the initial time window is generated through the time window algorithm model, and adaptively adjusted based on the current equipment load status to obtain the adjusted time window;

[0035] Real-time monitoring to determine whether the code reading results generated within the adjustment time window meet the scoring requirements generated based on the credibility scores of each visual device;

[0036] If the code reading result within the adjustment time window does not meet the scoring requirements, the current code system recognition process is skipped and the current code system is marked as pending re-inspection.

[0037] In one embodiment, the spatial projection relationship of the target object under different viewing angles is established and the observation data of each viewing angle is integrated to obtain the three-dimensional spatial position and posture parameters of the target object, including constructing the following optimization objective function to solve the optimal three-dimensional spatial position:

[0038]

[0039] Among them, N represents the total number of visual devices involved in fusion, P irepresents the two-dimensional observation coordinates of the target object in the i-th visual device image, π i (X, Θ) represents the projection function of the i-th visual device, the input is the three-dimensional position X and posture Θ, and the output is the two-dimensional coordinate projected onto the image plane of the device, α i Represents the observation weight coefficient of the i-th visual device, which is dynamically calculated by the image quality index of the device. and represents the predicted position and predicted posture of the target object output by the motion trajectory prediction model, and β and γ represent the regularization coefficients of position and posture, respectively.

[0040] Secondly, this application also provides an industrial code reading control high-speed response system based on equipment collaborative optimization, including:

[0041] Multi-source data synchronization module, used to obtain image data and 3D position coordinates of multiple visual devices, and generate multi-view image sequences with unified timestamps through clock calibration;

[0042] The motion trajectory modeling module is used to extract the two-dimensional bounding box coordinates of the target object in the multi-view image sequence and calculate its motion state parameters to establish a motion trajectory prediction model;

[0043] The spatial projection relationship building module is used to combine the motion trajectory prediction model with the three-dimensional position coordinates of the visual device to establish the spatial projection relationship of the target object under different viewing angles;

[0044] The 3D position and posture fusion calculation module is used to fuse the observation data from each perspective based on the spatial projection relationship to obtain the 3D spatial position and posture parameters of the target object;

[0045] The device parameter dynamic control module is used to calculate the lighting environment parameters, imaging object distance and motion blur coefficient based on the three-dimensional spatial position and posture parameters of the target object, and adjust the exposure time, gain coefficient and focal length parameters of each visual device in real time accordingly;

[0046] The code standardization processing module is used to use the adjusted visual devices to collect images, detect the code area in the image and extract geometric features, perform geometric correction based on the geometric features, and generate a standardized code image;

[0047] The decoding and recognition output module is used to decode and recognize standardized code images and generate code reading results.

[0048] In a third aspect, the present application also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned industrial code reading control method based on device collaborative optimization when executing the computer program.

[0049] In a fourth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned industrial code reading control method based on device collaborative optimization.

[0050] The above-mentioned industrial code reading control method and high-speed response system, computer equipment and storage medium based on equipment collaborative optimization generate a multi-perspective image sequence with a unified time stamp through clock calibration, combine the target object motion trajectory prediction model with the three-dimensional coordinates of the visual equipment to construct a spatial projection relationship, realize high-precision fusion of multi-perspective observation data, and effectively solve the three-dimensional spatial position and posture parameters of the target object; based on the posture parameters, the lighting environment, imaging object distance and motion blur coefficient are solved in real time, and the exposure time, gain coefficient and focal length parameters of each visual device are dynamically and collaboratively optimized to overcome image blur and light interference in high-speed motion scenes; through geometric correction, standardized code images are generated and decoded for recognition, breaking through the limitations of the single device's perspective while ensuring the reliability of the output results, achieving the triple technical effects of improved dynamic adaptability, expanded effective perspective coverage and enhanced multi-device collaborative efficiency, and systematically solving the problem of low recognition rate caused by fixed parameters, single perspective and lack of collaboration in traditional industrial code reading. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 This is a flow chart of an industrial code reading control method based on equipment collaborative optimization of the present invention;

[0053] Figure 2 This is a module diagram of an industrial code reading control high-speed response system based on equipment collaborative optimization of the present invention;

[0054] Figure 3 This is a module diagram of an industrial code reading control high-speed response system based on equipment collaborative optimization in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0056] The implementation environment of the present invention includes: multiple visual devices deployed based on different spatial distributions, which can be integrated within a single barcode reader or multiple barcode readers, and both the single barcode reader and the multiple barcode readers are connected to a computing terminal and server via a high-speed communication interface (such as Industrial Ethernet, Camera Link, USB4, or GigE PoE+). When integrated within a single barcode reader, the visual devices are arranged in an array within the housing (e.g., 2-8 industrial camera modules with optical axes arranged at preset angles such as 0°, 45°, and 90°), achieving clock synchronization and data transmission through a shared backplane circuit, and then communicating with the computing terminal via a single high-speed interface. When integrated within multiple barcode readers, each barcode reader contains at least one visual device, and each barcode reader is connected to the computing terminal and server via its own high-speed communication interface, forming a central processing unit. On a high-speed industrial assembly line, when a target object enters the collaborative perception area, the multi-view images collected by the code reader are synchronously processed by the computing terminal to generate posture parameters. The server then calculates the illumination / motion blur coefficients based on these parameters and sends control instructions to the visual device. At the same time, through the collaboration between the code standardization processing module and the decoding and recognition output module, closed-loop optimization of hardware resources and software modules is achieved.

