A method and system for visual acquisition and processing of large-format fabric top-view images
By employing multi-camera collaborative control and image fusion technology, the accuracy and synchronization issues of large-format fabric image acquisition were resolved, enabling high-precision cutting and positioning, and improving garment quality and operational efficiency.
Patent Information
- Application Number
- CN202610944440.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-29
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies suffer from insufficient accuracy, difficulty in multi-camera coordination and synchronization, and inaccurate consistency between visual and cutting coordinates when acquiring large-format fabric images, resulting in low cutting positioning accuracy and affecting garment quality.
Multiple top-mounted cameras work together to establish a transformation matrix between pixel coordinates and physical coordinates by calibrating feature points, enabling synchronous image acquisition and geometric correction. Combined with image fusion processing under a unified coordinate system, a seamless stitched image is generated and converted into layout data under a cropping coordinate system, driving the cropping equipment to perform automated cropping.
It effectively expands the imaging field of view, corrects lens distortion and installation tilt errors, improves stripe alignment accuracy and garment quality, and enhances equipment adaptability and operational efficiency.
Smart Images

Figure CN123123744A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent cutting equipment technology, and in particular to a method and system for visual acquisition and processing of large-format fabric top-mounted images. Background Technology
[0002] In the field of garment cutting, stripe and check alignment is a crucial step affecting the quality of finished garments. When cutting fabrics with striped or checkered patterns, it is essential to ensure that the stripes and checks on each piece are continuously aligned after sewing; otherwise, the appearance of the finished garment will be severely affected. Currently, the industry typically uses visual inspection solutions to solve the stripe and check alignment problem: images of the fabric are captured by a camera, image processing algorithms are used to identify the stripes and checks on the fabric, and the cutting bed is controlled to complete the cutting based on the identification results.
[0003] However, existing solutions mostly rely on a single camera for imaging, while the cutting table area is usually much larger than the imaging field of view of a single camera, meaning a single shot cannot cover the entire cutting area. To obtain complete fabric image information, it is often necessary to move the camera or the fabric multiple times, reducing work efficiency, and multiple imaging introduces cumulative errors from image stitching. Furthermore, a certain tilt angle is unavoidable during camera installation, and combined with the distortion characteristics of the lens itself and the principle of perspective imaging, the acquired images are prone to geometric distortion. This distortion directly affects the accuracy of stripe recognition, thereby reducing the accuracy of cutting positioning. Summary of the Invention
[0004] In view of the above, the present invention addresses the shortcomings of the prior art by providing a method and system for large-format fabric top-view visual acquisition and processing, aiming to solve the problems of the prior art not meeting expectations in terms of large-format fabric image acquisition accuracy, multi-camera collaborative synchronization, and visual and cutting coordinate unification.
[0005] To address the aforementioned problems, in a first aspect, the present invention provides a method for visual acquisition and processing of large-format fabric images from above, comprising the following steps:
[0006] S1. Control at least two top-mounted cameras to acquire calibration images, extract image feature points and establish a mapping relationship with the corresponding physical reference points, and generate a transformation matrix between the pixel coordinates of each camera and the physical coordinates of the cutting bed;
[0007] S2. Send a synchronization trigger command to multiple cameras so that each camera completes image acquisition at the same time and obtains multiple channels of raw image data;
[0008] S3. Using the transformation matrix corresponding to each camera, perform geometric correction on each channel of original image data respectively;
[0009] S4. Transform the multi-path corrected images to a unified coordinate system and perform fusion processing in the overlapping areas of adjacent camera fields of view to form a complete stitched image covering the large-format cutting bed;
[0010] S5. Convert the coordinate data of the stitched image into layout data in the cutting bed coordinate system and send it to the cutting control unit to drive the cutting equipment to perform automated cutting operations.
[0011] In one possible implementation, step S1 includes: S101. Laying a calibration feature plate with known physical coordinates on the cutting table, the calibration feature plate covering the complete field of view of a single camera; S102. Controlling each top-mounted camera to sequentially acquire calibration feature images through an image acquisition interface; S103. Extracting the center pixel coordinates of feature points in the image and establishing a one-to-one mapping relationship with the corresponding physical reference point coordinates; S104. Dividing the image into several local regions, each local region containing at least four physical reference points; S105. For each local region, based on the physical coordinates of the physical reference points in that region and the pixel coordinates of the corresponding image feature points in the image, using a perspective transformation model to calculate the transformation matrix between pixel coordinates and physical coordinates region by region.
