Textile fabric color difference detection equipment

Abstract

The invention belongs to the technical field of textile fabrics, and particularly relates to textile fabric color difference detection equipment, which comprises an image acquisition module, an image processing and feature calculation module, an intelligent judgment and learning module and a man-machine interaction and report module, and is characterized in that the image acquisition module comprises a standard light source and an area array imaging colorimeter; the image processing and feature calculation module is configured to execute an algorithm of a textile fabric color difference detection method, the intelligent judgment and learning module stores and operates a hierarchical judgment tree model, manages an online optimization process, and obtains a color difference detection result through combination of area collection of an area array imaging colorimeter and a self-adaptive matching algorithm based on texture alignment degree. The problem that in traditional point measurement, points replace faces, and space uniformity cannot be evaluated is solved, overall color space information of the cloth is obtained through face domain imaging, interference of inherent textures of the fabric can be effectively stripped through a self-adaptive matching algorithm, and the generated color difference image purely reflects the dyeing difference.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of textile fabrics, and particularly relates to a textile fabric color difference detection equipment. BACKGROUND

[0002] In the textile printing and dyeing industry, color difference control is the core of quality management. At present, objective detection mainly relies on a contact type spectrophotometer, which calculates the difference (Delta E) with a standard color by measuring the colorimetric value of several points on the cloth. However, this method has a principle defect: Point-to-surface measurement blind area: point measurement can only reflect the local average colorimetric value, and cannot capture the spatial dimension defects such as color spots, yin and yang colors, gradual strip marks and local pollution that exist on the cloth surface. This leads to a serious deviation between the instrument data (Delta E qualified) and the quality inspector's visual judgment (appearance unqualified), which becomes the technical root of trade disputes.

[0003] Evaluation dimension missing: the existing technology can only quantify the color deviation, and lacks effective objective evaluation indexes for the color uniformity that determines the appearance grade of the cloth, and cannot distinguish uniform color deviation from color spots that are annoying.

[0004] No diagnostic ability: when determining unqualified, it cannot provide visual and quantitative information such as spatial distribution, shape and severity of defects, and it is difficult to guide the production link to carry out accurate process optimization. SUMMARY

[0005] The application aims to provide a textile fabric color difference detection equipment and method to solve the above technical problems.

[0006] To achieve the above purpose, the specific technical scheme of a textile fabric color difference detection equipment of the application is as follows: A textile fabric color difference detection equipment, comprising an image acquisition module, an image processing and feature calculation module, an intelligent judgment and learning module and a man-machine interaction and report module, the image acquisition module comprising a standard light source and a surface array imaging colorimeter, the image processing and feature calculation module being configured to execute the algorithm of the textile fabric color difference detection method, the intelligent judgment and learning module storing and running a hierarchical decision tree model and managing an online optimization process, and the man-machine interaction and report module being used for parameter setting, result display and report export.

[0007] Further, a textile fabric color difference detection method, comprising the steps of: S1. Standard color digital model construction: acquiring the surface area image of a standard cloth sample under a standard light source, and establishing a digital model containing pixel-level colorimetric value and texture statistical features after correction.

[0008] S2. Collecting the information of the sample area to be tested: Collect the image of the sample to be tested under the same conditions, and correct to generate the LAB color image to be tested.

[0009] S3. Multi-dimensional color difference feature extraction: the extracted feature vector; S4. Intelligent comprehensive judgment and diagnosis: input the feature vector extracted in step S3 into a preset hierarchical decision tree model. The model simulates the logic of quality inspection experts, and sequentially judges the overall color deviation, color uniformity, and local significant defects, and outputs the grade judgment and related process diagnosis suggestions. The system supports online optimization mechanism based on feedback of judgment parameters.

[0010] S5. Visual diagnosis report generation: output the visual report containing pseudo-color color difference cloud map, feature data, judgment conclusion and process suggestion.