[0057] In one embodiment, Figure 1 As shown, an industrial code reading control method based on device collaborative optimization is provided. This embodiment uses a terminal as an example to illustrate the execution subject of this method. It should be understood that this method can also be applied to a server, or a collaborative processing system composed of a terminal and a server, and is implemented through data interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0058] S01, acquiring image data and three-dimensional position coordinates of multiple visual devices, and generating a multi-view image sequence with a unified timestamp through clock calibration.

[0059] Among them, image data of multiple visual devices (such as industrial cameras, smart sensors, where multiple visual devices can be integrated into a barcode reader) can be collected in real time through interfaces such as industrial Ethernet and high-speed buses (such as Camera Link), and the three-dimensional spatial coordinates of each device can be obtained (which can be determined through pre-hand-eye calibration or lidar positioning). At the same time, the clock calibration step can use (such as the IEEE1588 clock synchronization protocol or hardware clock calibration module) to uniformly calibrate the acquisition clocks of all visual devices, generate an accurate and unified timestamp for each frame of image data, and form a multi-view image sequence with temporal and spatial consistency.

[0060] S02, extracting the two-dimensional bounding box coordinates of the target object in the multi-view image sequence, and calculating its motion state parameters to establish a motion trajectory prediction model.

[0061] Among them, target detection algorithms such as YOLO, SSD or Faster R-CNN can be used to identify the target object in each frame of the multi-view image sequence and extract its two-dimensional bounding box coordinates (such as coordinates (x, y, z, h)). Kalman filtering or particle filtering algorithms are used to smooth the bounding box coordinates across frames and calculate the motion state parameters of the target object, including linear velocity such as (v x , v y ), acceleration such as (a x , a y ) and angular velocity. Motion trajectory prediction models can be established using linear regression, LSTM neural networks, or physical models (such as uniform velocity / uniform acceleration motion models). By learning from historical trajectory data, the target object's position in future frames can be predicted, providing a basis for motion trends for subsequent spatial projection relationship construction and dynamic adjustment of device parameters, solving the real-time matching problem of image acquisition in high-speed motion scenes.

[0062] S03, combining the motion trajectory prediction model with the three-dimensional position coordinates of the visual device to establish the spatial projection relationship of the target object under different viewing angles.

[0063] Among them, based on the established motion trajectory prediction model, the predicted three-dimensional position and posture parameters of the target object at the future moment are obtained. Combined with the three-dimensional spatial coordinates (including the origin position and coordinate axis direction of the device coordinate system) obtained by hand-eye calibration or lidar pre-calibration of each visual device, the pinhole camera projection model is used to construct the spatial projection relationship under different perspectives. For example, for the i-th visual device, its projection function π is defined i This function maps the target object's 3D spatial coordinates (X, Θ) (where X is the position vector and Θ is the pose parameter) to the device image plane using a homogeneous coordinate transformation, the camera's intrinsic parameter matrix (focal length, principal point coordinates, etc.), and the extrinsic parameter matrix (rotation matrix R and translation vector T), generating 2D projection coordinates. By fusing the motion trajectory prediction results with the device's 3D coordinates, a cross-view projection equation is established, enabling precise mapping of the target object's position within each device's field of view, providing a geometric constraint foundation for subsequent multi-view data fusion.

[0064] S04, fusing the observation data from each perspective based on the spatial projection relationship to obtain the three-dimensional spatial position and posture parameters of the target object.

[0065] Among them, based on the spatial projection relationship, an optimization objective function of multi-view observation data fusion can be constructed, and the three-dimensional spatial position and attitude parameters of the target object can be solved by minimizing the weighted sum of observation error and prediction constraints: error model construction, for example, for N visual devices, the error between the two-dimensional observation coordinates of the target object in the image of each device and the output coordinates of the projection function is calculated to form an observation error; the predicted position and predicted attitude output by the motion trajectory prediction model are combined to constrain the solution space, and the optimal three-dimensional position and attitude parameters that minimize the objective function are solved through nonlinear optimization methods such as gradient descent and LM (Levenberg-Marquardt) algorithm.

[0066] S05, based on the three-dimensional spatial position and posture parameters of the target object, calculate the lighting environment parameters, imaging object distance and motion blur coefficient, and adjust the exposure time, gain coefficient and focal length parameters of each visual device in real time accordingly.