[0012] In one possible implementation, step S2 includes: S201. Creating an independent acquisition thread for each camera, and each thread entering a synchronous waiting state after completing the camera parameter configuration; S202. After the main thread detects that all acquisition threads are ready, it broadcasts a unified trigger signal; S203. After receiving the trigger signal, each acquisition thread performs imaging acquisition within a preset time window, so that the imaging time difference between each camera is less than a preset threshold.
[0013] In one possible implementation, step S4 includes: S401. Mapping the pixel coordinates of the multi-path corrected images to the global coordinate system of the cutting bed based on the transformation matrix of each camera; S402. Determining the overlapping area between the fields of view of adjacent cameras; S403. Calculating the fusion weight according to the pixel position within the overlapping area, and performing weighted fusion on the corresponding pixels of adjacent cameras according to the fusion weight to generate a seamless stitched image.
[0014] In one possible implementation, step S5 includes: S501. Converting image coordinates into physical coordinates of the cutting bed based on the offset between the spliced image and the origin of the cutting bed; S502. Generating cutting trajectory pattern data in the cutting bed coordinate system based on the fabric stripe recognition result; S503. Sending the pattern data to the cutting control software through a communication interface to drive the cutting blade to perform cutting along the corrected trajectory.
[0015] Secondly, the present invention also provides a large-format fabric top-view visual acquisition and processing system, comprising the following modules:
[0016] The calibration module is used to control at least two top-mounted cameras to acquire calibration images, extract image feature points and establish a mapping relationship with the corresponding physical reference points, and generate a transformation matrix between the pixel coordinates of each camera and the physical coordinates of the cutting bed.
[0017] The synchronization control module is used to send synchronization trigger commands to multiple cameras, so that each camera completes image acquisition at the same time and obtains multiple channels of raw image data.
[0018] The correction module is used to perform geometric correction on each channel of original image data using the transformation matrix corresponding to each camera.
[0019] The stitching module is used to convert the multi-channel corrected images to a unified coordinate system and perform fusion processing in the overlapping areas of adjacent camera fields of view to form a complete stitched image covering a large-format cutting bed.
[0020] The output module is used to convert the coordinate data of the stitched image into layout data in the cutting bed coordinate system and send it to the cutting control unit to drive the cutting equipment to perform automated cutting operations.
[0021] In one possible implementation, the calibration module includes: laying a calibration feature plate with known physical coordinates on the cutting table surface, the calibration feature plate covering the complete field of view of a single camera; controlling each top-mounted camera to sequentially acquire calibration feature images via an image acquisition interface; extracting the center pixel coordinates of feature points in the images and establishing a one-to-one mapping relationship with the corresponding physical reference point coordinates; dividing the image into several local regions, each local region containing at least four physical reference points; for each local region, based on the physical coordinates of the physical reference points in that region and the pixel coordinates of the corresponding image feature points in the image, using a perspective transformation model to calculate the transformation matrix between pixel coordinates and physical coordinates region by region.
[0022] In one possible implementation, the synchronization control module includes: creating an independent acquisition thread for each camera; each thread entering a synchronization waiting state after completing camera parameter configuration; the main thread broadcasting a unified trigger signal after detecting that all acquisition threads are ready; and each acquisition thread performing imaging acquisition within a preset time window after receiving the trigger signal, so that the imaging time difference between each camera is less than a preset threshold.
[0023] In one possible implementation, the stitching module includes: mapping the pixel coordinates of the multi-path corrected images to the global coordinate system of the cutting bed based on the transformation matrix of each camera; determining the overlapping area between the fields of view of adjacent cameras; calculating the fusion weight according to the pixel position within the overlapping area, and performing weighted fusion on the corresponding pixels of adjacent cameras according to the fusion weight to generate a seamless stitched image.
[0024] In one possible implementation, the output module includes: converting image coordinates into physical coordinates of the cutting bed based on the offset between the spliced image and the origin of the cutting bed; generating cutting trajectory pattern data in the cutting bed coordinate system based on the fabric stripe recognition result; and sending the pattern data to the cutting control software through a communication interface to drive the cutting blade to perform cutting along the corrected trajectory.
