Dyeing cloth cover color difference detecting and sorting equipment based on machine vision and method of dyeing cloth cover color difference detecting and sorting equipment

Abstract

The invention relates to dyed cloth cover color difference detecting and sorting equipment based on machine vision and a method thereof, and belongs to the technical field of textile printing and dyeing quality detection.The dyed cloth cover color difference detecting and sorting equipment comprises a flattening traction module, an online detection module, a sorting execution module and a central control module; the online detection module comprises a hyperspectral imaging unit and a color area array imaging unit, the hyperspectral imaging unit is used for collecting cloth surface reflection spectrum information, and the color area array imaging unit is used for collecting cloth surface color texture images. The high-precision and high-stability color difference analysis is achieved based on hyperspectrum, the measurement precision which is advanced in the industry and can be traced to the spectrum reference is provided, the standard light source illumination unit which is integrated is matched, the consistency of detection conditions and the stability of results at different times and different batches are ensured, and data authority and reliability are achieved.

Description

Technical Field

[0001] This invention relates to the field of textile printing and dyeing quality inspection technology, and in particular to a machine vision-based color difference detection and sorting device and method for dyed fabrics. Background Technology

[0002] In the textile printing and dyeing industry, color difference and surface quality of dyed fabrics are core indicators for measuring product grade. Currently, the mainstream testing methods in the industry have the following shortcomings:

[0003] First, relying on manual visual inspection, where quality inspectors compare standard samples under a standard light source box, is highly subjective and greatly affected by personnel experience and fatigue; it is inefficient and cannot meet the needs of continuous large-scale production; moreover, it cannot provide objective and quantitative quality data, which is not conducive to quality traceability and process optimization.

[0004] Secondly, for automated detection using a single sensor, one approach is to use spot or portable colorimeters for random sampling, which cannot achieve full-area, continuous detection and carries the risk of missed detections. Another approach is to use a vision system based on RGB color industrial cameras, such as the existing method for sorting textile yarn tubes based on machine vision (publication number CN201410612225A). This method uses camera images to identify colors to control sorting. However, the texture, luster, and subtle changes in ambient light of textiles can significantly affect the stability of RGB color values, resulting in insufficient accuracy and poor repeatability of color difference measurement. In particular, it is prone to misjudgment in sensitive color areas such as dark colors and highly saturated colors. In addition, such solutions usually only focus on color differences and are difficult to detect complex appearance and texture defects such as stains, weave defects, and streaks with high precision at the same time.

[0005] Third, although advanced quality inspection systems have introduced deep learning models, their color difference detection modules are mostly still based on improved RGB imaging technology, which is essentially still "image-based color comparison". In terms of absolute accuracy and anti-interference ability of color measurement, they are inferior to physical measurement methods based on spectroscopy.

[0006] In summary, existing technologies struggle to simultaneously achieve color difference measurement accuracy comparable to professional spectrophotometers and comprehensive appearance defect detection while conducting online, full-width inspections, and to form an automated closed loop from inspection and judgment to sorting. Therefore, this paper proposes a machine vision-based color difference detection and sorting device and method for dyed fabrics to address the aforementioned issues. Summary of the Invention

[0007] This invention provides a machine vision-based color difference detection and sorting device and method for dyed fabrics, which solves the problems of insufficient detection accuracy and susceptibility to interference of single sensors, the inability to comprehensively evaluate appearance quality due to a single detection dimension, high false judgment rate, and disconnect between detection and sorting. By integrating hyperspectral imaging technology and color machine vision technology, it achieves multi-dimensional, high-precision, and quantitative online detection of the color and appearance of dyed fabrics, and performs automated sorting based on the fusion judgment results.

[0008] The solution of the present invention to solve the above technical problems is as follows: a color difference detection and sorting device for dyed fabric based on machine vision, including a flattening and traction module, an online detection module, a sorting execution module and a central control module, wherein the flattening and traction module flattens the fabric to be detected and pulls it through the detection area at a uniform speed;

[0009] The online detection module includes a hyperspectral imaging unit and a color area array imaging unit. The hyperspectral imaging unit acquires the reflectance spectrum information of the fabric surface, and the color area array imaging unit acquires the color texture image of the fabric surface.