[0067] Among them, the lighting environment parameters: according to the spatial position relationship between the target object and each light source (if the system deploys an auxiliary light source, the distance can be calculated by the three-dimensional coordinates of the light source and the object position), the light intensity is estimated in combination with the inverse square law of distance, or the ambient light change is indirectly deduced through the image grayscale distribution statistics (such as mean square error, brightness mean), and the lighting compensation coefficient is generated; the imaging object distance: the three-dimensional space distance (i.e., object distance) from the target object surface coding area to each visual device lens is calculated through the Euclidean distance formula. Motion blur coefficient: combined with the motion state parameters of step S02 and the current object distance, the motion blur coefficient is calculated through the formula k m =v×t exp Estimate the degree of motion blur, where t exp The default value for the current exposure time. This coefficient is used to quantify the impact of the object's displacement during exposure on image clarity. For exposure time: If the light intensity is high or the motion blur coefficient is large, shorten the exposure time to reduce motion smear; if the light is insufficient, extend the exposure time and combine it with gain adjustment to avoid noise amplification. The exposure time adjustment range follows the device hardware response curve (e.g., 10μs to 10ms); for gain coefficient control: In low-light scenes, increase the gain proportionally (e.g., ISO value), but set an upper threshold (e.g., 400) to control noise; when there is sufficient light, reduce the gain to ensure the optimal image signal-to-noise ratio. For focal length parameter control: call the lens autofocus algorithm (e.g., stepper motor drive) based on the real-time object distance, and calculate the focal length adjustment amount through the triangulation principle to make the code area clearly imaged on the focal plane of the camera sensor.

[0068] S06, using the adjusted visual devices to capture images, detect the code area in the image and extract geometric features, perform geometric correction based on the geometric features, and generate a standardized code image; the code includes common barcode types, and its geometric features include module size, locator shape or row and column arrangement rules.

[0069] Among them, for one-dimensional barcodes (such as Code128) among common barcode types, Hough Transform can be used to detect their parallel line features, and the main direction of the barcode can be located by parameter space accumulation. For two-dimensional barcodes (such as QR codes and PDF417) among common barcode types, template matching and corner point detection (such as Harris corner points) can be used to locate their feature locators (such as the three rectangular positioning boxes of QR codes and the L-shaped pattern of PDF417); for barcodes containing complex backgrounds, Gaussian filtering can be used for denoising, and then threshold segmentation (such as Otsu algorithm) or morphological operations can be used to separate the code area; geometric feature extraction: for one-dimensional barcodes, dimensional features such as module width and bar-to-space ratio are extracted; for two-dimensional barcodes, locator shapes (such as squares and trapezoids), row and column arrangement rules (such as row spacing and column offset of PDF417) and module geometric relationships (such as the relative position vectors of adjacent modules) are extracted. An image distortion model is established based on the intrinsic parameter matrix (focal length, principal point coordinates) obtained from camera calibration and the radial / tangential distortion coefficients. For perspective distortion (such as trapezoidal deformation caused by oblique viewing angles), the perspective transformation matrix is ​​constructed by extracting the deviation between the actual and ideal coordinates of the four corner points of the symbol area. Geometric correction of the symbol area is performed using bilinear or bicubic interpolation algorithms. For one-dimensional codes, the tilt angle is corrected along the scanning direction to ensure that the barcode lines are perpendicular to the scan line. For two-dimensional codes, the irregular quadrilateral symbol area is mapped to a regular rectangle through perspective transformation to ensure module row and column alignment. The corrected symbol area is cropped to a preset size (e.g., 512×512 pixels), and the image grayscale range is uniformly mapped to [0, 255]. Histogram equalization or linear stretching is used to adjust the brightness distribution to eliminate brightness differences between images captured by different devices (e.g., localized overbrightness caused by light reflection on one device). A standardized symbol image is generated.

[0070] S07: Decode and recognize the standardized code image to generate a code reading result.

[0071] For two-dimensional barcodes in standardized code images of various visual devices, lightweight convolutional neural networks (such as ResNet) can be used to extract code features and perform end-to-end decoding. For one-dimensional barcodes, edge detection combined with a bar-space width matching algorithm can be used to locate barcode lines through Hough transform, and decoding can be performed according to the standard character set mapping. The algorithm's automatic switching mechanism is to identify the code type based on code geometry features (such as the shape of the two-dimensional locator and the one-dimensional bar-space ratio) and dynamically call the corresponding decoding module. For example, when three square locating boxes are detected, a dedicated QR code decoding process is initiated. After the corresponding decoding and recognition, the code reading result is generated.

[0072] The above-mentioned industrial code reading control method based on device collaborative optimization obtains image data and three-dimensional coordinates from multiple vision devices and generates a multi-view image sequence with a unified timestamp. It then establishes a motion trajectory prediction model based on the two-dimensional bounding box coordinates of the target object. It then constructs a spatial projection relationship based on the three-dimensional coordinates of the vision devices and fuses the multi-view data to calculate the three-dimensional position and posture parameters of the target object. It then calculates the illumination, object distance, and motion blur coefficient in real time to dynamically adjust the exposure time, gain, and focal length parameters of each device. The adjusted device captures images to detect the code area and performs geometric correction to generate a standardized image. The decoded code reading result is then output. Through multi-device collaborative optimization and dynamic parameter adjustment, the method overcomes the limitations of traditional single-device code reading in terms of dynamic adaptability, view coverage, and collaborative capabilities. It utilizes the redundancy of multi-view data to improve the accuracy of pose calculation, dynamically adjusts imaging parameters to overcome image quality degradation caused by high-speed motion and sudden changes in illumination, and uses geometric correction to eliminate the effects of view distortion. This method improves the decoding success rate and efficiency under complex working conditions, systematically solving the low recognition rate problem of traditional industrial code reading caused by fixed parameters, a single view, and lack of collaboration.