[0025] As described above, the large-format fabric top-view visual acquisition and processing method and system provided by the present invention have the following beneficial effects:
[0026] By employing multi-camera collaborative control and integrating multiple visual images, the imaging field of view is effectively expanded, covering large-format cutting areas and meeting the operational needs of various cutting beds. Region-by-region transformation matrix calibration effectively corrects lens distortion, installation tilt, and perspective errors, achieving sub-pixel positioning accuracy for a single camera. A synchronous triggering mechanism ensures the temporal consistency of multi-camera acquisition, and combined with image mapping based on a unified physical coordinate system and overlapping area fusion processing, precise stitching of multiple images is achieved, eliminating misalignment across camera fields of view and stitching seams. The entire process from calibration to stitching output is automated, significantly reducing the difficulty of manual debugging and improving equipment adaptability and operational efficiency. Attached Figure Description
[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0028] Figure 1 This is a flowchart illustrating a method for visual acquisition and processing of large-format fabric top-view images, provided in an embodiment of the present invention. Detailed Implementation
[0029] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which constitute a part of the present invention and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.
[0030] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0031] A specific embodiment of the present invention, such as Figure 1 As shown, this invention discloses a method for visual acquisition and processing of large-format fabric images from top-down photography, comprising the following steps:
[0032] S1. Control at least two top-mounted cameras to acquire calibration images, extract image feature points and establish a mapping relationship with the corresponding physical reference points, and generate a transformation matrix between the pixel coordinates of each camera and the physical coordinates of the cutting bed;
[0033] S2. Send a synchronization trigger command to multiple cameras so that each camera completes image acquisition at the same time and obtains multiple channels of raw image data;
[0034] S3. Using the transformation matrix corresponding to each camera, perform geometric correction on each channel of original image data respectively;
[0035] S4. Transform the multi-path corrected images to a unified coordinate system and perform fusion processing in the overlapping areas of adjacent camera fields of view to form a complete stitched image covering the large-format cutting bed;
[0036] S5. Convert the coordinate data of the stitched image into layout data in the cutting bed coordinate system and send it to the cutting control unit to drive the cutting equipment to perform automated cutting operations.
[0037] In this embodiment of the invention, step S1 includes: S101. Laying a calibration feature plate with known physical coordinates on the cutting table surface, the calibration feature plate covering the complete field of view of a single camera. Using the cutting table's own motion mechanism and marking tools, drawing an array of cross-shaped grid points on the cutting table surface as physical reference points. The spacing between the cross-shaped grid points can be preset according to the field of view of each camera, preferably 50mm to 200mm; the number of array points is determined according to the cutting table width, ensuring that the grid array covers the entire field of view of each camera.
[0038] S102. Control each top-mounted camera to sequentially acquire calibration feature images via the image acquisition interface. The GrabOneImageImpl image acquisition interface controls each top-mounted camera to acquire a calibration image containing the crosshair grid array, with each camera obtaining a complete calibration image. During the acquisition process, ensure that the position, focal length, aperture, and other parameters of each camera remain constant to guarantee the validity of the calibration parameters.
[0039] S103. Extract the center pixel coordinates of feature points in the image and establish a one-to-one mapping relationship with the corresponding physical reference point coordinates. Using image processing algorithms, such as template matching, corner detection, and centroid method, identify the image feature points corresponding to each crosshair grid point in the calibration images of each camera, and obtain the pixel coordinates of each feature point in the image coordinate system.
[0040] S104. Divide the image into several local regions, each containing at least four physical reference points.
[0041] S105. For each local region, based on the physical coordinates of the physical reference point in that region and the pixel coordinates of the corresponding image feature point in the image, the transformation matrix between pixel coordinates and physical coordinates is calculated region by region using a perspective transformation model.
[0042] In this embodiment of the invention, step S2 includes: S201. Creating an independent acquisition thread for each camera, and each thread entering a synchronous waiting state after completing the camera parameter configuration; S202. After the main thread detects that all acquisition threads are ready, it broadcasts a unified trigger signal; S203. After receiving the trigger signal, each acquisition thread performs imaging acquisition within a preset time window, so that the imaging time difference between each camera is less than a preset threshold.
[0043] In this embodiment of the invention, step S4 includes: S401. Based on the transformation matrix of each camera, uniformly map the pixel coordinates of the multi-path corrected image to the global coordinate system of the cutting bed; S402. Determine the overlapping area between the fields of view of adjacent cameras; S403. Within the overlapping area, calculate the fusion weight according to the pixel position, and perform weighted fusion on the corresponding pixels of adjacent cameras according to the fusion weight to generate a seamless stitched image.
[0044] In this embodiment of the invention, step S5 includes: S501. Converting image coordinates into physical coordinates of the cutting bed based on the offset between the spliced image and the origin of the cutting bed; S502. Generating cutting trajectory pattern data in the cutting bed coordinate system based on the fabric stripe recognition result; S503. Sending the pattern data to the cutting control software through a communication interface to drive the cutting blade to perform cutting along the corrected trajectory.