[0011] Further, the method steps of step S3 multi-dimensional color difference feature extraction are: S3.1 Perception-optimized color difference map generation: an adaptive matching algorithm based on texture alignment degree is used to calculate the pixel-level color difference between the image to be tested and the standard model, and a color difference distribution matrix free from texture interference is generated .

[0012] S3.2 Global and statistical distribution feature calculation: the mean , standard deviation of , and further the skewness and kurtosis of the histogram are calculated to finely describe the color difference distribution pattern.

[0013] S3.3 Visual significant defect quantification: fuse color difference intensity and visual significance model to identify defect regions; extract shape-color joint descriptor for each region; and calculate the spatial aggregation index of all defect regions.

[0014] Further, the spatial aggregation index of the defect region is the global Moran index.

[0015] Further, a textile fabric color difference detection device also includes a system for executing the textile fabric color difference detection method; a device for continuously conveying the fabric on the production line; and a synchronization control device for triggering the image acquisition module to continuously or periodically shoot during the conveying process.

[0016] The advantages of the present application are: 1. By "area acquisition of area array imaging colorimeter" combined with "adaptive matching algorithm based on texture alignment", the problem of traditional point measurement "point instead of area" and inability to evaluate spatial uniformity is solved. The color space information of the whole fabric is obtained by area imaging, and the adaptive matching algorithm can effectively remove the interference of the inherent texture of the fabric, so that the generated color difference map purely reflects the dyeing difference.

[0017] 2. By "multi-dimensional color difference feature extraction (including skewness, kurtosis, and spatial aggregation index)" combined with "hierarchical decision tree model", the problem of single evaluation dimension and inability to intelligently diagnose in the prior art is solved. Skewness and kurtosis can finely depict the shape of color difference distribution and distinguish uniform color deviation from objectionable color spots. Spatial aggregation index (such as global Moran's index) can determine whether the defect is diffusely distributed or aggregated.

[0018] 3. By "multi-scale visual significant defect fusion detection algorithm" combined with "feedback-based decision parameter online optimization mechanism", the problems of complex defect missed detection and insufficient system adaptability are solved. The multi-scale fusion algorithm ensures that defects of different sizes and contrasts (from large area gradual change to small stains) can be effectively captured, improving the accuracy of detection. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a schematic diagram of the overall structure of the present application; Figure 2 is a flowchart of the textile fabric color difference detection method of the present application; Figure 3 is a flowchart of the multi-dimensional color difference feature extraction method of the present application; DETAILED DESCRIPTION The technical solutions of the present application will be described below in conjunction with the drawings. Obviously, the described embodiments are part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0020] In the description of the present application, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like indicate the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore it cannot be understood as a limitation on the present application. In addition, the terms "first", "second", "third" are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0021] Example 1 As shown in Figure 1 , a textile fabric color difference detection device, comprising an image acquisition module, an image processing and feature calculation module, an intelligent decision and learning module, and a man-machine interaction and report module, the image acquisition module comprises a standard light source and a plane array imaging colorimeter, the image processing and feature calculation module is configured to execute the algorithm of the textile fabric color difference detection method, the intelligent decision and learning module stores and runs the hierarchical decision tree model, and manages the online optimization process, the man-machine interaction and report module is used for parameter setting, result display and report export.

[0022] Among them, there is also a system for executing the textile fabric color difference detection method; a device for continuously conveying cloth on the production line; a synchronous control device for triggering the image acquisition module to take continuous or timed pictures during the conveying process.

[0023] Embodiment 2 As shown in Figure 2 , a textile fabric color difference detection method, comprising the steps of: S1. Standard color digital model construction: Collect the surface area image of the standard cloth sample under the standard light source, and establish a digital model containing pixel-level colorimetric value and texture statistical characteristics after correction.

[0024] S2. Information acquisition of sample surface area to be measured: Collect the image of the sample to be measured under the same conditions, and correct to generate the LAB colorimetric image to be measured.