[0073] In one embodiment, after adjusting the exposure time, gain coefficient and focal length parameters of each visual device in real time, the method further includes:

[0074] S11, determining an initial code reading task allocation plan based on the image acquisition performance parameters of each visual device, where the image acquisition performance parameters include resolution, frame rate, and field of view;

[0075] S12, real-time monitoring of the processing status data of each visual device for the assigned task, the processing status data including image processing delay, decoding failure rate and resource utilization rate;

[0076] S13, calculating the task weight adjustment coefficient of each visual device according to the processing status data;

[0077] S14, performing weighted optimization on the initial task allocation plan based on the task weight adjustment coefficient to generate an optimized allocation plan.

[0078] Specifically, through the image acquisition performance parameters of each visual device, such as resolution (determines the minimum code element parsing capability), frame rate (affects the acquisition frequency of high-speed motion scenes), and field of view (determines the effective coverage area), the adaptability of the device to different code areas is calculated, and an initial code reading task allocation plan is generated; real-time monitoring of processing status data such as image processing delay (reflecting computing load), decoding failure rate (characterizing adaptability to current working conditions), and resource occupancy rate (such as CPU / GPU usage rate) of each device is carried out, and a task weight adjustment model is constructed based on the above data. For example, the weight of devices with processing delay exceeding the threshold or high failure rate is reduced; the initial plan is weighted and optimized through weight coefficients, and code recognition tasks are reallocated to devices with high adaptability, forming a dynamically adjusted optimized allocation plan. Through the coordinated analysis of device performance and real-time load, intelligent scheduling of multi-device tasks is achieved, avoiding device overload or resource waste, improving the overall decoding efficiency and reliability of the system under complex working conditions, and solving the recognition delay and resource redundancy problems caused by the lack of collaborative ability when traditional multiple devices work independently.

[0079] In one embodiment, an initial code reading task allocation scheme is determined based on image acquisition performance parameters of each visual device, including:

[0080] S21, calculating the resolvability of the minimum symbol size according to the resolution parameters of each visual device;

[0081] S22, calculating the effective coverage area of ​​each visual device based on the relationship between the field of view angle parameter and the spatial position of the target object;

[0082] S23, generating device allocation parameters by integrating the resolution and the effective coverage area;

[0083] S24, allocating the code area to be identified to the corresponding visual device according to the allocation parameters, and generating an initial task allocation plan.

[0084] For example, the resolution of each visual device for the minimum symbol size is calculated based on its resolution parameter. Specifically, the geometric relationship between the device pixel size and the actual symbol size is used to determine the minimum symbol width that the device can clearly distinguish (for example, when the device resolution is 2048×1536 pixels and the field of view width is 100mm, the pixel resolution is 0.0488mm / px, and the minimum resolvable symbol size is determined to be 2×0.0488mm); based on the relationship between the field of view angle parameter and the spatial position of the target object, the trigonometric function is used to calculate the minimum symbol size of each visual device at the current position. The effective coverage area under different postures is calculated, such as using the field of view angle to calculate the projection coverage of the device on the plane where the target object is located; the resolvability and effective coverage area are normalized and weighted to generate device allocation parameters that comprehensively reflect the device's adaptability to the code area. The resolvability weight is used to characterize the device's ability to distinguish fine code elements, and the effective coverage area weight is used to characterize the device's spatial coverage capability of the code area; based on the allocation parameters, the code area to be identified is divided into several sub-areas and assigned to corresponding highly adaptable visual devices to generate an initial code reading task allocation plan. Through quantitative analysis of device performance parameters, the initial optimized allocation of code recognition tasks is achieved, avoiding the waste of device resources or insufficient resolution capabilities caused by traditional fixed allocation methods, laying the foundation for subsequent dynamic task adjustments.

[0085] In one embodiment, after generating the optimized allocation plan, the method further includes:

[0086] S31, writing the code reading result into the target association database;

[0087] S32, if there are differences in the decoded content between N consecutive code reading results associated with the same target object ID in the target association database (N≥2), a manual re-inspection process is triggered;

[0088] S33, updating the exposure time, gain coefficient, focal length parameter and device allocation parameter of each visual device according to the manual re-inspection result.