[0010] The sorting execution module guides the fabric to different storage areas according to the sorting instructions. The sorting execution module is located at the end of the fabric conveying path.

[0011] The central control module includes a color difference analysis submodule and a comprehensive judgment submodule. The central control module is electrically connected to the online detection module and the sorting execution module, respectively. The central control module receives and processes detection data, executes color difference analysis and comprehensive judgment algorithms, and generates and sends sorting instructions.

[0012] The usage method includes the following steps:

[0013] S1, Calibration stage: The central control module collects the reference data of the standard fabric sample under the hyperspectral imaging unit and the color area array imaging unit, and sets the judgment threshold for each quality level.

[0014] S2, during the actual online detection stage, the fabric to be inspected is transported through the flattening and traction module, and the hyperspectral imaging unit and color area array imaging unit of the online detection module are simultaneously triggered to collect data.

[0015] S3, Data Analysis Stage: The central control module calculates the spectral color difference diagram and ΔE statistical value of the current fabric, and analyzes the texture and color uniformity in the color image.

[0016] S4, Fusion and Judgment Stage: Combining color difference statistics and image analysis results, the quality level of the fabric in the current detection unit is determined according to preset logic.

[0017] S5, sorting execution stage: The central control module converts the grade judgment result into a control command, driving the sorting execution module to complete the sorting action when the fabric reaches the corresponding position.

[0018] S6, the data recording stage, stores the test data, judgment results and images of each piece of fabric, and generates a quality report.

[0019] Based on the above technical solution, the present invention can be further improved as follows.

[0020] Furthermore, the online detection module also includes an integrated multi-source illumination unit. This illumination unit provides a uniform and stable lighting environment that meets standard color difference observation conditions, and simultaneously illuminates the hyperspectral imaging unit and the color area array imaging unit. Both the hyperspectral imaging unit and the color area array imaging unit are positioned above the detection area. By providing illumination with a high color rendering index, a specific color temperature (such as a simulated D65 standard light source), and uniform and stable illumination, the color of the fabric being tested can be accurately presented, eliminating the influence of ambient light fluctuations on the detection results. This provides reliable input conditions for hyperspectral imaging and color imaging, making the results of tests conducted at different times and in different batches comparable, and ensuring the objectivity and repeatability of the color difference measurement data.

[0021] Furthermore, the sorting execution module includes a programmable logic controller and a robotic arm. The robotic arm moves the fabric at the corresponding position to the designated sorting channel according to the instructions of the central control module. The robotic arm is highly flexible and can adapt to more complex sorting layouts or action requirements.

[0022] Furthermore, the color difference analysis submodule is used to calculate the ΔE value of each point on the fabric and the standard sample in the CIELAB color space based on hyperspectral data; the comprehensive judgment submodule is used to integrate the distribution statistical characteristics of the ΔE value with the texture anomaly characteristics of color image analysis to generate a comprehensive quality score and grade code, realizing the optimization from single index judgment to multi-dimensional comprehensive intelligent evaluation. It not only focuses on whether the overall color is qualified (average ΔE), but also on the uniformity of color (ΔE standard deviation, distribution), and combines texture defect information. This integrated judgment method greatly simulates the comprehensive judgment logic of experienced quality inspectors, significantly improves the scientificity, accuracy and reliability of grade judgment, and avoids misjudgment caused by relying solely on color difference or solely on images.

[0023] Furthermore, the judgment logic executed by the comprehensive judgment submodule is as follows: when the average ΔE of the fabric is less than the first threshold and there are no significant texture defects, it is judged as Grade A; when the average ΔE is between the first and second thresholds, or there are slight local color differences / slight texture defects, it is judged as Grade B; when the average ΔE is greater than the second threshold, or there are serious color difference spots, stains, or weaving defects, it is judged as Grade C. This reflects the principle of giving equal importance to color and appearance. For example, even if the overall color difference is acceptable (the average ΔE does not exceed the standard), but there are obvious stains, it will be downgraded; conversely, slight local defects but excellent overall color can still obtain a higher grade. This logic makes the sorting results more in line with actual quality requirements, improving the rationality of product grading and customer satisfaction.