[0025] S3. Multi-dimensional color difference feature extraction: The extracted feature vector is S4. Intelligent comprehensive decision and diagnosis: The feature vector extracted in step S3 is input into a preset hierarchical decision tree model. The model simulates the logic of quality inspection experts, and sequentially judges the overall color deviation, color uniformity, and local significant defects, and outputs the grade judgment and the associated process diagnosis suggestion. The system supports the online optimization mechanism of the decision parameters based on feedback.

[0026] S5. Visual diagnosis report generation: Output the visual report containing pseudo-color color difference cloud map, feature data, decision conclusion and process suggestion.

[0027] Embodiment 3 As shown in Figure 3 , the method steps of the step S3 multi-dimensional color difference feature extraction are: S3.1 Perception optimization color difference map generation: An adaptive matching algorithm based on texture alignment degree is used to calculate the pixel-level color difference between the image to be measured and the standard model, and generate a color difference distribution matrix free from texture interference .

[0028] S3.2 Global and statistical distribution feature calculation: Calculate the mean value of the color difference , the standard deviation , and further calculate the skewness and kurtosis of the histogram of the color difference to finely describe the distribution pattern of the color difference.

[0029] S3.3 Visual significant flaw quantification: fuse the color difference intensity and visual saliency model to identify the flaw region; extract the shape-chroma joint descriptor for each region; and calculate the spatial aggregation index of all flaw regions.

[0030] The spatial aggregation index of the flaw region is the global Moran's I.

[0031] Embodiment 4 The adaptive color difference matching algorithm based on texture alignment degree in step S3.1: Input: gray image of the sample to be tested , gray reference image of the standard sample and global average chroma value .

[0032] Texture alignment degree evaluation: Divide and into several non-overlapping local blocks (such as 32x32 pixels) respectively.

[0033] For each pair of spatial position corresponding blocks, calculate the normalized cross-correlation coefficient (NCC).

[0034] Calculate the average value of NCC of all block pairs .

[0035] Matching mode decision and execution: Set the texture alignment threshold θ (for example, θ = 0.75).

[0036] If ≧ θ, it is determined that the texture is highly alignable, and the "local reference window mode" is adopted. For each pixel point (m, n) in the sample image, take a WxW neighborhood window (such as W = 15) at the same position in the standard reference image, calculate the average value of the standard chroma of all pixels in the window as the colorimetric reference of the point.

[0037] Otherwise, the "global average mode" is adopted. All pixels are compared with the global average value of the standard color .

[0038] Output: based on the selected mode, calculate the CIEDE2000 color difference of each pixel, and generate the de-textured color difference distribution matrix .

[0039] Example 5 The multi-scale visual saliency fusion detection algorithm in step S3.3 is as follows: Input: LAB image of the sample to be tested And its color difference matrix D.

[0040] Construction of the Gaussian Pyramid: On The a* and b* channels are used to construct L-level Gaussian pyramids (e.g., L=4).

[0041] Multi-scale feature map generation: at each layer l: calculate The standard deviation of the a* and b* values ​​of each pixel in its 5×5 neighborhood is used to generate the color contrast map of this layer. .

[0042] Downsample the global color difference matrix D to the same level as... Same size, get .

[0043] Calculate the saliency plot of this layer. (⊙ indicates element-wise multiplication).

[0044] Cross-scale fusion: Each layer Upsampled to the original image resolution, denoted as .

[0045] Final defect saliency map The result is obtained by weighted summation of the upsampled images from each layer: The lower layers (high resolution) have a higher weight.

[0046] Defect area extraction: Binarization is performed using an adaptive threshold, followed by connected component analysis to identify independent defect regions. And calculate the morphology-chromaticity joint descriptor for each region (such as area, circularity, and region average). .

[0047] Example 6 Among them, the hierarchical decision tree model and its online optimization mechanism in step S4 are as follows: Model structure definition: Root node: Determine >2.0? If yes, determine "severe color cast"; if not, proceed to the next level.