[0089] Specifically, after generating the optimized allocation plan, the code reading results are written into the target-associated database according to the target object ID index to achieve structured storage and traceability of the decoded data; if there are differences in the decoding content (such as character misalignment, verification failure, etc.) in the N consecutive code reading results associated with the same target object ID in the database, the manual review process is automatically triggered, and the operator reviews the original image of the code and the decoding process to avoid systematic error transmission; based on the manual review results, the exposure time, gain coefficient, focal length parameters and other imaging parameters of each visual device are reversely updated, and the device allocation parameters (such as task weight coefficient, coverage area division) are adjusted to form a closed-loop optimization mechanism of data storage-anomaly detection-manual calibration-parameter update. By combining database-associated anomaly detection with manual review, the problem of decreased decoding accuracy caused by equipment parameter drift or failure of task allocation strategy during long-term operation is solved, so that the system can continue to maintain optimal performance under complex working conditions. It is especially suitable for industrial traceability scenarios with extremely high requirements for code reading reliability.

[0090] In one embodiment, decoding and identifying a standardized code image to generate a code reading result includes:

[0091] S41, performing decoding recognition on the standardized coded image and calculating the decoding confidence of each visual device;

[0092] S42: If the decoding confidence of any visual device is lower than the preset threshold, generate the code reading result according to the following steps:

[0093] S43, excluding decoding results of visual devices with decoding confidence lower than a preset threshold, and obtaining a preferred decoding result;

[0094] S44, generating a corresponding weighting coefficient according to the confidence value of the visual device of the preferred decoding result;

[0095] S45 , performing weighted aggregation on the preferred decoding results based on the weighting coefficients to generate a code reading result including a credibility score.

[0096] Exemplarily, when performing decoding and recognition on standardized code images and generating code reading results, a convolutional neural network (such as ResNet) or a traditional feature matching algorithm can be used to decode the standardized images of each visual device, and at the same time, the decoding confidence of the corresponding device is calculated based on the probability value or matching index output by the decoding algorithm (such as using the softmax output probability of CNN or the check code pass rate of the traditional algorithm); if the decoding confidence of any device is lower than the preset threshold (0.7), the decoding result of the device is excluded, and the preferred decoding result with high confidence is retained; the confidence value of the device corresponding to the preferred result is normalized to generate a weighting coefficient (confidence 0.9 corresponds to weight 0.6, 0.8 corresponds to weight 0.4); the preferred decoding results are aggregated based on the weighting coefficient, and if the results are consistent, they are directly output; if there are differences, the content with the highest proportion is selected according to the weight, and the weighted average confidence is added as the credibility score to generate the code reading result. Through confidence screening and weighted fusion of multi-view decoding results, decoding errors caused by poor imaging quality of some equipment can be effectively suppressed, the bit error rate under complex working conditions can be reduced, and code reading results with quantitative reliability indicators can be output to meet the stringent requirements of industrial scenarios for data accuracy.

[0097] In one embodiment, the method further comprises:

[0098] S51, obtaining historical decoding time data, device processing performance parameters and code complexity level to form a calculation basis for the preset time window;

[0099] S52, generating an initial time window using a time window algorithm model based on the calculation basis, and adaptively adjusting the time window based on the current device load state to obtain an adjusted time window;

[0100] S53, real-time monitoring to determine whether the code reading results generated within the adjustment time window meet the scoring requirements generated based on the credibility scores of each visual device;

[0101] S54: If the code reading result within the adjustment time window does not meet the scoring requirements, the current code system recognition process is skipped and the current code system is marked as pending re-inspection status.

[0102] Specifically, real-time optimization of the code reading process is achieved by constructing an adaptive time window mechanism: historical decoding time data (such as the time distribution of the past 1,000 decodings), device processing performance parameters (CPU occupancy, memory bandwidth, etc.) and code complexity level (divided according to code element density and error correction level) are obtained to form the calculation basis for the preset time window; based on the above data, an initial time window is generated through a time window algorithm model (such as exponential moving average combined with load prediction), and the window size is dynamically adjusted according to the current device load status (such as the GPU decoding queue length) to obtain an adaptive adjustment time window; real-time monitoring is carried out to determine whether the code reading results generated within the window meet the scoring requirements (threshold 0.7) generated based on the credibility scores of each visual device (such as weighted average confidence); if no results meeting the scoring requirements are obtained within the adjustment time window, the current code recognition process is skipped and marked as pending re-inspection. By collaboratively analyzing historical data and real-time loads, the system dynamically adapts to decoding time constraints in high-speed motion scenarios, avoiding the accumulation of processing delays caused by equipment overload or complex code systems, and improving system response efficiency. At the same time, a pending re-inspection marking mechanism ensures the subsequent tracing of abnormal code systems, balancing the real-time and reliability requirements of industrial code reading.

[0103] In one embodiment, a spatial projection relationship of a target object under different viewing angles is established and observation data from each viewing angle is integrated to obtain the three-dimensional spatial position and posture parameters of the target object, including S61, constructing the following optimization objective function to solve the optimal three-dimensional spatial position:

[0104]

[0105] Among them, N represents the total number of visual devices involved in fusion, P i represents the two-dimensional observation coordinates of the target object in the i-th visual device image, π i (X, Θ) represents the projection function of the i-th visual device, the input is the three-dimensional position X and posture Θ, and the output is the two-dimensional coordinate projected onto the image plane of the device, α i Represents the observation weight coefficient of the i-th visual device, which is dynamically calculated by the image quality index of the device. and represents the predicted position and predicted posture of the target object output by the motion trajectory prediction model, and β and γ represent the regularization coefficients of position and posture, respectively.