[0024] Furthermore, the fusion judgment in step S4 adopts a weighted decision or machine learning model. The input features include the average ΔE, standard deviation, maximum value, area ratio of color difference region, as well as image texture consistency index and defect pixel ratio. The weighted decision model can quickly adapt to the quality standards of different customers or products by adjusting the weights, while the machine learning model can automatically discover the deep correlation between complex features and the final quality level by learning a large amount of historical data, forming a more accurate judgment model that goes beyond simple rules. This enables the system to have the ability to continuously optimize and learn autonomously, and to cope with various complex and ever-changing defect types.

[0025] The beneficial effects of this invention are as follows: This invention provides a machine vision-based color difference detection and sorting device and method for dyed fabrics, which has the following advantages:

[0026] 1. It features high precision and stability, based on hyperspectral color difference analysis, providing industry-leading measurement accuracy traceable to spectral standards. Combined with an integrated standard light source illumination unit, it ensures consistency of testing conditions and stability of results under different times and batches, making the data authoritative and reliable.

[0027] 2. With comprehensive and integrated testing dimensions, the system combines "laboratory-level color measurement" and "factory-level appearance inspection" capabilities, breaking the functional boundaries of traditional equipment. This multi-dimensional testing capability enables it to more accurately reflect the overall quality level of products and meet the dual requirements of high-end customers for color and appearance.

[0028] 3. The system achieves intelligent and highly reliable decision-making. It goes beyond simple pass / fail judgment and can perform multi-level (A / B / C) fine sorting. By integrating judgment logic, the system can weigh the severity of color deviation and physical defects like an expert and make more reasonable decisions that are closer to advanced human experience, which significantly improves the rationality of sorting results and customer satisfaction.

[0029] 4. It can effectively improve production efficiency. From fabric flattening and traction, synchronous detection, intelligent judgment to automatic sorting, the entire process does not require manual intervention, realizing true unmanned operation. This not only liberates workers from heavy and subjective inspection work, but also perfectly matches the inspection speed with the production line rhythm, greatly improving the overall production efficiency and the consistency of quality control.

[0030] 5. The process is data-driven and traceable. The system automatically records and links the original data, process images, judgment results and sorting logs of each piece of fabric, generating a complete digital quality archive. This provides strong data support for production process retrospection, quality responsibility definition, production parameter optimization and providing customers with objective quality proof, helping enterprises achieve digital and refined management.

[0031] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings.

[0032] The attached diagram lists the components represented by each number as follows:

[0033] 1. Flattening and traction module; 2. Online detection module; 201. Hyperspectral imaging unit; 202. Color area array imaging unit; 203. Multi-source illumination unit; 3. Sorting execution module; 301. Programmable logic controller; 302. Robotic arm; 4. Central control module. Attached Figure Description

[0034] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0035] Figure 1 This is a schematic diagram of a machine vision-based dyed fabric color difference detection and sorting device and method according to an embodiment of the present invention.

[0036] Figure 2 A front view of a machine vision-based dyed fabric color difference detection and sorting device and method according to an embodiment of the present invention;

[0037] Figure 3 This is a flowchart of a machine vision-based color difference detection and sorting device and method for dyed fabrics, provided as an embodiment of the present invention. Detailed Implementation

[0038] The following is in conjunction with the appendix Figure 1-3The principles and features of the present invention are described below. The examples given are for illustrative purposes only and are not intended to limit the scope of the invention. The invention is described more specifically in the following paragraphs by way of example with reference to the accompanying drawings. The advantages and features of the invention will become clearer from the following description. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0040] like Figure 1-2 As shown, the present invention provides a machine vision-based dyed fabric color difference detection and sorting device and method, including a flattening and traction module 1, an online detection module 2, a sorting execution module 3 and a central control module 4 (industrial computer), and each component is connected to each module through an industrial network.

[0041] The flattening and traction module 1 consists of a servo drive system, a tension sensor, a correction device, and multiple guide rollers to ensure that the fabric to be inspected passes through the inspection area smoothly, without wrinkles, and without deviation at a set speed (e.g., adjustable from 0 to 50 meters per minute).

[0042] The online detection module 2 is encapsulated in a light-shielding dark box, and an illumination unit, a hyperspectral imaging unit, and a hyperspectral imaging unit are integrated at the top inside.