[0045] In another specific embodiment of the present invention, the present invention also provides a large-format fabric top-view visual acquisition and processing system, comprising the following modules:
[0046] The calibration module is used to control at least two top-mounted cameras to acquire calibration images, extract image feature points and establish a mapping relationship with the corresponding physical reference points, and generate a transformation matrix between the pixel coordinates of each camera and the physical coordinates of the cutting bed.
[0047] The synchronization control module is used to send synchronization trigger commands to multiple cameras, so that each camera completes image acquisition at the same time and obtains multiple channels of raw image data.
[0048] The correction module is used to perform geometric correction on each channel of original image data using the transformation matrix corresponding to each camera.
[0049] The stitching module is used to convert the multi-channel corrected images to a unified coordinate system and perform fusion processing in the overlapping areas of adjacent camera fields of view to form a complete stitched image covering a large-format cutting bed.
[0050] The output module is used to convert the coordinate data of the stitched image into layout data in the cutting bed coordinate system and send it to the cutting control unit to drive the cutting equipment to perform automated cutting operations.
[0051] In this embodiment of the invention, the calibration module includes: laying a calibration feature plate with known physical coordinates on the cutting table surface, the calibration feature plate covering the complete field of view of a single camera; controlling each top-mounted camera to sequentially acquire calibration feature images through an image acquisition interface; extracting the center pixel coordinates of feature points in the images and establishing a one-to-one mapping relationship with the corresponding physical reference point coordinates; dividing the image into several local regions, each local region containing at least four physical reference points; for each local region, based on the physical coordinates of the physical reference points in that region and the pixel coordinates of the corresponding image feature points in the image, using a perspective transformation model to calculate the transformation matrix between pixel coordinates and physical coordinates region by region.
[0052] In this embodiment of the invention, the synchronization control module includes: creating an independent acquisition thread for each camera; each thread entering a synchronization waiting state after completing camera parameter configuration; the main thread broadcasting a unified trigger signal after detecting that all acquisition threads are ready; and each acquisition thread performing imaging acquisition within a preset time window after receiving the trigger signal, so that the imaging time difference between each camera is less than a preset threshold.
[0053] In this embodiment of the invention, the stitching module includes: mapping the pixel coordinates of the multi-path corrected images to the global coordinate system of the cutting bed based on the transformation matrix of each camera; determining the overlapping area between the fields of view of adjacent cameras; calculating the fusion weight according to the pixel position within the overlapping area, and performing weighted fusion on the corresponding pixels of adjacent cameras according to the fusion weight to generate a seamless stitched image.
[0054] In this embodiment of the invention, the output module includes: converting image coordinates into physical coordinates of the cutting bed based on the offset between the spliced image and the origin of the cutting bed; generating cutting trajectory pattern data in the cutting bed coordinate system based on the fabric stripe recognition result; and sending the pattern data to the cutting control software through a communication interface to drive the cutting blade to perform cutting along the corrected trajectory.
[0055] In summary, this invention solves the problems of insufficient field of view, geometric distortion of images, and misalignment across cameras in large-format cutting beds by combining region-by-region coordinate transformation matrix calibration with multi-camera synchronous control, thus significantly improving the alignment accuracy of stripes and the quality of finished garments.
[0056] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.
Claims
1. A large-format cloth top shot visual acquisition and processing method, characterized in that, Includes the following steps: S1. Control at least two top-mounted cameras to acquire calibration images, extract image feature points and establish a mapping relationship with the corresponding physical reference points, and generate a transformation matrix between the pixel coordinates of each camera and the physical coordinates of the cutting bed; S2. Send a synchronization trigger command to multiple cameras so that each camera completes image acquisition at the same time and obtains multiple channels of raw image data; S3. Using the transformation matrix corresponding to each camera, perform geometric correction on each channel of original image data respectively; S4. Transform the multi-path corrected images to a unified coordinate system and perform fusion processing in the overlapping areas of adjacent camera fields of view to form a complete stitched image covering the large-format cutting bed; S5. Convert the coordinate data of the stitched image into layout data in the cutting bed coordinate system and send it to the cutting control unit to drive the cutting equipment to perform automated cutting operations.