[0048] Second-level node: Judgment <0.5? If yes, proceed to the "Uniform" branch; if no, combine skewness and kurtosis to determine "Diffuse Color Variation" or proceed to the "Accumulated Defects" branch.

[0049] Third-level nodes (clustered branches): combined (Maximum defect area) and The (aggregation index) is judged as "major aggregation defect" or "minor color variation".

[0050] Each leaf node is associated with specific process diagnostic recommendations.

[0051] Online optimization: The system maintains a conflict sample pool, storing the feature vector F of samples whose manual review and system judgment are inconsistent. i With artificial label Y i .

[0052] When the number of samples accumulates to a set value N (e.g., 100), the optimization process is triggered: Threshold fine-tuning: targeting key thresholds in the decision tree (such as...) The threshold is set to 0.5. The coordinate descent method is used to search within a reasonable neighborhood, and the threshold is adjusted with the goal of minimizing the number of misclassified samples.

[0053] Or arbitrator training: with (F i Y i Using the training set, a lightweight random forest classifier is trained as an auxiliary arbitrator to provide a reference or trigger manual review when the system's decision is uncertain.

[0054] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A color difference detection device for textile fabrics, characterized in that, It includes an image acquisition module, an image processing and feature calculation module, an intelligent decision and learning module, and a human-computer interaction and reporting module. The image acquisition module includes a standard light source and an area array imaging colorimeter. The image processing and feature calculation module is configured with an algorithm for executing the color difference detection method for textile fabrics. The intelligent decision and learning module stores and runs a hierarchical decision tree model and manages the online optimization process. The human-computer interaction and reporting module is used for parameter setting, result display, and report export.

2. A method for detecting color difference in textile fabrics, characterized in that, Including the following steps: S1. Construction of Standard Color Digital Model: Collect surface images of standard fabric samples under standard light sources, and after correction, establish a digital model containing pixel-level chromaticity values ​​and texture statistical features. S2. Acquisition of surface area information of the sample to be tested: Acquire images of the sample to be tested under the same conditions, and generate the LAB chromaticity image to be tested after correction. S3. Multi-dimensional color difference feature extraction: Extracted feature vector; S4. Intelligent Comprehensive Decision and Diagnosis: The feature vector extracted in step S3 is input into a pre-defined hierarchical decision tree model. This model simulates the logic of a quality inspection expert, sequentially judging overall color deviation, color uniformity, and local significant defects, and outputting the level judgment and associated process diagnosis suggestions. The system supports an online optimization mechanism for decision parameters based on feedback. S5. Visualized Diagnostic Report Generation: Outputs a visualized report containing a pseudo-color color difference cloud map, feature data, judgment conclusions, and process recommendations.

3. The method for detecting color difference in textile fabrics according to claim 2, characterized in that: The steps for multi-dimensional color difference feature extraction in step S3 are as follows: S3.1 Perceptual Optimization Color Difference Map Generation: An adaptive matching algorithm based on texture alignment is used to calculate the pixel-level color difference between the image under test and the standard model, generating a texture-free color difference distribution. . S3.2 Calculation of Global and Statistical Distribution Characteristics: Calculation mean Standard deviation And further calculate the skewness of its histogram. With kurtosis To provide a detailed description of the color difference distribution pattern. S3.3 Visually Significant Defect Quantification: Integrating color difference intensity and visual saliency models to identify defect areas; Extract the morphology-chromaticity joint descriptor for each region; and calculate the spatial clustering index of all defective regions.

4. The method for detecting color difference in textile fabrics according to claim 3, characterized in that: The spatial clustering index of the defective region is the global Moran index.

5. A color difference detection device for textile fabrics, characterized in that: It also includes a system for performing the textile fabric color difference detection method according to any one of claims 2-4; and an apparatus for continuously conveying fabrics on a production line. A synchronization control device for triggering the image acquisition module to perform continuous or timed shooting during transmission.

Images