[0043] Lighting unit: Uses high color rendering index (CRI>95) LED light sources arranged in a ring, forming a uniform and stable D65 standard light source lighting environment through a diffuser plate;

[0044] Hyperspectral imaging unit 201: Uses a pushbroom hyperspectral camera with a spectral range covering the 400-700nm visible light band;

[0045] Color area array imaging unit 202: Uses a high-resolution industrial area array CCD or CMOS camera, equipped with a telecentric lens to eliminate perspective error;

[0046] The hyperspectral camera and the color camera are connected to the central control module 4 via a hardware trigger line to achieve strictly synchronized acquisition and ensure that data is acquired from the same cross section of the fabric.

[0047] The sorting execution module 3 is located at the end of the production line. In this embodiment, a robotic arm controlled by a programmable logic controller (PLC) is used. According to the fabric grade (A / B / C), the corresponding robotic arm action puts the fabric roll into the corresponding conveyor belt and enters the qualified product area, downgraded product area or defective product area.

[0048] like Figure 3 As shown, the specific working principle and usage method of this invention are as follows:

[0049] S1: Calibration stage. After the system starts, it automatically performs spectral calibration: a standard white board is placed at the workstation, a hyperspectral camera is used to scan it, and a mapping relationship between DN value and absolute reflectance is established. Then, color calibration is performed: an X-RiteColorChecker color card is photographed, and a conversion model from the camera RGB space to the standard CIELAB space is established through a multinomial regression algorithm.

[0050] Finally, the operator places the customer-confirmed Grade A standard sample fabric into the system, which collects its "standard spectral fingerprint" and "standard appearance template image," and sets the judgment threshold for this batch (e.g., Grade A: ΔE00 mean < 1.5, and no defects; Grade B: 1.5 ≤ ΔE00 mean < 3.0, or small area of ​​minor defects; Grade C: ΔE00 mean ≥ 3.0, or serious defects).

[0051] S2: Online detection and real-time processing;

[0052] Production begins, the fabric passes through at a constant speed, and every 200 pulses from the encoder (corresponding to 20mm of fabric), the central controller simultaneously triggers two cameras;

[0053] Spectral data processing thread: performs dark current correction and reflectance calculation on hyperspectral line scan data, converts the spectral data of each pixel to CIELAB space using a pre-calibrated model, calculates the ΔE00 value of each pixel relative to the "standard spectral fingerprint" according to the CIEDE2000 formula, updates and displays the color difference cloud map of the current frame in real time, and calculates the mean, maximum value, standard deviation, etc.

[0054] S3, Data Analysis Stage, Image Processing Thread: After white balance correction and filtering of the color image, it is input into the pre-trained lightweight YOLOv5 defect detection neural network model. This model can identify and select categories such as "stain", "color spot", "warp stripe", "weft stripe", and "normal", and calculate the total defect area and the maximum defect size at the same time.

[0055] S4, Fusion Judgment Stage: After each logical unit (e.g., 1 meter long fabric) is inspected, the fusion judgment submodule is started. It combines the color difference statistical features (average ΔE00, over-standard area ratio, etc.) and image features (total defect area, most severe defect type, etc.) of all frames in the unit into a feature vector.

[0056] This embodiment uses the Gradient Boosting Decision Tree (GBDT) model for final judgment. This model has been trained with thousands of sets of historical data and can learn the nonlinear relationship between complex features and the final human judgment level (A / B / C), outputting the probability of the comprehensive level. The system takes the level with the highest probability as the judgment result of the 1-meter cloth segment.

[0057] S5, sorting execution stage: When a whole roll of fabric (e.g., 50 meters) has been inspected, the system will give the final grade of the roll of fabric based on the judgment of all the meter segments according to preset rules (e.g., "one vote veto" or "proportion limit").

[0058] The final grade, length, and other information are sent to the PLC. When the end of the fabric passes the SICK photoelectric sensor installed on the sorting table, the PLC triggers the robotic arm to complete the precise sorting.

[0059] S6, Data Recording Stage: Throughout the entire process, the system stores the original hyperspectral data (compressed storage), color difference map and defect annotation map of key frames, all judgment logs, and sorting action records corresponding to each piece of fabric into the MySQL database, using the unique roll number as the primary key.