2. The method of claim 1, wherein, Step S1 includes: S101. A calibration feature plate with known physical coordinates is laid on the cutting table surface, the calibration feature plate covering the complete field of view of a single camera; S102. Control each top-mounted camera to sequentially acquire calibration feature images via the image acquisition interface; S103. Extract the center pixel coordinates of feature points in the image and establish a one-to-one mapping relationship with the corresponding physical reference point coordinates; S104. Divide the image into several local regions, each local region containing at least four physical reference points; S105. For each local region, based on the physical coordinates of the physical reference point in that region and the pixel coordinates of the corresponding image feature point in the image, the transformation matrix between pixel coordinates and physical coordinates is calculated region by region using a perspective transformation model.
3. The method of claim 1, wherein, Step S2 includes: S201. Create an independent acquisition thread for each camera. After each thread completes the camera parameter configuration, it enters a synchronization waiting state. S202. After the main thread detects that all acquisition threads are ready, it broadcasts a unified trigger signal; S203. After receiving the trigger signal, each acquisition thread performs imaging acquisition within a preset time window, so that the imaging time difference between each camera is less than a preset threshold.
4. The method of claim 1, wherein, Step S4 includes: S401. Based on the transformation matrix of each camera, the pixel coordinates of the multi-channel rectified images are uniformly mapped to the global coordinate system of the cutting bed; S402. Determine the overlapping area between the fields of view of adjacent cameras; S403. Within the overlapping area, calculate the fusion weight based on the pixel position, and perform weighted fusion on the corresponding pixels of adjacent cameras according to the fusion weight to generate a seamless stitched image.
5. The method of claim 1, wherein, Step S5 includes: S501. Based on the offset between the stitched image and the origin of the cutting bed, convert the image coordinates into physical coordinates of the cutting bed; S502. Based on the fabric stripe recognition results, generate cutting trajectory layout data in the cutting table coordinate system; S503. The layout data is sent to the cutting control software through the communication interface, and the cutting blade is driven to perform cutting along the corrected trajectory.
6. A large-format cloth top shot visual acquisition and processing system, characterized by, Includes the following modules: The calibration module is used to control at least two top-mounted cameras to acquire calibration images, extract image feature points and establish a mapping relationship with the corresponding physical reference points, and generate a transformation matrix between the pixel coordinates of each camera and the physical coordinates of the cutting bed. The synchronization control module is used to send synchronization trigger commands to multiple cameras, so that each camera completes image acquisition at the same time and obtains multiple channels of raw image data. The correction module is used to perform geometric correction on each channel of original image data using the transformation matrix corresponding to each camera. The stitching module is used to convert the multi-channel corrected images to a unified coordinate system and perform fusion processing in the overlapping areas of adjacent camera fields of view to form a complete stitched image covering a large-format cutting bed. The output module is used to convert the coordinate data of the stitched image into layout data in the cutting bed coordinate system and send it to the cutting control unit to drive the cutting equipment to perform automated cutting operations.
7. The system of claim 6, wherein, The calibration module includes: A calibration feature plate with known physical coordinates is laid on the cutting table surface, the calibration feature plate covering the complete field of view of a single camera; The image acquisition interface controls each top-mounted camera to sequentially acquire calibration feature images; Extract the center pixel coordinates of feature points in the image and establish a one-to-one mapping relationship with the corresponding physical reference point coordinates; The image is divided into several local regions, each containing at least four physical reference points. For each local region, based on the physical coordinates of the physical reference points in that region and the pixel coordinates of the corresponding image feature points in the image, a perspective transformation model is used to calculate the transformation matrix between pixel coordinates and physical coordinates region by region.
8. The system of claim 6, wherein, The synchronization control module includes: An independent acquisition thread is created for each camera, and each thread enters a synchronous waiting state after completing the camera parameter configuration; After the main thread detects that all data acquisition threads are ready, it broadcasts a unified trigger signal. After receiving the trigger signal, each acquisition thread performs imaging acquisition within a preset time window, ensuring that the imaging time difference between each camera is less than a preset threshold.
9. The system of claim 6, wherein, The splicing module includes: Based on the transformation matrix of each camera, the pixel coordinates of the multi-channel rectified images are uniformly mapped to the global coordinate system of the cutting bed; Determine the overlapping area between the fields of view of adjacent cameras; Within the overlapping area, a fusion weight is calculated based on the pixel position, and the corresponding pixels of adjacent cameras are weighted and fused according to the fusion weight to generate a seamless stitched image.
10. The system of claim 6, wherein, The output module includes: Based on the offset between the stitched image and the origin of the cutting bed, the image coordinates are converted into the physical coordinates of the cutting bed; Based on the fabric stripe recognition results, cutting trajectory pattern data is generated in the cutting table coordinate system; The layout data is sent to the cutting control software via a communication interface, driving the cutting blade to perform cutting along the corrected trajectory.