[0106] Specifically, when establishing the spatial projection relationship of the target object under different perspectives and fusing the observation data to solve the three-dimensional pose, the weighted fusion and constraint solving of multi-source data are achieved by constructing an optimization objective function: for N visual devices involved in the fusion, the two-dimensional observation coordinates P of the target object in the image of each device are converted into i With the projection function π i(X, Θ) calculation value is used for error modeling, where the projection function π i Based on the device internal parameters (focal length, principal point) and external parameters (rotation matrix R, translation vector T), the three-dimensional position X and posture Θ are mapped to the image plane; the observation weight coefficient α i The image quality index (such as clarity and signal-to-noise ratio) of the device is dynamically calculated. The higher the quality, the greater the weight, so as to highlight the contribution of reliable data. The predicted position output by the motion trajectory prediction model is introduced. and predicted posture As a priori constraints, through the regularization term The divergence of the solution is suppressed, where β and γ are regularization coefficients that balance observation error and prediction constraints. The objective function is to minimize the sum of the weighted sum of observation error and the regularization term, and to find the optimal 3D position X and pose Θ that minimizes the objective function. This achieves redundant fusion of multi-view data and motion trend constraints. By constructing an optimization model that includes data consistency terms and prior constraints, and using a nonlinear optimization algorithm (such as the LM algorithm) for iterative solution, the multi-device observation error is controlled to the sub-pixel level, improving solution accuracy and effectively resolving pose ambiguity under complex viewpoints.

[0107] The above-mentioned industrial code reading control method based on equipment collaborative optimization obtains image data and three-dimensional coordinates of multiple visual devices and generates a multi-view image sequence with a unified time stamp. It establishes a motion trajectory prediction model based on the two-dimensional bounding box coordinates of the target object, constructs a spatial projection relationship based on the three-dimensional coordinates of the visual device and fuses the multi-view data to solve the three-dimensional position and posture parameters of the target object. It solves the illumination, object distance and motion blur coefficient in real time to dynamically adjust the exposure time, gain and focal length parameters of each device. The adjusted device is used to collect images to detect the code area and perform geometric correction to generate a standardized image. It combines the dynamic optimization of the initial task allocation plan, the confidence weighted aggregation of the code reading results and the adaptive time window. The mouth mechanism forms a closed loop between manual re-inspection and parameter update when the code reading results are abnormal, breaking through the limitations of traditional single-device code reading in dynamic adaptability, viewing angle coverage and collaborative capabilities, using multi-view data redundancy to improve the accuracy of pose solution, overcoming the image quality deterioration caused by high-speed motion and sudden changes in illumination by dynamically adjusting imaging parameters, eliminating the influence of viewing angle deformation with the help of geometric correction, and avoiding waste of equipment resources in combination with dynamic allocation of task loads. Errors are suppressed through confidence screening and weighted fusion, and the time window mechanism is used to balance real-time and reliability, thereby improving the decoding success rate and efficiency under complex working conditions, and systematically solving the problem of low recognition rate of traditional industrial code reading due to fixed parameters, single viewing angle and lack of collaboration.

[0108] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0109] Based on the same inventive concept, embodiments of the present application also provide a high-speed response system for industrial code reading control based on device collaborative optimization, for implementing the aforementioned method for industrial code reading control based on device collaborative optimization. The solution provided by this system is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the high-speed response system for industrial code reading control based on device collaborative optimization provided below can be found in the above-mentioned limitations of the method for industrial code reading control based on device collaborative optimization, and will not be repeated here.

[0110] In an exemplary embodiment, Figure 2 As shown, an industrial code reading control high-speed response system based on equipment collaborative optimization is provided, including:

[0111] The multi-source data synchronization module 101 is used to obtain image data and three-dimensional position coordinates of multiple visual devices and generate a multi-view image sequence with a unified time stamp through clock calibration;

[0112] The motion trajectory modeling module 102 is used to extract the two-dimensional bounding box coordinates of the target object in the multi-view image sequence and calculate its motion state parameters to establish a motion trajectory prediction model;

[0113] The spatial projection relationship building module 103 is used to combine the motion trajectory prediction model with the three-dimensional position coordinates of the visual device to establish the spatial projection relationship of the target object under different viewing angles;

[0114] The 3D position and posture fusion calculation module 104 is used to fuse the observation data from each perspective based on the spatial projection relationship to obtain the 3D spatial position and posture parameters of the target object;

[0115] The device parameter dynamic control module 105 is used to calculate the lighting environment parameters, imaging object distance and motion blur coefficient based on the three-dimensional spatial position and posture parameters of the target object, and adjust the exposure time, gain coefficient and focal length parameters of each visual device in real time accordingly;

[0116] The code standardization processing module 106 is used to use the adjusted visual devices to capture images, detect code areas in the images and extract geometric features, perform geometric correction based on the geometric features, and generate a standardized code image;

[0117] The decoding and recognition output module 107 is used to decode and recognize the standardized code image and generate a code reading result.