[0060] Operators can query quality reports for any batch, shift, or product type through a web interface. The reports can be automatically generated in PDF format and include quality trend charts and Pareto charts of defect distribution.

[0061] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Content not described in detail in this specification is prior art known to those skilled in the art.

[0062] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Those skilled in the art can readily implement the present invention based on the accompanying drawings and the above description. However, any modifications, alterations, or variations made by those skilled in the art without departing from the scope of the present invention, utilizing the disclosed technical content, are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. A machine vision-based dyed fabric color difference detection and sorting device, comprising a flattening and traction module (1), an online detection module (2), a sorting execution module (3), and a central control module (4), characterized in that: The flattening and traction module (1) flattens the fabric to be tested and pulls it through the testing area at a uniform speed; The online detection module (2) includes a hyperspectral imaging unit (201) and a color area array imaging unit (202). The hyperspectral imaging unit (201) acquires the reflectance spectrum information of the fabric surface, and the color area array imaging unit (202) acquires the color texture image of the fabric surface. The sorting execution module (3) guides the fabric to different storage areas according to the sorting instructions. The sorting execution module (3) is located at the end of the fabric conveying path. The central control module (4) includes a color difference analysis submodule and a comprehensive judgment submodule. The central control module (4) is electrically connected to the online detection module (2) and the sorting execution module (3) respectively. The central control module (4) receives and processes the detection data, executes the color difference analysis and comprehensive judgment algorithm, and generates and sends sorting instructions. The usage method includes the following steps: S1, the central control module (4) collects the reference data of the standard fabric sample under the hyperspectral imaging unit (201) and the color array imaging unit (202), and sets the judgment threshold for each quality level; S2, the fabric to be inspected is transported through the flattening and traction module (1), and the hyperspectral imaging unit (201) and color array imaging unit (202) of the online detection module (2) are simultaneously triggered to collect data; S3, the central control module (4) calculates the spectral color difference diagram and ΔE statistical value of the current fabric, and analyzes the texture and color uniformity in the color image at the same time; S4, combining color difference statistics and image analysis results, determines the quality level of the fabric in the current detection unit according to preset logic; S5, the central control module (4) converts the grade determination result into a control command, driving the sorting execution module (3) to complete the sorting action when the fabric reaches the corresponding position; S6 stores the inspection data, judgment results, and images for each piece of fabric and generates a quality report.

2. The machine vision-based color difference detection and sorting equipment for dyed fabrics according to claim 1, characterized in that, The online detection module (2) also includes an integrated multi-source illumination unit (203), which provides a uniform and stable illumination environment that meets the standard color difference observation conditions, and simultaneously provides illumination for the hyperspectral imaging unit (201) and the color area array imaging unit (202), both of which are located above the detection area.

3. The machine vision-based color difference detection and sorting equipment for dyed fabrics according to claim 1, characterized in that, The sorting execution module (3) includes a programmable logic controller (301) and a robotic arm (302). The robotic arm moves the fabric at the corresponding position to the designated sorting channel according to the instructions of the central control module (4).

4. The machine vision-based color difference detection and sorting equipment for dyed fabrics according to claim 1, characterized in that, The color difference analysis submodule is used to calculate the ΔE value of each point on the fabric and the standard sample in the CIELAB color space based on hyperspectral data; the comprehensive judgment submodule is used to integrate the distribution statistical characteristics of the ΔE value with the texture anomaly characteristics of the color image analysis to generate a comprehensive quality score and grade code.

5. The machine vision-based color difference detection and sorting device for dyed fabrics according to claim 4, characterized in that, The judgment logic executed by the comprehensive judgment submodule is as follows: when the average value of ΔE of the fabric is less than the first threshold and there are no significant texture defects, it is judged as Grade A; when the average value of ΔE is between the first and second thresholds, or there are local slight color differences / slight texture defects, it is judged as Grade B; when the average value of ΔE is greater than the second threshold, or there are serious color difference spots, stains, or weaving defects, it is judged as Grade C.

6. The machine vision-based color difference detection and sorting equipment for dyed fabrics according to claim 1, characterized in that, The fusion determination in step S4 adopts a weighted decision or machine learning model. The input features include the average value, standard deviation, maximum value, area ratio of color difference region, as well as image texture consistency index and defect pixel ratio.

Images