[0118] In one embodiment, the system further includes a task allocation plan generation module, which is configured to:

[0119] Determine the initial code reading task allocation plan based on the image acquisition performance parameters of each visual device. Image acquisition performance parameters include resolution, frame rate, and field of view.

[0120] Real-time monitoring of the processing status data of each visual device for the assigned task, including image processing delay, decoding failure rate and resource utilization;

[0121] Calculate the task weight adjustment coefficient of each visual device based on the processing status data;

[0122] The initial task allocation plan is weighted optimized based on the task weight adjustment coefficient to generate an optimized allocation plan.

[0123] In one embodiment, the task allocation solution generation module is further configured to:

[0124] Calculate the resolvability of the minimum symbol size based on the resolution parameters of each visual device;

[0125] Calculate the effective coverage area of ​​each visual device based on the relationship between the field of view angle parameters and the spatial position of the target object;

[0126] Fusing resolution and effective coverage area to generate device allocation parameters;

[0127] The code area to be identified is allocated to the corresponding visual device according to the allocation parameters, and the initial task allocation plan is generated.

[0128] In one embodiment, Figure 3 As shown, the system further includes a detection optimization module 108, which is used to:

[0129] Write the code reading results into the target associated database;

[0130] If there are differences in the decoded content between N consecutive code reading results associated with the same target object ID in the target association database (N≥2), the manual re-inspection process will be triggered;

[0131] Update the exposure time, gain factor, focal length parameters and device allocation parameters of each visual device based on the manual re-inspection results.

[0132] In one embodiment, the decoding recognition output module 107 is further configured to:

[0133] Perform decoding recognition on the standardized code image and calculate the decoding confidence of each visual device;

[0134] If the decoding confidence of any vision device is lower than the preset threshold, the code reading result is generated according to the following steps:

[0135] Eliminating decoding results of visual devices whose decoding confidence is lower than a preset threshold to obtain a preferred decoding result;

[0136] Generate a corresponding weighting coefficient according to the confidence value of the visual device of the preferred decoding result;

[0137] The preferred decoding results are weighted and aggregated based on the weighting coefficients to generate a code reading result including a confidence score.

[0138] In one embodiment, Figure 3 As shown, the system further includes a real-time response control module 109, which is used to:

[0139] Obtain historical decoding time data, equipment processing performance parameters and code complexity level to form the calculation basis for the preset time window;

[0140] Based on the calculation basis, the initial time window is generated through the time window algorithm model, and adaptively adjusted based on the current equipment load status to obtain the adjusted time window;

[0141] Real-time monitoring to determine whether the code reading results generated within the adjustment time window meet the scoring requirements generated based on the credibility scores of each visual device;

[0142] If the code reading result within the adjustment time window does not meet the scoring requirements, the current code system recognition process is skipped and the current code system is marked as pending re-inspection.

[0143] In one embodiment, the 3D pose fusion solution module 104 is further configured to:

[0144] The following optimization objective function is constructed to solve the optimal three-dimensional spatial position:

[0145]

[0146] Among them, N represents the total number of visual devices involved in fusion, P i represents the two-dimensional observation coordinates of the target object in the i-th visual device image, π i(X, Θ) represents the projection function of the i-th visual device, the input is the three-dimensional position X and posture Θ, and the output is the two-dimensional coordinate projected onto the image plane of the device, α i Represents the observation weight coefficient of the i-th visual device, which is dynamically calculated by the image quality index of the device. and represents the predicted position and predicted posture of the target object output by the motion trajectory prediction model, and β and γ represent the regularization coefficients of position and posture, respectively.

[0147] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the industrial code reading control method based on device collaborative optimization as described above are implemented.

[0148] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0149] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0150] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. An industrial code reading control method based on equipment collaborative optimization, characterized in that: The method comprises: Acquire image data and 3D position coordinates of multiple visual devices, and generate multi-view image sequences with unified timestamps through clock calibration; Extracting the two-dimensional bounding box coordinates of the target object in the multi-view image sequence, and calculating its motion state parameters to establish a motion trajectory prediction model; Combining the motion trajectory prediction model with the three-dimensional position coordinates of the visual device to establish the spatial projection relationship of the target object under different viewing angles; Based on the spatial projection relationship, the observation data from each perspective are fused to obtain the three-dimensional spatial position and posture parameters of the target object; Based on the three-dimensional spatial position and posture parameters of the target object, the lighting environment parameters, imaging object distance and motion blur coefficient are calculated, and the exposure time, gain coefficient and focal length parameters of each visual device are adjusted in real time accordingly; Using the adjusted visual devices to capture images, detecting the code area in the image and extracting geometric features, performing geometric correction based on the geometric features, and generating a standardized code image; the code includes common barcode types, whose geometric features include module size, locator shape, or row and column arrangement rules; The standardized code system image is decoded and recognized to generate a code reading result.

2. The method according to claim 1, characterized in that After adjusting the exposure time, gain coefficient and focal length parameters of each visual device in real time, the method further includes: Determine the initial code reading task allocation plan based on the image acquisition performance parameters of each visual device, wherein the image acquisition performance parameters include resolution, frame rate, and field of view; Real-time monitoring of the processing status data of each visual device for the assigned task, including image processing delay, decoding failure rate and resource utilization rate; Calculating a task weight adjustment coefficient for each visual device based on the processing status data; The initial task allocation plan is weighted optimized based on the task weight adjustment coefficient to generate an optimized allocation plan.

3. The method according to claim 2, characterized in that The initial code reading task allocation scheme is determined based on the image acquisition performance parameters of each visual device, including: Calculating the resolvability of the minimum symbol size according to the resolution parameters of each of the visual devices; Calculating the effective coverage area of ​​each visual device according to the relationship between the field of view angle parameter and the spatial position of the target object; generating device allocation parameters by fusing the resolvability and the effective coverage area; The code area to be identified is allocated to the corresponding visual device according to the allocation parameters to generate an initial task allocation plan.

4. The method according to claim 2, characterized in that After generating the optimized allocation plan, the method further includes: Writing the code reading result into a target associated database; If there are differences in the decoded content of N consecutive code reading results associated with the same target object ID in the target association database (N≥2), a manual re-inspection process is triggered; The exposure time, gain coefficient, focal length parameter and device allocation parameter of each of the visual devices are updated according to the manual re-inspection results.

5. The method according to claim 1, wherein The step of decoding and identifying the standardized code image to generate a code reading result includes: Performing decoding recognition on the standardized code image and calculating decoding confidence of each visual device; If the decoding confidence of any of the above-mentioned visual devices is lower than the preset threshold, the code reading result is generated according to the following steps: Eliminating decoding results of visual devices whose decoding confidence is lower than a preset threshold to obtain a preferred decoding result; generating a corresponding weighting coefficient according to a confidence value of the visual device of the preferred decoding result; The preferred decoding results are weighted and aggregated based on the weighting coefficients to generate a code reading result including a confidence score.

6. The method according to claim 5, characterized in that The method further comprises: Obtain historical decoding time data, equipment processing performance parameters and code complexity level to form the calculation basis for the preset time window; According to the calculation basis, an initial time window is generated through a time window algorithm model, and adaptively adjusted based on the current device load state to obtain an adjusted time window; Real-time monitoring of whether the code reading results generated within the adjustment time window meet the scoring requirements generated based on the credibility scores of the visual devices; If the code reading result within the adjustment time window does not meet the scoring requirement, the current code system recognition process is skipped and the current code system is marked as pending re-inspection status.

7. The method according to claim 1, characterized in that The establishment of the spatial projection relationship of the target object under different viewing angles and the fusion of the observation data from each viewing angle to obtain the three-dimensional spatial position and posture parameters of the target object include constructing the following optimization objective function to solve the optimal three-dimensional spatial position: Among them, N represents the total number of visual devices involved in fusion, P i represents the two-dimensional observation coordinates of the target object in the i-th visual device image, π i (X, Θ) represents the projection function of the i-th visual device, the input is the three-dimensional position X and posture Θ, and the output is the two-dimensional coordinate projected onto the image plane of the device, α i Represents the observation weight coefficient of the i-th visual device, which is dynamically calculated by the image quality index of the device. and represents the predicted position and predicted posture of the target object output by the motion trajectory prediction model, and β and γ represent the regularization coefficients of position and posture, respectively.

8. An industrial code reading control high-speed response system based on equipment collaborative optimization, characterized in that: The system comprises: Multi-source data synchronization module, used to obtain image data and 3D position coordinates of multiple visual devices, and generate multi-view image sequences with unified timestamps through clock calibration; A motion trajectory modeling module is used to extract the two-dimensional bounding box coordinates of the target object in the multi-view image sequence and calculate its motion state parameters to establish a motion trajectory prediction model; A spatial projection relationship building module is used to combine the motion trajectory prediction model with the three-dimensional position coordinates of the visual device to establish the spatial projection relationship of the target object under different viewing angles; A three-dimensional position and posture fusion solution module is used to fuse the observation data of each perspective based on the spatial projection relationship to obtain the three-dimensional spatial position and posture parameters of the target object; A device parameter dynamic control module is used to calculate the lighting environment parameters, imaging object distance and motion blur coefficient based on the three-dimensional spatial position and posture parameters of the target object, and adjust the exposure time, gain coefficient and focal length parameters of each visual device in real time accordingly; a code system standardization processing module, configured to use the adjusted visual devices to capture images, detect code system areas in the images and extract geometric features, perform geometric correction based on the geometric features, and generate a standardized code system image; The decoding and recognition output module is used to decode and recognize the standardized code image and generate a code reading result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images